Paid subscribers now get first access to something I’ve been developing for the past year: a complete course on systematic AI integration for content professionals and technical writers.

This isn’t just about casual ChatGPT tips or prompt hacks. It’s about building the operational framework you need to integrate AI strategically into your actual work …the kind that makes you more effective without replacing your judgment or expertise.

I’m making the beta version available to you paid subscribers, before the polished video course launches publicly. Here’s what that means and why it might interest you.

Subscribe now

What This Course Actually Is

Writing with Machines teaches content professionals how to move from casual AI use to systematic integration. It’s built around a core insight: AI works best when you can articulate your process clearly enough to identify exactly where it helps.

The course covers ten chapters:

1. Understanding Your Workflow - Making your invisible intellectual process visible and identifying friction points

2. What AI Actually Does - Pattern-matching vs. reasoning, and why the distinction matters

3. The Five Information Types - Structuring content so AI can process it effectively

4. Grounding AI in Knowledge - Building reliable knowledge bases instead of hoping for accurate responses

5. Prompt Design - Moving beyond trial-and-error to systematic prompt design

6. Pairing Prompts with Process - Matching specific prompts to specific workflow stages

7. Style and Temperature - Controlling AI output characteristics deliberately

8. The Structured Principles - How content operations thinking improves AI integration

9. Building Your Prompt Taxonomy - Organizing prompts as operational assets

10. Workflow Redesign - The capstone where you transform an actual process using everything you’ve learned

This isn’t just theory. Each chapter includes practical exercises, templates you can adapt, and frameworks you’ll use immediately. Chapter 10 has you redesign a real workflow from your work. That’s your deliverable.

What Beta Access Means

The beta course is complete. All ten chapters are written and functional. But it’s developmental. It’s text-based with exercises and templates rather than polished video lessons. You’re getting the systematic framework and practical tools, not production value.

Here’s what you get as a beta participant:

The exchange is simple: You get early access to systematic AI integration frameworks. I get insights that make the final course better. Both of us benefit.

Why Beta Access Now

I could wait until everything is polished and perfect before releasing anything. But that’s not how I work, and frankly, it’s not how good courses get built.

The best educational materials come from real interaction with real learners.

You’re professionals doing actual content work. You have real workflows, real constraints, real stakeholders. Your friction points are different. Your applications will be different. Your feedback will be invaluable.

Also, I’m learning things from my research and teaching this semester that are making the course better week by week. The disciplinary differences I wrote about in this week’s newsletter? That came from watching students map their workflows. Those insights are already improving how I teach the concepts.

If you wait for the polished version, you pay more and you miss the opportunity to shape it. If you join the beta, you get the systematic framework now (which you can use immediately) and you influence what the premium version becomes.

Who This Course Is For

What About Cost?

If you’re already paying for Cyborgs Writing. This beta access is included in your subscription. No additional payment required.

The premium video course, when it launches, will cost significantly more. Early beta participants will get preferred pricing on that, but right now, your subscription covers this beta access completely.

If cost is a barrier: I’m willing to provide access to people who can’t afford or prefer not to pay for the subscription but are genuinely committed to the process and feedback. If that’s you, email me at Lance.cummings@hey.com and we’ll work something out. The goal is to build an excellent course, not to restrict access unnecessarily.

Access beta form below.

Read more

Last semester, my students and I built a writing feedback chatbot for our technical communication course. In testing, it worked beautifully. Clear, specific feedback that maintained professional warmth. We deployed it.

Within two weeks, students started reporting inconsistent experiences. The same submission structure that earned detailed feedback on Monday produced superficial responses on Wednesday.

One student showed me screenshots. For example, the chatbot had praised her conclusion as “effectively synthesized” in the morning, then flagged the identical paragraph as “needing stronger connections” that afternoon. Same prompt. Same model version. Same student text.

This isn’t a bug. It’s just the way it is when working with language models. And it’s why prompt testing requires something more rigorous than “try it and see if it looks good.”

Evaluation as Craft

The ancient Greeks had a term for what we need: kritikē technē … Or the art of judgment. The word kritikē comes from krinō (to separate, to decide), and technē means a teachable craft. Together, they describe a disciplined practice for evaluating the worth, correctness, and fitness of language.

The grammarians who were a part of the library of Alexandria developed kritikē into a systematic discipline that created repeatable procedures for evaluating texts. Working with multiple manuscript copies of Homer, they faced a problem familiar to anyone testing AI outputs: variant versions of the same content, with no obvious way to determine which was best.

Their solution was to operationalize judgment. They established criteria, collected evidence across variants, applied standards consistently, and recorded their decisions so others could follow the reasoning. Judgment became a craft that could be taught, reproduced, and improved.

This is exactly what prompt testing requires. When we evaluate AI outputs, we’re not asking “does this sound good?” We’re asking whether the outputs meet specific criteria reliably enough for a particular purpose.

That question demands a method—articulated standards, systematic procedures, transparent documentation.

Why Structured Prompting Demands Systematic Testing

When people claim structured prompting is dead, they’re usually working with single interactions or more dialogic collaborations. Ask a question, get an answer, move on (or continue working on that one instance). In that context, casual prompting often works fine.

But the moment you’re building something that needs to perform reliably across users, sessions, and contexts, you’re no longer in single-interaction territory.

This could be:

• a classroom assistant

• a documentation helper, or

• a content generation workflow.

You’re building a system. And systems require consistency that casual prompting can’t guarantee.

My research on prompt format bears this out. When I tested the same complex task across four different structures, the outputs varied dramatically. Not just in efficiency (processing time ranged from 64 to 120 seconds) but in character. The unstructured prompt produced exploratory, wandering responses. JSON triggered mechanical, compliance-document prose. Natural structure with clear sections generated focused, efficient communication.

Each format created different statistical conditions for the model’s token prediction. JSON tokens co-occur with technical documentation patterns in training data, so generating JSON-formatted input increases the probability of formal, exhaustive output patterns. Unstructured conversational input co-occurs with exploratory discussion, so the model follows those statistical tendencies.

Structure gives you leverage over consistency, but only if you verify that your structure actually produces the consistency you need. And structure isn’t the only variable. Temperature settings, model selection, and context length all affect output character. (I explore temperature’s effects in a separate lesson.) Testing helps you understand how these variables interact for your specific use case.

Testing Prompts vs. Testing Code

When we test software, we’re verifying logical operations. Given input X, does the function return output Y? The relationship is deterministic. Run the test a thousand times, get the same result a thousand times.

Prompt testing operates on different principles. We’re examining rhetorical reliability, not verifying logic. Does the relationship we’ve established between human intention and machine interpretation remain stable across time, context, and varied inputs?

This distinction matters because it changes what we’re looking for. Code either passes or fails. Prompts exist on a spectrum of reliability, and our job is to understand where on that spectrum a given prompt sits for a given purpose.

The writing feedback chatbot didn’t “fail” in any binary sense. It produced plausible feedback every time. The question was whether that feedback remained consistent enough to be pedagogically useful … And whether students could trust that the evaluation criteria were being applied reliably rather than arbitrarily.

That’s a question of judgment, not logic. And answering it requires a method for judgment.

What We’re Judging

When you test a prompt systematically, you’re evaluating three aspects of the human-AI collaboration you’ve created.

Stance stability. Every prompt establishes a rhetorical stance, or a position from which the AI speaks. “You are a writing tutor who provides constructive feedback focused on argument structure and evidence use” isn’t just an instruction. It’s establishing a consistent voice and perspective.

But does that stance actually persist?

With our classroom chatbot, testing revealed that the constructive-tutor stance held firm for the first few exchanges in a conversation, then gradually drifted toward generic encouragement.

This could happen because the model’s training data contains patterns where tutoring interactions soften over time, or because earlier instructions lose influence as the context window fills with conversation.

Whatever the mechanism, the effect was measurable: stance drift under extended use. Testing helped us identify where drift occurred so we could add stabilizing elements—periodic reinforcement of the evaluative criteria, structural markers that maintained the rigorous-feedback pattern.

Interpretive framework reliability. Your prompt doesn’t just tell the AI what to do. It shapes how inputs get processed. When our chatbot prompt said “evaluate based on the technical communication rubric criteria,” we were creating conditions where rubric-related language would influence the output. But those conditions had gaps we didn’t anticipate.

The rubric worked well for standard assignments because the model had seen similar patterns. But when students submitted creative approaches, like an infographic, the statistical patterns broke down. The model couldn’t match rubric language to unfamiliar input formats, so it defaulted to surface-level observations about grammar and formatting. Testing with diverse input types revealed these blind spots. The fix wasn’t clarifying instructions—it was providing examples of the rubric applied to non-standard formats, giving the model patterns to match against.

Collaborative boundaries. Every prompt creates what I think of as a collaborative space, or the zone where human intention and machine capability overlap productively. Testing maps the edges of this space.

For the classroom chatbot, we needed to know: What types of student writing produce useful feedback? Where does the feedback quality drop off? What submission characteristics cause confusion or generic responses? Which edge cases does the prompt handle gracefully, and which break it entirely?

These boundaries aren’t obvious from the prompt text. They emerge only through running varied inputs through the system and observing where reliability holds and where it fractures.

Knowing what to judge is only half the challenge. The Alexandrian grammarians understood this. They didn’t just identify what made a text authentic or well-formed. They also developed procedures for making those judgments systematically: comparing variants, marking uncertainties, documenting reasoning.

Prompt testing requires both dimensions. We need criteria for evaluation—what counts as stable stance, reliable interpretation, appropriate boundaries. And we need procedures for applying those criteria.

The Rhetorical Appeals as Evaluation Criteria

The Alexandrian grammarians faced a problem we might recognize: they had no original to compare against. When scholars assessed a line of Homer, they weren’t checking it against some authoritative master copy—none existed. Homer was oral tradition committed to writing centuries after composition, and every manuscript was a copy of copies, each with its own variants and corruptions.

So how did these scholars develop criteria for judgment? By immersion in the corpus itself. They studied patterns across many manuscripts, inferring what Homeric diction typically looked like, identifying metrical conventions from the poems themselves, developing a sense of stylistic consistency through deep familiarity with the work. They’re standards emerged from the body of texts, then got applied back to evaluate individual passages.

We’re doing something similar with AI outputs. There’s no “ideal response” to compare against—just multiple outputs from which we infer what “good” looks like for a particular purpose. Our criteria emerge from examining what works, identifying patterns that characterize successful responses, and then applying those standards to evaluate new outputs.

But rhetoric offers a framework that accelerates this process: the three appeals. Aristotle identified ethos (credibility), pathos (emotional engagement), and logos (reasoning) as the fundamental dimensions of persuasive communication. These aren’t just persuasion techniques—they’re categories for evaluating whether communication works.

We’ve applied them to speeches, text, digital media … And now AI outputs.

When we adapt them for prompt testing, they become three lenses for examining output quality.

Ethos Testing: Can the Output Be Trusted?

Ethos in classical rhetoric establishes the speaker’s credibility and character. For AI outputs, we’re not assessing whether the model has credibility (it doesn’t, inherently), but whether the outputs are trustworthy enough for the intended purpose.

Trustworthiness breaks down into two components: consistency and accuracy.

Consistency asks whether the same prompt produces comparable outputs across multiple runs. This matters because inconsistent outputs can’t be trusted for systematic use. If a documentation prompt generates comprehensive coverage on one run and superficial summaries on the next, you can’t build a workflow around it.

Testing for consistency is straightforward: run the same prompt with the same input multiple times and compare the outputs. But “same” doesn’t mean identical. The question is whether variation falls within acceptable bounds for your purpose.

Consider a blog title generator. Testing the same article summary five times might produce five different titles—but if all five maintain brand voice, include relevant keywords, and target the right audience, that variation is a feature for brainstorming purposes. The prompt has sufficient ethos for generating options.

Contrast that with a product description prompt. If testing reveals 30% variation in which technical specifications get mentioned, the prompt lacks the consistency required for that task. Product descriptions need completeness, not creativity. The prompt would need explicit checklists and verification steps until testing shows reliable coverage of required elements.

Accuracy asks whether the outputs are factually correct and appropriately grounded. This is particularly critical for prompts that draw on domain knowledge or make claims that could be verified.

Testing for accuracy requires reference points—either human expert review or comparison against known-correct information. For our classroom chatbot, we tested accuracy by having instructors evaluate whether the AI’s feedback aligned with how they would assess the same submissions. Where the AI and instructors diverged significantly, we examined whether the prompt’s criteria were unclear or whether the model was introducing its own evaluation standards.

The ethos question for any prompt is: Can I trust this output enough to use it for its intended purpose? Testing answers that question with evidence rather than hope.

Pathos Testing: Is the Emotional Register Appropriate?

Pathos in classical rhetoric involves emotional appeal—engaging the audience’s feelings appropriately for the context. For AI outputs, we’re testing whether the tone and emotional register remain appropriate across different inputs and contexts.

This matters more than many practitioners realize. Tone inconsistency can undermine otherwise solid content. A customer service prompt that sounds helpful for simple questions but becomes condescending for complex ones will damage relationships regardless of how accurate the information is.

Imagine an automated feedback system for student writing. The prompt might maintain an encouraging tone when reviewing strong work but shift to patronizing reassurance for weaker submissions. Phrases like “You tried your best” and “Don’t worry, writing is hard” appearing only in responses to struggling students would unintentionally signal that the system had already judged them as less capable.

In this scenario, the prompt’s ethos could be fine—consistent, accurate feedback. But its pathos would be off, treating different students with different levels of respect based on submission quality.

Testing pathos requires diverse inputs that trigger different emotional contexts. For a feedback system, this means testing with:

• Strong submissions (does it avoid excessive praise that might seem hollow?)

• Weak submissions (does it maintain respect while identifying problems?)

• Frustrated student language (does it respond with patience rather than matching the frustration?)

• Confused questions (does it clarify without condescension?)

For a customer service prompt, you’d test across complaint types, customer tones, and issue severity. For documentation, you might test whether the prompt maintains appropriate professional distance when explaining both mundane features and exciting new capabilities.

The pathos question is: Does the emotional register remain appropriate across the full range of likely inputs? Testing reveals where tone calibration breaks down.

Logos Testing: Is the Reasoning Sound?

Logos in classical rhetoric involves logical argument, or the structure and validity of reasoning. For AI outputs, we’re testing whether the logical framework established in the prompt actually governs how outputs get generated.

This goes beyond checking factual accuracy (that’s ethos). Logos testing examines whether the prompt’s stated priorities, decision rules, and evaluation criteria actually shape the output—or whether they get overridden by other patterns in the model’s training.

Consider a documentation prompt that claims to prioritize accuracy but consistently chooses simpler explanations over precise ones. The prompt might include both “maintain technical accuracy” and “explain in accessible language.” In practice, accessibility could win out every tim—the AI sacrificing precision for readability without letting the user know.

This wouldn’t be a failure of the model. It would be a logical contradiction in the prompt that testing reveals. “Accurate and accessible” sounds reasonable until you encounter cases where accuracy requires technical precision that isn’t accessible. Without guidance for resolving that tension, the model could default to patterns from its training data, where accessible explanations are more common than technically precise ones.

Testing for logos means deliberately creating inputs that force your prompt’s priorities into conflict:

• If your prompt says “be concise but thorough,” test with topics that can’t be covered both concisely and thoroughly. Which wins?

• If your prompt prioritizes “user benefit” and “technical accuracy,” test with features where the accurate description doesn’t sound beneficial. What happens?

• If your prompt establishes an evaluation hierarchy (“first check X, then Y, then Z”), test with inputs where X and Y suggest different conclusions. Does the hierarchy hold?

The logos question is: When the prompt’s instructions compete, does the output resolve conflicts the way I intend? Testing surfaces hidden contradictions and reveals which instructions actually govern behavior.

Combining the Three Lenses

Most prompt testing requires all three lenses, but their relative weight depends on purpose.

For a research summarization prompt, logos dominates. You need the reasoning structure to govern output reliably. Ethos matters for accuracy, but pathos is less critical since emotional register in research summaries is relatively narrow.

For a customer-facing chatbot, pathos may matter most. Users will forgive minor inconsistencies or occasional reasoning gaps if the tone feels right. They won’t forgive condescension or inappropriate cheerfulness when they’re frustrated.

For a compliance documentation prompt, ethos is paramount. Consistency and accuracy are non-negotiable. Pathos and logos matter, but trustworthiness is the threshold requirement.

When designing your testing approach, identify which appeals are critical for your use case and weight your testing accordingly. A prompt can have strong ethos but weak pathos (consistent and accurate but tonally inappropriate), or strong logos but weak ethos (sound reasoning but inconsistent execution). Testing across all three reveals the full picture.

From Criteria to Procedures

Knowing what to evaluate doesn’t tell you how to evaluate it. The Alexandrian grammarians understood this. They developed not just standards for judgment but systematic procedures for applying those standards: methods for comparing variants, marking uncertainties, and documenting reasoning so others could follow or challenge their conclusions.

These procedures translate surprisingly well to prompt testing. The Alexandrians were solving a version of our problem: multiple variant texts, no definitive original, and the need for judgments that could be taught, reproduced, and defended.

Recension: Multi-Run Comparison

The Alexandrians framed this work as diorthōsis and ekdosis. They collated multiple manuscript “witnesses” to identify variants, marked doubtful lines, and recorded their comparative reasoning in commentaries. Rather than trusting any single copy, they corrected the text and then issued a stabilized edition, choosing readings based on consistent patterns across the evidence.

For prompt testing, this becomes multi-run comparison. Never evaluate a prompt based on a single output. Run the same prompt with the same input multiple times and compare results.

This sounds obvious, but it’s surprisingly rare in practice. Most prompt development follows a pattern: write prompt, test once, adjust if the output looks wrong, test once more, deploy. That’s like Alexandrians looking at one manuscript and declaring it authoritative.

Multi-run comparison reveals what single tests hide. When I tested my four prompt formats, I didn’t just run each once. Each variation ran under controlled conditions, with metrics tracked across runs. The patterns emerged from comparison, not from any single output.

For practical testing, I recommend a minimum of three runs for informal evaluation and five or more for anything you’ll deploy in production. Compare outputs looking for:

Structural consistency. Do the outputs follow the same organization? If your prompt specifies a format, does that format hold across runs, or does it drift?

Coverage variation. Do all runs address the same key points, or do some outputs omit information that others include? For the blog title generator, variation in titles is fine. For product descriptions, variation in which features get mentioned is a problem.

Tonal range. Do all outputs stay within the same emotional register, or do some runs produce noticeably different tones? This is your pathos check.

Priority adherence. When the prompt contains competing instructions, do all runs resolve the conflict the same way? This is your logos check.

The goal isn’t identical outputs. That’s neither possible nor desirable. The goal is understanding the range of variation your prompt produces and determining whether that range falls within acceptable bounds for your purpose.

Athetesis: Marking Uncertainty

When Alexandrian grammarians encountered lines they suspected were spurious or corrupted, they didn’t simply delete them. They marked them with an obelos—a horizontal line indicating doubt. This kept them visible in the text. Future scholars could see the judgment, assess the reasoning, and reach their own conclusions.

This practice, called athetesis, prioritized transparency over tidiness. A marked line told readers: “This is questionable, but I’m preserving it so you can evaluate my judgment.”

For prompt testing, this becomes uncertainty flagging. When you identify problems in AI outputs, mark them explicitly rather than silently fixing them or discarding the output entirely.

This matters for two reasons. First, patterns of uncertainty reveal prompt weaknesses. If you’re consistently flagging the same type of problem, you’ve identified where your prompt needs revision. Silent fixes hide these patterns.

Second, flagged outputs become training data for your own judgment. Over time, a collection of marked outputs teaches you (and your team) what to watch for. The Alexandrians built scholia (or commentary traditions) around their marked texts. You can build similar institutional knowledge around flagged AI outputs.

A simple flagging system might include markers like:

• H for hallucination (unsupported claims or fabricated details)

• T for tone problems (inappropriate emotional register)

• I for incompleteness (missing required elements)

• C for contradiction (conflicts with prompt instructions or internal inconsistency)

• D for drift (departure from established stance or format)

The specific markers matter less than consistent use. Pick a system and apply it across all your testing so patterns become visible.

Scholia: Documenting Your Reasoning

The Alexandrians didn’t just mark problems—they explained their judgments. Marginal notes called scholia documented why a line was suspect, what alternatives existed, and how the editor had reasoned through the decision. These annotations accumulated over generations, creating a scholarly conversation around the text.

For prompt testing, this becomes documented evaluation. Don’t just record that an output passed or failed—record why.

This is where most testing falls apart. Teams run prompts, glance at outputs, make a gut judgment, and move on. Nothing gets written down. A month later, no one remembers why certain prompt versions were rejected or what problems the current version was designed to solve.

Documented evaluation doesn’t require elaborate systems. A simple log capturing the following for each test proves valuable:

The input used. What specific content did you feed the prompt? Save it so tests can be reproduced.

The output received. Keep the full output, not just a summary or judgment.

Your assessment. Did it pass or fail on ethos, pathos, logos? What specific problems did you identify? Use your flagging system.

Your reasoning. Why did you judge it this way? What would have made it better? This is the scholia—the part that teaches future evaluators (including future you) how to think about the prompt’s performance.

When you revise a prompt based on testing, document what you changed and why. Link the revision to the specific test failures that motivated it. This creates a trail that makes prompt development cumulative rather than circular.

Putting Procedures into Practice

These Alexandrian procedures provide the methodological foundation. But you still need practical workflows for implementing them. The approach you choose depends on your technical resources and scale.

Read more

In our previous lessons, we explored how structured prompts can be woven into conversational interfaces through chatbots.

Now, let's take the next step in our PromptOps journey: creating AI agents and automated workflows that operate more independently within your content systems.

While chatbots are powerful tools, they still require your active involvement. You must open the interface, type your query, and wait for a response.

But what if your AI tools could operate more autonomously—taking action at specific times or in response to specific events?

Imagine you're a technical writer maintaining documentation for software that updates monthly. Currently, you manually update your documentation when new features are released.

What if instead, your system could automatically detect new release notes, analyze them, and draft updated documentation without your intervention?

Or perhaps you're an educator who's created a chatbot that provides feedback on student writing.

Rather than requiring students to copy and paste their work into a chat interface, what if they could simply submit their assignments to a shared folder, triggering an AI system to provide initial feedback before you review it?

This shift from reactive to proactive AI implementation represents a significant advancement in how we collaborate with machines, but raises an important question:

What if we could take the structured prompting techniques we've developed and apply them in systems that don't wait for our commands, but instead recognize when action is needed and respond accordingly?

Subscribe now

The Promise of AI Agents

This is where AI agents and automated workflows enter the picture. But let's be clear about what these terms actually mean, as they're often surrounded by unnecessary mystique:

AI Agent: At its simplest, an AI agent is just an AI system that can perform specific tasks with some degree of independence. It's not a sentient being or a complex autonomous system – it's essentially a chatbot with a job description and the ability to act when certain conditions are met rather than waiting for direct commands.

Automated Workflow: This is a series of connected steps where an AI agent works together with other digital tools to accomplish tasks with minimal human intervention. These workflows follow an "if this, then that" logic – when specific triggers occur, the workflow activates, performs its tasks, and delivers results.

Together, these concepts represent a natural progression in machine rhetorics:

1. We started with structured prompts that generate single responses.

2. We evolved to chatbots that maintain context across multiple exchanges.

3. Now we're developing agents that operate proactively within defined workflows.

The transition from chatbots to automation isn't about abandoning our conversation-based tools but extending them beyond the chat interface into our broader content operations. It's about taking the same principles of structured prompting that make chatbots effective and applying them in systems that can operate even when we're not actively engaged with them.

For content professionals, educators, and technical communicators, this evolution offers tremendous potential to streamline repetitive tasks while preserving your expertise for the work that truly requires human insight and creativity.

Consider these common scenarios that might resonate with your daily work:

You're a technical writer frustrated with how a summarization tool misses key points that matter to your product's users. The tool consistently overlooks details about accessibility features and security considerations that your readers specifically care about.

As a content creator, you're tired of a style checker that enforces rules that don't align with your publication's voice. It flags your intentional sentence fragments as errors and pushes for formal language when your audience responds better to a conversational tone.

Instead of abandoning these tools or accepting their limitations, you can rebuild them—creating versions that better align with your specific needs and preferences.

A technical writer might create a workflow that first runs content through a general summarization tool, then applies a custom filter that ensures accessibility and security information is preserved or even emphasized.

A content creator could build a system that applies house style rules after the general check, effectively overriding generic suggestions with publication-specific guidance.

This creative approach echoes a long tradition in writing and publishing:

Dissatisfied with the books being published? Write your own. Unhappy with information available on the web? Create your own content. Concerned about biased AI systems? Build alternatives that reflect your values.

The ability to design AI workflows gives you tremendous agency in shaping how technology serves your writing and content processes.

Cunning Intelligence in Action

The ancient Greeks had a concept called metis—often translated as "cunning intelligence" or "adaptive wisdom." This was the kind of practical intelligence embodied by figures like Odysseus, who used creativity and resourcefulness to overcome challenges.

Unlike theoretical knowledge (episteme) or technical skill (techne), metis represented the ability to adapt, improvise, and find clever solutions to complex problems.

When you design AI workflows, you're exercising a modern form of metis. You're not just accepting the tools and processes given to you; you're actively reshaping them to better serve your needs and goals. You're finding creative ways to make AI systems work for your specific context, even when they weren't explicitly designed for your particular use case.

Consider this real-world example from a technical writing team at a large software company.

The team was responsible for documenting an increasingly complex product that received updates every two weeks. Their challenge wasn't just keeping up with the volume of changes but ensuring consistency across hundreds of interconnected help articles. Standard documentation tools didn't offer effective ways to track these relationships or identify which articles needed updating when features changed.

Rather than accepting these limitations, the team created a custom workflow that:

1. Monitored the development team's code repository for changes to specific feature areas

2. Automatically extracted key terms and functions from the updated code

3. Searched their documentation for all articles mentioning these terms

4. Generated a prioritized list of potentially affected documents with confidence scores

5. Created first drafts of change notes for the highest-confidence matches

This system didn't just save time—it transformed how the team approached documentation. Instead of frantically reviewing all documents after each release, they could focus their attention precisely where it was needed most. The system wasn't perfect, occasionally missing connections or suggesting unnecessary updates, but it dramatically improved their process.

What made this approach a true expression of metis wasn't just the technology but how it embodied the team's unique expertise and knowledge of their documentation ecosystem.

They built a system that reflected their understanding of how their particular product's features connected to their specific documentation structure—knowledge that no general-purpose AI tool could possibly possess.

This represents a significant shift in our relationship with AI:

• Instead of being constrained by the limitations of existing tools, you create new ones tailored to your specific needs.

• Rather than adapting to how machines work, you adapt the machines to how you and your team work.

• Instead of accepting biased or problematic AI outputs, you design systems that align with your values and priorities.

By building AI agents and workflows, you become an architect of automated processes rather than merely a user of them. You're essentially writing the instructions not just for what content should be created, but for how your entire content system should operate.

The technical writing team didn't just create a tool—they extended their collective intelligence and expertise into an automated system that embodied their best practices and deepest knowledge of their product.

The system didn't replace their expertise; it amplified it, allowing them to apply their judgment and insight more effectively by eliminating hours of tedious document review.

This form of metis—the ability to creatively reshape technology to serve human needs—represents one of the most powerful aspects of machine rhetorics. It's not about mastering AI for its own sake, but about bending these technologies to better serve your specific content operations challenges.

The Anatomy of an AI Workflow

Today's marketplace is flooded with AI automation tools with impressive-sounding capabilities. From "intelligent content optimization platforms" to "autonomous marketing assistants" to "self-improving documentation systems," these tools often present themselves as revolutionary technologies with proprietary approaches.

But here's a secret that will help you navigate this landscape and eventually build your own systems: beneath their sleek interfaces and marketing language, virtually all AI automation tools share the same fundamental architecture. They all require the same four core components to function, regardless of their specific application or industry focus.

Let's examine these universal building blocks while looking at some popular tools in the market:

1. Trigger

This is the event or condition that initiates the workflow. A trigger might be:

• A specific time (every morning at 8 AM)

• A new piece of data (a fresh article published on a website)

• A user action (clicking a browser extension button)

• A system event (a new file being added to a folder)

Real-world examples:

• Zapier's automation platform calls these "Triggers" explicitly in their interface

• HubSpot's marketing automation calls them "Enrollment criteria"

• Microsoft Power Automate refers to them as "When this happens..."

• IFTTT (If This Then That) builds the trigger concept directly into its name

For technical writers, a valuable trigger might be when documentation files are modified in your content management system, prompting an automated quality check. Content creators might use the publication of competing content as a trigger to generate differentiation analyses. Educators could set up triggers based on assignment submission deadlines to automatically distribute work for peer review.

2. AI Agent

This is the artificial intelligence component that performs the core task. The agent leverages structured prompting techniques but operates without direct human supervision.

Real-world examples:

• Copy.ai calls this their "AI writer"

• Jasper refers to its "AI assistant"

• GitHub Copilot calls it their "AI pair programmer"

• Grammarly's system functions as an "AI writing assistant"

Despite the different names, these are all essentially AI systems following structured instructions to process information and generate outputs. The quality and usefulness of these systems depend largely on how well their prompts are structured—exactly what we've been exploring throughout this course.

For example, a technical documentation team might employ an agent that reviews newly written procedures and suggests clarifications based on readability scores and common user questions. Content marketers might use an agent that transforms long-form content into multiple social media posts with appropriate hashtags. Educational content developers could implement an agent that converts lecture notes into interactive quiz questions to reinforce learning.

3. Connections

These are the integration points between your AI agent and other systems. Connections might include APIs to access external data, file system interfaces, email services, or database connections.

Real-world examples:

• Make.com (formerly Integromat) calls these "Modules"

• Airtable's Automations refers to them as "Actions"

• Notion AI integrates with your existing Notion database

• Salesforce's Einstein connects with your customer data

These connections are what make AI agents truly powerful. A technical writing workflow might connect to your product roadmap system to automatically flag documentation that needs updating based on upcoming feature changes. Content creators might establish connections to analytics platforms to inform content optimization decisions. Educators could connect their grading systems with content repositories to automatically suggest resources based on student performance patterns.

4. Output

This is the result produced by the workflow. The output might be a document stored in a specific location, an email sent to stakeholders, data added to a spreadsheet, or even a trigger for another workflow.

Real-world examples:

• Buffer delivers social media posts to multiple platforms

• Canva's Magic Write generates design copy

• Fathom AI produces meeting summaries

• Notion AI creates formatted content in your workspace

For technical communicators, an output might be a weekly report highlighting potential inconsistencies across your documentation suite. Content creators might receive a personalized digest of trending topics in their niche with AI-generated content ideas tailored to their publication's style. Educational content developers might produce automatically differentiated versions of the same learning materials targeted at various student proficiency levels.

Understanding the Patterns Behind the Products

By recognizing these common patterns across different AI automation tools, you gain several advantages:

1. Better evaluation skills: You can more effectively assess whether a new tool actually offers unique capabilities or just repackages familiar components.

2. Improved integration abilities: Understanding these shared structures helps you connect different tools into more powerful custom workflows.

3. Greater independence: When you understand the basic architecture, you can build your own solutions when commercial tools don't quite meet your needs.

4. Strategic perspective: Rather than getting caught up in feature comparisons, you can focus on how well a tool's architecture aligns with your specific content operations.

This perspective also reveals an important truth: the most sophisticated AI automation isn't necessarily the one with the most advanced AI model, but the one that most effectively orchestrates these four components to address your specific content challenges.

As we continue our exploration of automation, remember that you're not just learning to use tools—you're learning to think about content operations in a way that allows you to design your own tools when necessary, combining these universal building blocks in unique ways that address your specific needs.

Starting Simple: Your First AI Workflow

While the potential of AI workflows is vast, complexity is the enemy of reliability. Just as we developed chatbots iteratively, we'll approach workflow automation with a similar mindset:

• Start with a single, well-defined task

• Make sure it works consistently before adding complexity

• Add components incrementally, testing at each stage

• Focus on creating value, not technical sophistication

For technical writers, a simple first automation might monitor your product's GitHub repository for documentation-related issues and compile them into a daily digest. This targeted workflow accomplishes one specific task reliably rather than attempting to automate your entire documentation process at once.

Content creators might begin with a workflow that identifies trending hashtags in your niche and suggests content ideas based on them. This focused automation addresses a specific pain point—ideation—before expanding to more complex aspects of your content creation process.

Even these simple automations can dramatically improve your productivity and effectiveness. A workflow that checks RSS feeds for new content in your field, summarizes the articles, and emails you the results each morning might take just a few hours to build but save you hours of work each week.

The Writer as System Designer

Throughout this course, we've emphasized that working with AI is fundamentally an act of writing—crafting language to achieve specific outcomes. Building automated workflows extends this perspective, positioning you as both a writer and a system designer.

This evolution is particularly meaningful for professionals who work with content.

Technical writers have always served as bridges between complex systems and their users. Now, you're not just documenting systems—you're designing them. The same skills that help you create clear documentation—understanding user needs, breaking complex processes into manageable steps, ensuring consistency—are valuable assets in designing effective AI workflows.

Content creators already think systematically about audience journeys and engagement funnels. Designing AI workflows leverages this strategic thinking in new ways, allowing you to automate parts of the content lifecycle while maintaining your distinctive voice and perspective.

When you build an AI workflow, you're still writing—not just prompts for the AI, but instructions for how various components should interact. You're creating a system that processes language, generates language, and delivers language to specific destinations. You're essentially writing a meta-narrative that guides how content flows through your professional ecosystem.

Case Study: The Content Monitor

Let's explore how a chatbot might evolve into an automated workflow.

Read more

One of the most accessible ways to leverage your structured prompts is through chatbots - conversational agents designed to perform specific tasks through natural dialogue.

The prompt library you've developed in our previous lesson serves as the foundation for creating more sophisticated AI applications.

That’s why it is important to master structured prompting … not so you can get this one AI interaction to work … but so you can builder better and bigger AI content systems.

Why consider creating a chatbot? The applications are remarkably versatile across different professional contexts:

For content creators, chatbots can serve as specialized assistants that streamline repetitive aspects of your workflow.

Imagine a chatbot that helps you brainstorm newsletter topics based on your audience analytics, transforms your long-form content into social media posts tailored to different platforms, or analyzes your writing for clarity and engagement. Rather than switching between multiple tools, a well-designed chatbot brings these capabilities into a conversational interface that fits naturally into your creative process.

Educators can leverage chatbots to extend their teaching presence beyond class hours.

A chatbot could provide personalized feedback on student drafts using your specific rubric, create differentiated practice activities based on individual learning needs, or even simulate discussion scenarios for students to practice critical thinking. These applications don't replace the educator—they amplify your ability to provide timely, consistent support to students.

For technical writers, chatbots can become powerful quality control tools that help maintain documentation standards.

A specialized bot might analyze technical content for reading level and jargon, verify that procedures follow your organization's required structure, or even generate template-based documentation for common scenarios. These applications help technical writers focus on high-value communication challenges while ensuring consistency across large documentation sets.

What makes these applications powerful isn't just the technology—it's the specialized focus on specific, high-value tasks. This brings us to our first crucial insight about effective chatbot design.

Subscribe now

Understanding Chatbots as Tools, Not Personas

When designing chatbots, the most common mistake is creating a persona rather than a tool. People often approach chatbot design by thinking, "I want to create a writing coach" or "I need a research assistant." This approach seems intuitive but often leads to disappointing results, because you are leaving too much to chance … or the models default training.

Instead, think of your chatbot as a specialized tool designed to perform a specific task or set of closely related tasks. Rather than creating a "content marketing assistant," consider creating a "headline optimizer" or a "content repurposing guide" that excels at one clearly defined function.

This distinction between tools and personas is crucial for several reasons:

1. Focus creates effectiveness: The more narrowly defined a chatbot's purpose, the more effectively it can perform its specific function. A chatbot designed to "help with writing" faces an impossibly broad domain, while one designed to "identify passive voice and suggest active alternatives" can be finely tuned to excel at that specific task.

2. Clear user expectations: Users understand exactly what the chatbot can and cannot do, leading to more satisfying interactions. When expectations match capabilities, users experience the chatbot as helpful rather than frustrating.

3. Reduced complexity: Designing for a specific task simplifies the development process and reduces the likelihood of errors. Each additional capability introduces new edge cases and potential issues; focusing on a single capability keeps the design manageable.

4. Better measurement: It's easier to evaluate and improve a chatbot's performance when its purpose is clearly defined. With a narrow focus, you can create specific success metrics and test cases that would be impossible for a general-purpose assistant.

Consider the difference between these two approaches:

General Approach: "I'll create a social media assistant chatbot that helps with all aspects of social media management."

Focused Approach: "I'll create a chatbot that helps transform long-form blog content into engaging Twitter threads while preserving key points and maintaining brand voice."

The general approach sounds appealing but presents enormous challenges. How would you train it to handle every aspect of social media? How would users know what to ask? How would you measure success?

The focused approach, by contrast, defines a clear problem with specific inputs, outputs, and success criteria. You can design prompts specifically for analyzing blog content, identifying key points, and reformatting those points as Twitter-appropriate content. Users know exactly what to expect, and you can measure success by how effectively the chatbot preserves key information while adapting to the platform's constraints.

Remember, AI excels when properly constrained. By limiting your chatbot's scope, you're not reducing its value—you're enhancing it. The most useful tools in your workshop aren't the multi-purpose ones—they're the specialized tools designed to excel at specific tasks.

This tool-oriented mindset also helps avoid one of the most common pitfalls in AI interaction: the illusion of general intelligence.

When users perceive a chatbot as a general assistant (a persona), they develop unrealistic expectations about its capabilities. When the chatbot inevitably fails to meet these expectations, users become frustrated and distrustful.

By presenting your chatbot as a specialized tool with clear boundaries, you set appropriate expectations from the start.

The Anatomy of an Effective Chatbot

When we look at chatbot design as technical writers, educators, or content creators, we're really looking at a new form of communication. Just as we carefully craft our documents, lesson plans, or content strategies, we need to thoughtfully design our chatbots to serve our audiences effectively.

Every effective chatbot consists of three core components that mirror what we already know about good communication:

1. Purpose: A clear definition of what the chatbot does and doesn't do

2. Process: A structured approach to accomplishing its task

3. Presentation: How the chatbot communicates with users

Let's explore each of these components in a way that connects to our everyday work.

Defining Your Chatbot's Purpose

As content professionals, we know that defining the purpose is always the first step in any project. For your chatbot, this means identifying the specific problem it solves.

I've found that the most useful chatbots address very specific needs rather than trying to be all-purpose assistants. For example:

• For content creators: "This chatbot helps transform long-form blog posts into engaging Twitter threads while preserving key points."

• For educators: "This chatbot helps students identify credible sources for their research papers by analyzing citation information."

• For technical writers: "This chatbot simplifies complex procedures by breaking them down into step-by-step instructions with visual cues."

Notice how each purpose identifies both the user and the specific task. This clarity is something we strive for in all our communication, and it's equally important in chatbot design.

When defining your purpose, ask yourself questions that connect to your professional practice:

• Who specifically will use this chatbot? (Just as you would define your audience for any content)

• What particular problem are they trying to solve? (Similar to defining the scope of a document)

• What falls outside the chatbot's capabilities? (Like setting boundaries in a content brief)

• How will users know if the chatbot has successfully helped them? (Similar to defining learning objectives or content goals)

Document these answers clearly. In my experience working with various content teams, this documentation becomes invaluable as your project evolves—just like a good creative brief or project plan.

Designing Your Chatbot's Process

The process component describes how your chatbot will accomplish its purpose—the conversation flow that guides users from question to answer.

This is where your structured prompt blocks become especially valuable. When I've designed chatbots for my own work, I've found that a good process typically follows these steps:

1. Initial orientation: Introducing what the chatbot does and setting expectations

2. Information gathering: Collecting necessary inputs from the user

3. Processing: Applying your prompt blocks to transform inputs into outputs

4. Output delivery: Providing results in a useful format

5. Refinement: Allowing for feedback and adjustments

For each step, you'll need to design appropriate prompt blocks. For instance, if I were creating a chatbot to help transform blog content into social media posts, I might include:

• A welcome message that explains what the chatbot does and what I'll need to provide

• Questions about my target platform and audience (since what works on LinkedIn differs from Twitter)

• Analysis of my blog content to identify key points

• Generation of platform-appropriate content that maintains my voice

• Options to refine the generated content based on my feedback

These prompt blocks can come directly from your prompt library, combined in sequences that accomplish your chatbot's specific purpose.

What's fascinating to me is how this process mirrors good instructional design or technical writing. We're essentially creating a conversation that guides users through a process, helping them achieve their goals through well-structured steps—much like a well-designed tutorial or lesson plan.

Crafting Your Chatbot's Presentation

As content professionals, we know that how we say something is often as important as what we say. The presentation component addresses how your chatbot communicates with users. This includes:

• Tone and voice: How formal or casual should the chatbot be?

• Interaction style: Should the chatbot ask multiple questions or process everything at once?

• Error handling: How should the chatbot respond when it doesn't understand or can't complete a task?

• Visual elements: What formatting will make the chatbot's outputs most useful?

Your presentation choices should align with your users' expectations and the context in which they'll use the chatbot. In my work with educational technology, I've found that a chatbot for students might use a more casual, encouraging tone, while a chatbot for professional researchers might prioritize precision and efficiency.

This is no different from how we adapt our writing style for different documentation types or audience needs. Just as we might use a friendly, conversational tone for a blog post but a more formal approach for technical documentation, our chatbots need to communicate in ways that feel appropriate to their context and users.

When designing my first chatbot to help with student feedback, I found that building in friendly encouragement along with specific guidance made the experience feel more supportive—similar to how I might balance critique with encouragement in face-to-face feedback sessions.

By thoughtfully designing these three components—purpose, process, and presentation—you create conversations that not only accomplish tasks but do so in a way that feels natural and helpful to users. The most effective chatbots don't just process information; they guide users through tasks in a clear, contextually appropriate manner that respects their needs and time constraints.

This approach to chatbot design draws on everything we already know about good communication—clarity of purpose, well-structured process, and appropriate presentation. We're simply applying these principles to a new medium of conversation.

The Connection to Rhetoric

What we're doing when we design chatbots isn't entirely new. In fact, we're drawing on principles that date back thousands of years to the ancient art of rhetoric—the study of effective communication and persuasion.

When I first began exploring AI, I was struck by how much chatbot design resembles classical rhetorical principles. The purpose component mirrors what rhetoricians call "exigence"—the situation that calls for a response. The process component resembles "arrangement"—how arguments are structured for maximum impact. And the presentation component connects to "style" and "delivery"—how messages are expressed and conveyed.

This connection isn't just academic—it's practical. For thousands of years, humans have been studying how to communicate effectively, developing frameworks that help us organize our thoughts, present information clearly, and connect with audiences. Now we're applying these same timeless principles to a new medium.

As content professionals, we're already practicing rhetoric every day, whether we use that term or not. When we structure a document for clarity, choose words carefully for our audience, or design learning experiences that guide students step by step, we're using rhetorical techniques.

Chatbot design simply extends these skills into conversational interfaces. Instead of arranging information on a page, we're arranging it in a dialogue. Instead of choosing a writing style, we're designing a conversational style. The medium has changed, but the fundamental principles remain the same.

Understanding this connection can help us leverage what we already know about effective communication as we venture into this new territory. Our experience with audience analysis, information design, and clear communication gives us a head start in creating effective chatbots.

Planning Your Chatbot: A Step-by-Step Approach

Now that we understand the core components of effective chatbot design, let's explore a systematic approach to planning your chatbot. I'll walk you through each step of the process while sharing how I applied these principles to create an educational feedback chatbot for my writing courses.

Read more

Despite some predictions that prompt engineering might become obsolete with advancing AI technology, the reality for content professionals working at scale is quite different … whether your a content developer, creator, or educator.

As recently noted in a webinar on AI & Content Ops — The value that writers bring isn't the writing... it's the workflow mindset.

The best writers reflect on their processes, or workflows, and adapt to become better writers. Content operations is simply scaling this mindset for a classroom, team, or organization.

This approach inevitable leads to what many call a prompt library when working with AI, because this workflow mindset is precisely what distinguishes random prompting from systematic prompt operations.

When building a prompt library, we should aim to follow the FAIR principles that guide effective knowledge management systems … a key way of thinking in content operaations:

Findable: Your prompts should be easily discoverable through consistent organization and robust search capabilities.

Accessible: Team members should be able to access relevant prompts when needed, without complicated permissions or technical barriers.

Interoperable: Prompts should work across different contexts and integrate with your existing tools and workflows.

Reusable: Well-designed prompts should be adaptable for multiple purposes, reducing duplication and saving time.

These FAIR principles help transform scattered prompts into organizational assets that preserve knowledge, ensure consistency, and improve efficiency across your content operations.

Subscribe now

From Theory to Implementation: Choosing Your Tool

The taxonomy you developed in our previous lesson provides the conceptual structure for your prompt library. Now we need a practical system to implement that structure.

Many options exist, each with different strengths:

Enterprise Knowledge Systems: Tools like Confluence or Microsoft Loop can integrate prompt libraries into existing knowledge bases. This works well if your team already uses these platforms, but they weren't specifically designed for prompt management.

Specialized Prompt Management: Applications like Promptitude offer dedicated features for prompt organization, version control, and collaborative development. These provide robust capabilities but may require subscription costs and learning new interfaces.

Knowledge Graph Tools: Platforms like Anytype allow you to create sophisticated relationships between prompts and other knowledge objects with low cost. This approach offers powerful flexibility but comes with a steeper learning and maintenance curve.

Simple But Effective Options: A system of text files in organized folders can work with good naming conventions, though it lacks advanced features like tagging and search.

For our implementation example, we'll use Twos—a versatile productivity app that provides an accessible entry point to prompt library management while demonstrating principles that apply to any system.

I'm currently exploring more advanced implementations with tools like Promptitude and Anytype, which I'll cover in future lessons or tutorials.

Understanding the Object-Oriented Foundation

Before we dive into specific implementation, it's important to understand the theoretical approach that makes systems like Twos so versatile. Many modern productivity tools employ what's essentially an "object-oriented" approach to organizing knowledge.

In this approach, individual pieces of information are treated as discrete objects (or "things" in Twos terminology) that can:

• Be categorized in multiple ways simultaneously

• Contain various properties (metadata)

• Be related to other objects

• Exist in multiple contexts without duplication

For most people using Twos, these "things" are events, todos, and notes. However, the beauty of this object-oriented approach is that a "thing" can be anything—including prompts and prompt components.

When we build a prompt library in such a system, each prompt becomes an object with properties (metadata), relationships (to categories and other prompts), and the ability to appear in multiple contexts without being duplicated.

This approach provides the flexibility and interoperability that makes a prompt library truly useful.

As we've explored in previous lessons, effective prompts are often modular, composed of distinct "prompt blocks" that serve specific functions. These blocks might include:

• Role blocks: Defining the AI's persona or perspective

• Context blocks: Providing background information

• Audience blocks: Defining who the content is for

• Style blocks: Setting the tone and writing approach

• Task blocks: Specifying what the AI should do

The modular nature of prompts makes them perfect candidates for an object-oriented organizational approach. Each prompt block can be stored, categorized, and reused across multiple complete prompts, allowing for consistent elements while enabling customization for specific needs.

You can take this mindset to any tool that fits you and your team’s workflow, but in the rest of this article, I’ll show you how to adapt a simple, structured knowledge management like Twos into a FAIR prompt library.

By the way, Twos is completely free and one of the most innovative and accessible knowledge management platforms out their for non-experts.

Setting Up Your Prompt Library in Twos

Twos organizes information through a straightforward system of days, lists, and "things"—individual pieces of information that can be organized, tagged, and retrieved when needed.

This structure aligns perfectly with how prompt libraries function. You just need to create a new kind of thing called a prompt.

Twos works on both a vertical and horizontal organizing principle. Vertically, you organize information in hierarchical lists and sublists—think of them as folders that can contain both items and other folders. Horizontally, you use tags to create connections across this hierarchy, allowing items to be found through multiple pathways.

When you open Twos, you'll see a main screen that defaults to "Today"—this is your daily note where you can capture anything that comes to mind. But for a prompt library, we'll create a dedicated organizational structure. You'll start by creating a main "Prompt Library" list, then add category sublists beneath it, followed by specific prompt sublists under each category.

What makes Twos particularly powerful for prompt libraries is its treatment of "things." Each thing (like a prompt block or complete prompt) can be tagged, searched, and—most importantly—reused across multiple lists without duplication. When you update a prompt block in one location, it updates everywhere it appears.

The search functionality in Twos is impressively robust. You can search not only by text but also by tags and even combinations of tags (like "#beginner AND #conceptual"). This makes finding the exact prompt components you need quick and efficient, even as your library grows.

I'll be creating a detailed video tutorial soon that will walk through setting up a complete prompt library in Twos, showing exactly how to implement the concepts we've discussed.

In the meantime, you can check out this general Twos tutorial to familiarize yourself with the platform's basic functionality and interface.

Read more

Imagine you're developing training materials for a new employee onboarding program. Your company has just updated its project management software, internal communication protocols, and security procedures.

While you could simply ask AI to help generate training guides, relying on its default knowledge would be risky. The AI might draw on generic onboarding practices or outdated business procedures, creating materials that don't reflect your company's specific workflows and culture.

Consider the complexity of what new employees need to learn: how to use multiple software tools, when to use different communication channels, and which security practices apply to different situations. Your training materials need to be accurate, accessible, and aligned with your company's specific practices. More importantly, they need to build on each other in ways that help employees develop confidence and competence progressively.

This scenario illustrates a fundamental challenge in AI-assisted content development: how to effectively provide and structure knowledge in our prompts. While AI models are trained on vast amounts of internet data, this default knowledge comes from unknown sources and may be outdated, biased, or simply incorrect.

For content developers and training teams, relying solely on this default knowledge isn't just risky – it can lead to confusion, inefficiency, and potential security issues.

Think about how you develop training materials traditionally.

You consult company policies, interview experienced employees, review existing documentation, and consider what has worked well in previous training sessions. You might reference similar programs or compare different teaching approaches. You consider your audience's background and common challenges.

Working effectively with AI requires bringing all these knowledge sources together in structured ways that help the AI understand and use them effectively.

Subscribe now

How AI Processes Knowledge

To use knowledge effectively in our content development workflows, we first need to understand how AI processes the information we provide. Think of AI as a new team member who has read vast amounts of content from the internet but needs guidance about your specific organization and requirements.

Modern AI language models convert text into mathematical representations called vectors - essentially turning words and concepts into numbers that they can process.

These vectors don't just represent individual words; they capture relationships between concepts. For instance, in employee training, the AI understands that "onboarding" relates to concepts like "orientation," "training modules," and "new hire paperwork" because these ideas frequently appear together in its training data.

When you provide new knowledge in a prompt, the AI processes it by connecting your specific information with its general understanding.

Let's say you're creating onboarding materials about your company's unique approach to project management. The AI already understands basic project management concepts from its training, but it needs your specific knowledge about how your company handles project timelines, team collaboration, and resource allocation.

This process is similar to how an experienced trainer might adapt general business practices to fit your company's specific culture. Just as that trainer draws connections between standard practices and your unique implementations, the AI attempts to bridge its general knowledge with your specific requirements.

For longer documents or comprehensive resources like employee handbooks and training manuals, AI systems use a technique called Retrieval Augmented Generation (RAG).

Imagine having an extremely thorough research assistant who can instantly search through all your company documentation, find relevant information, and incorporate it into new training materials. When you upload documents like company policies, existing training guides, or process documentation, the system breaks them down into smaller, manageable pieces and converts each into vectors.

For example, if you're developing a module on communication protocols, RAG could help you:

• Find relevant sections from your company's communication policy

• Reference examples of effective internal communications

• Pull specific guidelines from your style guide

• Identify common communication challenges from HR documentation

However, just like you wouldn't dump a stack of policy documents on a new hire's desk without guidance, you need to help the AI understand which parts of these materials are most relevant to your current task. Simply uploading documents isn't enough - you need to provide context about how this knowledge should be used in your current content development project.

Understanding this process helps us develop more effective strategies for structuring information in our prompts.

In the next section, we'll explore practical approaches to organizing knowledge that help AI create more accurate and useful content for your specific needs.

Making Knowledge Work Effectively

Understanding how AI processes knowledge leads us to an important question: How do we provide information in ways that help AI create truly useful content?

The answer lies in thoughtful structure and clear organization, much like how we organize training materials to build understanding progressively.

Remember our onboarding program example. When an experienced trainer develops new modules, they don't simply present information randomly. They organize content carefully, establish key concepts first, and build on that foundation. We need to take a similar approach when providing knowledge to AI.

One powerful method borrows from a technology called XML, which uses tags to label different types of information. While you don't need to understand XML technically, this concept of "tagging" knowledge helps organize information in ways that AI can use more effectively. Think of it like creating clear sections in a training manual, each serving a specific purpose.

For instance, instead of providing a jumbled list of information about your communication protocols, you might structure it like this:

<communication_policy> Our company uses Slack for immediate team communication, email for formal communications and external contacts, and Microsoft Teams for video meetings and file sharing. All sensitive information must be shared through our secure internal portal. </communication_policy>

<common_challenges> New employees often struggle with knowing which communication channel to use when. They frequently default to email when real-time collaboration would be more effective, or use public channels for sensitive discussions that should be private. </common_challenges>

<success_metrics> Employees should be able to:

- Choose appropriate communication channels for different situations
- Follow security protocols for sensitive information
- Effectively collaborate using our digital tools
- Maintain professional communication standards across all platforms </success_metrics>

This structured approach helps the AI understand not just what each piece of information is, but how it should be used in creating training materials. When you later ask the AI to help develop specific modules or exercises, it can draw on this clearly organized knowledge to create more targeted and effective content.

Let's look at how this plays out in a real content development scenario. Imagine you're creating a series of microlearning modules about communication tools. You might structure your prompt like this:

<learning_context> These modules will be delivered as 5-minute video scripts, focusing on one communication tool per module. Learners will watch these during their first week, before they start actively using the tools. </learning_context>

<tool_specifics> Microsoft Teams features we use most:

- Channel-based team discussions
- Video meetings with screen sharing
- Document collaboration
- Integration with other Microsoft tools </tool_specifics>

<audience_background> Our new hires typically have experience with basic email and chat tools, but many haven't used Teams in a professional setting. They're often anxious about making mistakes in company-wide channels. </audience_background>

By providing knowledge in this structured way, you help the AI understand:

• The format and constraints of the content it needs to create

• The specific features that need to be covered

• The audience's starting point and concerns

This allows the AI to generate content that's not just accurate, but truly useful for your specific situation and audience.

This structured approach to knowledge organization in prompts connects directly back to our earlier discussions about taxonomies.

Taxonomies help us organize our prompt libraries and guide how we structure knowledge within individual prompts. Think of it as creating a miniature knowledge taxonomy for each content development task.

Remember how we discussed organizing prompts into categories that reflect natural workflows and relationships? The same principles apply when organizing knowledge for specific content tasks.

Your knowledge structure should reflect both the natural relationships between different pieces of information and the practical ways that information will be used in content creation.

For instance, in our training materials example, we might organize our knowledge following similar patterns to our prompt taxonomy:

Core Knowledge - This includes fundamental information that shapes everything else: company policies, essential procedures, and key objectives. Like the main departments in our prompt taxonomy, this forms the foundation of our content.

Contextual Knowledge - This provides important background and situational information, much like how our taxonomy creates connections between related prompts. It helps the AI understand how to apply the core knowledge effectively.

Implementation Knowledge - This includes specific examples, common scenarios, and practical applications. Similar to how our taxonomy helps us find the right prompts for specific tasks, this knowledge helps the AI create content that bridges theory and practice.

By aligning our knowledge organization with our broader taxonomical approach, we create consistent patterns that both humans and AI can follow more easily. This consistency helps ensure that whether we're organizing our prompt library or structuring knowledge within individual prompts, we're working within a coherent, systematic framework.

As you develop your own approach to knowledge structuring, consider how it might align with your existing content organization systems.

How can your prompt taxonomy inform the way you structure knowledge? How might your knowledge organization patterns suggest new ways to organize your prompt library?

Direct Knowledge vs. Reference Materials: Finding the Right Balance

When developing content at scale, you'll often work with both small, focused pieces of information and larger bodies of documentation. Understanding when and how to use each type of knowledge is crucial for effective content development.

Think of it like planning a training program. Sometimes you need to provide specific instructions directly to a trainer - key points that must be covered, particular examples to use, or specific language choices.

Other times, you give them access to comprehensive training manuals and let them draw on that broader resource as needed. Working with AI requires similar judgment about when to provide knowledge directly and when to use reference materials.

Direct Knowledge Inclusion

Direct knowledge works best when you need precise control over how information is used. You're essentially telling the AI, "Use exactly this information in exactly this way." This approach is particularly valuable when accuracy and consistency are crucial.

For example, imagine you're creating a series of safety procedure guides. You might provide direct knowledge like this:

<safety_procedures> Emergency evacuation requires:

1. Immediate cessation of all activities
2. Exit through marked emergency routes only
3. Assembly at designated meeting points
4. Attendance check by department supervisors </safety_procedures>

<critical_context> Recent safety audit identified confusion about assembly points. Multiple employees attempted to use unauthorized shortcuts during our last drill. New materials must emphasize following marked routes only. </critical_context>

<compliance_requirements> All safety documentation must include:

- Clear step-by-step instructions
- Visual guides or diagrams
- Emergency contact information
- References to relevant safety regulations </compliance_requirements>

By providing this knowledge directly in your prompt, you ensure these crucial details are immediately available and prominently featured in the generated content.

Working with Reference Materials

For comprehensive resources like company handbooks, technical specifications, or extensive training materials, file references often work better. Instead of copying large amounts of text into your prompts, you can upload these documents and guide the AI to relevant sections.

However, simply uploading documents isn't enough. You need to provide context about how these materials should be used. Consider this approach:

<reference_materials> I've uploaded our complete employee handbook and safety manual. Focus specifically on:

- Chapter 3: Emergency Procedures
- Appendix B: Assembly Point Locations
- Section 7.2: Department Supervisor Responsibilities </reference_materials>

<content_goals> Create step-by-step guides that:

- Simplify complex procedures without omitting crucial steps
- Incorporate relevant visuals from the reference materials
- Maintain consistency with existing documentation </content_goals>

This combination of focused guidance and comprehensive references helps the AI create content that's both accurate and well-integrated with your existing materials.

Finding the Right Balance

The key to effective knowledge management is understanding when to use each approach. Consider using direct knowledge when:

• Specific wording or details must be precisely maintained

• You're working with critical information that can't risk misinterpretation

• You need to ensure certain elements are prominently featured

• You're establishing new standards or procedures

Use reference materials when:

• Working with comprehensive documentation that provides important context

• Creating content that needs to align with existing materials

• Dealing with complex topics that require drawing from multiple sources

• Maintaining consistency across a large body of content

Often, the most effective approach combines both methods, using direct knowledge to guide the AI's focus while keeping broader reference materials available for context and alignment. This hybrid approach helps ensure your content is both precisely accurate and consistently integrated with your larger documentation ecosystem.

Think of it as building a bridge between your immediate content needs and your broader knowledge base. Direct knowledge forms the essential structure, while reference materials provide the supporting framework that connects it to your larger content landscape.

➡️ Want to dive deeper into implementing these approaches in your content workflows? Paid subscribers get access to specific frameworks and examples, as well as tools like worksheets to help you get started on developing your prompt operations.

Subscribe now to unlock the full lesson and join our community of content developers mastering AI-assisted content creation.

Subscribe now

Read more

Think of this template as a map for organizing the different types of knowledge you need in your content development.

Like any good map, it helps you plan your route while staying flexible enough to accommodate different journeys.

To use this template, cut and paste it into your writing tool.

Start With Your Purpose

First, clearly define what you're trying…

Read more

This January, all proceeds from paid subscriptions will support my friend David's GoFundMe. After losing his mother, caring for his father, and losing his family home, David is fighting to keep his parents' two beagles. Your subscription will help him raise $3,000 for professional dog training—offering hope after his most challenging year.

➡️ You can also check out his GoFundMe directly here.

Subscribe now

In our last lesson, we explored frameworks for structuring your prompt operations. Before we start building a prompt library or creating templates, we need to step back and think systematically about how we'll organize our prompts … or at least get started. This is where taxonomies comes in.

So you might be wondering what a taxonomy is and what that has to do with content or AI. Simply put, its a way of organizing things … in this case prompts.

Think about how a grocery store organizes its products. It's not random - there's a careful system that groups similar items together and creates logical pathways for shoppers.

Produce is grouped together, with fruits in one section and vegetables in another. Dairy products have their own area. Frozen foods are grouped by type - vegetables, ready meals, desserts.

This organization isn't just about putting things in neat rows - it's about understanding how people use these items and navigate the physical space.

For example, a grocery store's layout creates meaningful connections. Fresh vegetables sit next to fresh fruits, forming a produce section. Nearby, you'll often find the deli with fresh-made salads.

This shows how these chains are understanding the ways shoppers think about and use these items together, not just like with like.

Someone making a salad might need tomatoes (produce), croutons (bakery), and salad dressing (condiments), so these sections are often within easy reach of each other.

The same kind of multidimensional thinking applies to organizing prompts. While you might have clear categories like "content creation" or "content analysis," the real power comes from understanding how these categories relate to each other in practice.

Just as the grocery store creates useful adjacencies between related products, your prompt taxonomy should reflect the natural connections in your content workflows.

And just as grocery stores have different sections for different types of products, your prompt taxonomy will have different categories for different types of prompts.

Why Taxonomy Matters in Prompt Operations

Creating a taxonomy is about being organized, but its also about creating a system that supports your workflows. When a grocery store decides where to place items, they think about how people shop - placing frequently bought items like milk toward the back, grouping items that are often bought together, and putting impulse purchases near the checkout.

Similarly, when creating your prompt taxonomy, you need to think about how you and your team work with prompts:

• What prompts do you use most often?

• Which prompts are typically used together?

• How do different team members think about and search for prompts?

• What relationships exist between different types of content work?

Let's see how the grocery store principle works with prompts in a content marketing operation.

Imagine you regularly create product announcements across different channels. You might have prompts for social media posts, email newsletters, blog articles, and press releases. A simple categorical organization might just list these as different types of content:

• Social Media Prompts

• Email Prompts

• Blog Prompts

• Press Release Prompts

But this one-dimensional organization misses important relationships in how these prompts work together. In practice, when launching a new product, you'll need prompts that help you maintain consistent messaging across all these channels while adapting the tone and format for each one.

A more useful organization would also consider these relationships. You might group your prompts both by channel (like grocery store departments) and by common workflows (like recipe ingredients). For a product launch, you might have:

Product Launch Bundle:

• Core messaging prompt (develops key points)

• Channel adaptation prompts (adjusts tone and format)

• Audience targeting prompts (customizes for different segments)

• Timeline prompts (creates content schedule)

These prompts naturally go together, just like spaghetti ingredients. While they could live in different "departments" of your library, they should be easy to find and use together across your prompt library. Your taxonomy might even include pre-made "recipe cards" - templates that combine these prompts in proven workflows.

This multidimensional organization means someone working on a product launch can easily find not just individual prompts, but entire workflows. Just as a grocery store might feature a display with everything needed for a summer barbecue, your prompt taxonomy should support common content creation scenarios.

This kind of thoughtful organization becomes even more valuable as your prompt library grows. A good taxonomy helps new team members understand not just what prompts are available, but how they work together to achieve specific content goals.

Creating Your Basic Taxonomy Structure

Before we explore complex relationships between prompts, let's start with a basic hierarchical organization - like creating the main departments and aisles of our prompt store.

A linear taxonomy starts with broad categories and moves to increasingly specific subcategories. For content operations, we typically begin with the primary functions our prompts serve. For instance, a content marketing team might start with these top-level categories:

• Generation Prompts: These create new content from scratch

• Transformation Prompts: These adapt existing content for new purposes

• Enhancement Prompts: These improve or optimize existing content

• Analysis Prompts: These evaluate content effectiveness

Within each category, we then identify specific types of prompts. For example, under Generation Prompts, you might have:

• Ideation Prompts: Help brainstorm content ideas and angles

• Outline Prompts: Structure content organization

• Draft Prompts: Create initial content versions

• Polish Prompts: Refine and complete content pieces

Let's see how this works with a real example. Imagine you're organizing prompts for a technical documentation team. Your initial taxonomy might look like this:

• Documentation Prompts

  • Product Description Prompts

  • Feature Overview Prompts

  • Technical Specification Prompts

• Integration Guide Prompts

  • Tutorial Prompts

  • Getting Started Prompts

  • Basic Usage Prompts

  • Advanced Usage Prompts

• Troubleshooting Prompts

  • Common Issues Prompts

  • Diagnostic Prompts

  • Resolution Prompts

• Reference Prompts

  • API Documentation Prompts

  • Configuration Prompts

  • Command Reference Prompts

Each level becomes more specific while maintaining clear relationships to its parent category. This hierarchical structure provides a foundation for finding and organizing prompts. For instance, if a team member needs to create API documentation, they can navigate directly to Reference Prompts → API Documentation Prompts to find relevant prompt templates.

The key to creating an effective basic taxonomy is balancing breadth and depth. Too many top-level categories can be overwhelming, while too few can make it difficult to find specific prompts. Similarly, going too deep with subcategories can make the system overly complex, while staying too shallow might not provide enough organization.

Start by identifying 3-5 main categories that cover your core content workflow. Then within each category, create no more than 3-4 subcategories. You can always add more granularity as your prompt library grows, but starting with a manageable structure helps ensure the system remains usable.

Understanding Relationships in Your Taxonomy

Just as a grocery store exists in three-dimensional space, your prompt taxonomy needs to capture multiple types of relationships. But visualizing these relationships can be challenging when you're just getting started. This is where tools like mind mapping and knowledge graphs become invaluable.

Think about how a grocery store creates useful adjacencies. The bakery might be near both the coffee section (morning routine) and the deli counter (lunch combinations), while also connecting to the broader center aisles where you find packaged breads and pastries. These multiple connections reflect different ways shoppers think about and use these products.

Mind mapping helps us visualize similar connections between prompts and prompt blocks. Starting with a central concept like "Product Launch Content," you might branch out to identify different types of needed content: social media, email, blog posts, press releases. But then you can add another layer showing how these connect through shared elements like tone guidelines, key messaging, or audience personas. Drawing these connections helps reveal patterns you might miss in a simple hierarchical list.

For example, a mind map for product launch prompts might reveal that your "audience" prompt blocks connect not just to product documentation, but also to sales content, customer support scripts, and technical blog posts. This insight might lead you to create more versatile audience prompts that serves multiple content workflows.

Knowledge graphs take this concept even further by allowing us to specify the nature of these relationships. In a knowledge graph, you don't just show that prompts are connected – you describe how they're connected. A "tone guidance" prompt might "inform" multiple content creation prompts, while "audience persona" prompts "constrains" how those same prompts operate.

Let's look at how this might work in practice. Consider these relationship types in your prompt taxonomy:

Hierarchical Relationships: A high-level "brand voice" prompt might contain or govern several more specific tone and style prompts. Just as the produce department contains the fruit section, which contains the apple display.

Adjacent Relationships: Your "product feature description" prompt naturally connects to both "technical specification" prompts and "benefit statement" prompts. These adjacent prompts often work together but serve different purposes, like finding pasta near sauce.

Workflow Relationships: Some prompts naturally sequence together. A "research synthesis" prompt might feed into an "outline generation" prompt, which then connects to various drafting prompts. This is similar to how a grocery store might organize ingredients in the order you'll use them in a recipe.

Functional Relationships: Prompts that serve similar purposes might be grouped together even if they're used in different contexts. All your "audience analysis" prompts might be connected, whether they're used for blog posts, email campaigns, or social media.

Creating a visual map of these relationships serves several purposes:

1. It helps you identify gaps in your prompt collection

2. It reveals opportunities for prompt reuse across different contexts

3. It makes it easier to train new team members on your prompt system

4. It supports the development of more sophisticated prompt workflows

You can start mapping these relationships with simple tools like paper and pencil, or digital mind mapping software like Miro or Context Mind. As your system grows more complex, you might want to explore dedicated knowledge graph tools (like Anytype) that can help you manage and visualize more sophisticated relationship networks.

Remember that, like a grocery store's layout, your relationship map should reflect how people actually work with prompts. Pay attention to which prompts tend to be used together, which ones support or depend on each other, and which ones might benefit from being more closely connected.

In our next several sections, we'll explore exactly how to build your own taxonomy from the ground up. I'll guide you through practical exercises and strategies for organizing your prompts, including:

• A step-by-step process for identifying your key prompt categories

• Common pitfalls to avoid when structuring your taxonomy

• Real examples of taxonomies for different content scenarios

• Advanced techniques for evolving your system as your needs grow

➡️ Paid subscribers get access to these detailed guides along with exercises and examples to help implement your own prompt taxonomy.

Subscribe now

Read more

This worksheet accompanies Lesson 9: Developing a Prompt Taxonomy (coming soon). You can cut and paste this into your own doc and fill in the spaces, or just add your answers to a separate sheet. If you need a little help … guess what? You can feed this to your AI and have it help you!

I’m also always available for consultation.

Read more

This January, all proceeds from paid subscriptions will support my friend David's GoFundMe. After losing his mother, caring for his father, and losing his family home, David is fighting to keep his parents' two beagles. Your subscription will help him raise $3,000 for professional dog training—offering hope after his most challenging year.

➡️ You can also check out his GoFundMe directly here.

Subscribe now

In our journey through AI content operations, we've explored the foundational elements of effective prompting—from understanding transformer technology to mastering style and temperature controls.

As AI technology develops, some experts predict the eventual disappearance of traditional prompt engineering, suggesting that more sophisticated knowledge bases and AI architectures will make explicit prompting obsolete.

And this may be true for the vast majority of people who don’t necessarily work with content at scale as technical writers, educators, or content creators.

Yet, understanding the fundamental principles of structured prompting remains crucial—not just for immediate content generation, but as a foundation for building more consistent AI content operations and workflows, which most likely will lead to more complex knowledge systems.

But learning how to structure the workflows around your prompts (not just the prompts themselves) is the first step … especially when you are looking to scale your content as a business or organization.

Taking a structured approach is how we ensure consistency, scalability, and quality across our AI-assisted content workflows.

The consequences of unstructured prompting extend beyond mere inconsistency:

1. Resource Inefficiency: Teams repeatedly "reinvent the wheel" when crafting prompts

2. Quality Inconsistency: Similar prompts produce varying results across different team members

3. Limited Scalability: Success becomes dependent on individual expertise rather than systematic approaches

4. Difficult Knowledge Transfer: New team members struggle to replicate successful prompts

These challenges mirror broader issues in content operations, where lack of structure can impede organizational effectiveness. Adding AI to the mix simply compounds the issues that are already there.

This isn’t just for content developers or tech writers. An educator crafting differentiated learning materials faces challenges similar to a content creator developing multi-platform narratives—both must manage complexity while maintaining rhetorical effectiveness across diverse contexts.

These shared challenges reveal the universal value of structured approaches, whether you're designing a curriculum, developing a content strategy, or crafting compelling narratives.

Thinkers and practitioners in content development have already been thinking about this. For example, Mark Baker’s book, Structured Writing: Rhetoric & Process, shows the close connection between structure and rhetoric. Even though its not specifically written for AI operations, the principles certainly apply. And that’s what I want to explore today.

Drawing from Baker's analysis of structured writing and my experiences across educational and creative contexts, we can identify specific methodologies that enhance development of AI writing systems.

Just as we use style guides and templates to standardize traditional writing, we need frameworks to systematize our AI interactions.

The Power of Structure in AI Operations

Everyone that deals with content has to think about this question: How do we systematically address the complexity of content at scale in ways that enhance how our AI tools work?

This often means systemizing the ways we provide context to both humans and machines … much of which is rhetorical.

Mark Baker's analysis of structured writing provides a valuable theoretical framework for understanding these challenges. He observes that "all writing is structured" and that the key differentiation lies in how we add structure "over and above the basic requirements of grammar, to exercise some control over the rhetoric or processing of the content."

If we want to exert control over how AI systems use our context and integrate rhetorical approaches, then we need to develop frameworks to help us do this more systematically.

Think about your typical global software company's documentation team. Many of the challenges they face are indicative of all forms of systematic content creation.

These teams need to maintain consistency across multiple levels of their documentation ecosystem: user guides for different products, varying technical depths for diverse audiences, and content adapted for multiple platforms—from detailed online documentation to quick-start guides and mobile interfaces.

Initially, each writer might approach these challenges independently, leading to what Baker would identify as unmanaged complexity in the content system.

A writer documenting the enterprise version of a product might adopt a highly technical tone, while another writer, working on the same feature for the small business version, might take a more conversational approach.

These variations, multiplied across products, platforms, and audience levels, can create a documentation landscape that becomes increasingly difficult to maintain and navigate.

This challenge resonates deeply across different content domains. Consider the educator developing a curriculum that must work across multiple course sections, accommodate different learning modalities, and maintain pedagogical consistency while adapting to diverse student needs.

Or examine the content creator crafting narratives that must maintain brand voice across social media platforms while adapting to each platform's unique characteristics and audience expectations.

The introduction of structured prompt operations can provide what Baker calls "repeatable rhetorical structures." These structures don’t merely standardize the content—they enhance their effectiveness by allowing creators to focus their expertise on content rather than constantly reinventing organizational patterns … or in our case, AI prompts.

Five Principles for Structured Prompt Operations

Drawing from both practical experience and Baker's theoretical frameworks, I've identified five key principles that guide structured prompt operations.

These principles aim to partition complexity effectively while ensuring consistent, high-quality outputs.

1. The Modular Mindset

Modularity in prompt design parallels the partitioning principles central to structured writing. Baker emphasizes that partitioning complexity doesn't eliminate it—rather, it directs complexity to where it can be handled most effectively.

Consider an educational content developer creating a series of lesson plans. Rather than crafting unique prompts for each lesson, they might develop modular components, or prompt blocks, for:

• Learning objective generation

• Activity development

• Rubric creation

• Different learning levels

These prompt blocks can be combined and customized while maintaining pedagogical consistency across the curriculum.

2. Connection Mapping

Baker discusses how structure creates "context that you can use to simplify processing." In prompt operations, this principle manifests in the systematic mapping of relationships between prompt components.

A technical writing team might implement this approach for software release documentation:

• Core feature prompt blocks linked to various audiences

• Technical specifications blocks connected to use case scenarios

• Troubleshooting context blocks mapped to specific user questions

• Installation prompt blocks tied to various system requirement descriptions

This mapping enables content developers to generate comprehensive documentation that maintains consistency while addressing different user needs.

3. Semantic Tagging

Of course, this can get complicated quickly, which is why we need to semantically tag each prompt block.

Semantic tagging helps create clear relationships between content components that can be later used to organize a prompt library taxonomically … or even a knowledge base.

For example, a content marketing team might develop tags that identify:

• Content purpose (education, engagement, conversion)

• Audience segment (technical, managerial, end-user)

• Content type (how-to, concept explanation, reference)

• Subject matter domain (specific product features or topics)

This granular approach helps identify and manage consistent elements across the content system.

4. Design for Reuse

Creating what Baker calls "repeatable rhetorical structures" becomes crucial for scaling content operations. So we have to think about how each prompt block might be reused or re-purposed.

For example, a university course development team might demonstrate this principle by:

• Creating reusable prompt patterns for different types of learning materials

• Developing concept explanations that can be reused across courses

• Creating a library of learning objective prompt blocks that can be reused across assignments

• Establishing standard pattern for student engagement activities through a prompt block or prompt template

This usually means keeping prompt components as discrete as possible to more easily identify those repeatable units.

5. Centralized Management

At some point, these prompt blocks or modular components need to be stored somewhere with easy access for the entire team or organization … what most people are calling prompt libraries.

These spaces need to also allow for iteration. As prompts and AI models change, teams need to be able to adapt this library and make notes about their effectiveness.

A large technical documentation team might implement this principle by:

• Creating a searchable repository of proven prompt patterns in Confluence

• Developing a Google Doc with all their audience prompt blocks

• Developing a system for testing and updating different blocks in an excel worksheet

• Or even use a fancy prompt library tool, like Promptitude, as the AI system gets more complex

Through these examples, we see how structured prompting transforms ad-hoc efforts into systematic processes that support both rhetorical quality and operational efficiency.

The key insight remains: structure isn't about constraining creativity—it's about managing complexity to enable consistent, high-quality content creation at scale.

For my premium subscribers, I'll explore advanced implementations of these principles through detailed use cases and frameworks in the rest of this lesson. You'll learn how this adapts across different kinds of organizations, along with practical strategies for developing your own frameworks and prompts that scale.

Subscribe now

Read more

This worksheet accompanies Lesson 8: Principles of Structured Prompt Operations. You can cut and paste this into your own doc and fill in the spaces, or just add your answers to a separate sheet. If you need a little help … guess what? You can feed this to your AI and have it help you!

I’m also always available for consultation.

When we first approach str…

Read more

This is a subsection of Cyborgs Writing. If you are not interested in PromptOps, you can unsubscribe from this section in your Substack settings and still receive other content.

In our previous lesson, we explored how rhetoric shapes our interactions with AI, particularly through structured prompt blocks. One of these crucial blocks is [STYLE] - but this raises an intriguing question:

What exactly is style, and how does AI interpret and generate it?

Style has been a subject of debate among rhetoricians for centuries. While we often talk about style as something we "have," it's more accurate to think of style as something we "do" - an active process of adapting our communication to fit specific relationships and contexts.

Our style shifts naturally when we're:

• Writing an email to a colleague versus texting a friend

• Explaining a concept to experts versus newcomers

• Crafting formal documentation versus brainstorming ideas

This dynamic nature of style presents a fascinating challenge when working with AI. When we use a [STYLE] block in our prompts, we're essentially asking AI to perform a specific kind of linguistic adaptation.

This brings up a host of questions that I’m not sure can be entirely answered … but they can be tested.

How does AI interpret style directives? Can AI truly adapt to contextual nuances? Does AI have enough contextual awareness to make appropriate stylistic choices?

And most importantly: How can we guide AI's word choices to achieve different styles while maintaining coherence and purpose?

What is Style?

Style is one of the most complex concepts in rhetoric and writing. As Kate Ronald explains in her essay "Style: The Hidden Agenda in Composition Classes," style isn't just about word choice or sentence structure - it's about establishing a "presence" on the page, creating a genuine connection between writer and reader.

This is why combining explicit style guidance with AI tools, like temperature control, becomes crucial when working with AI, especially at scale when developing AI writing systems. Rather than expecting AI to understand style intuitively, we can use these tools together to guide its outputs more effectively.

In the creator economy, I consider this a key aspect of creating authenticity — or ethos. Whether you are a creator or working for a business, ethos is going to be a key to succeeding in a world where people expect authentic interaction, not just content.

And with the rise of AI, this ethos or authenticity will be even more difficult to achieve.

That’s why we need to take a closer look at style.

Ronald describes style as "somebody's home in this paper, somebody wants to say something—to me, to herself, to the class, to the community." When we ask AI to adopt a particular style through our prompts, we're really asking it to establish a specific kind of presence or relationship with its audience.

This is far more complex than just adjusting word choice or sentence patterns. Style emerges from the dynamic relationship between:

• The writer's voice and authenticity

• The reader's needs and expectations

• The context and purpose of the writing

• The conventions of different communities

As Ronald points out, even seasoned writing teachers struggle with teaching style explicitly because it's so deeply contextual. So I’m not sure why we expect AI to magically achieve this for us.

She notes that while we "recognize good style when we hear it," defining exactly what makes it good is much harder. This same challenge applies to working with AI - we can recognize when an AI's output hits the right stylistic notes, but explicitly instructing it to achieve that style is more complex.

This is where temperature settings become a valuable tool. Rather than relying entirely on direct instructions about style (which, as Ronald shows, is difficult even for human writers), temperature allows us to influence and fine-tune how the AI constructs its responses at a more fundamental level, especially in response to our [STYLE] blocks.

By adjusting the temperature, we can guide the AI toward different kinds of stylistic choices while still allowing space for the natural emergence of voice and presence that characterize effective writing.

Temperature: A Tool for Stylistic Control

Imagine you're at a coffee shop watching a master coffee blender develop specific flavors of coffee. This might mean having buckets full of different kinds of beans … a palette, so to speak.

Each time they make a blend, they need to select beans from these buckets. When making a standard house coffee, they'll consistently choose the most common beans - the ones they know will produce a reliable, familiar cup.

But for their "special choice," they might reach deeper into their palette and combine some unique or rare beans to create something more unexpected.

This is similar to how temperature works in AI language models. When generating text, the AI has a vast collection of possible word choices - like those buckets of coffee beans. For each word in a sequence, the AI assigns probabilities to different possible "next words." Some words are highly probable (like those common coffee beans) while others are less likely choices.

Temperature controls how the AI makes these word selections. Usually, this is a number between 0-2, with 0 being the most predictable. Some apps though don’t show this, but instead give you a “creativity” setting.

This is what the setting looks like in OpenAI’s playground - a place where you can create assistants and adjust many different kinds of settings (no subscription required).

At low temperatures (like our standard house coffee):

• The AI becomes very "conservative" in its choices

• It consistently picks the words with the highest probability

• It's like telling the blender "Just use the regular beans"

• This results in more predictable, consistent outputs

At high temperatures (like the coffee blender choice):

• The AI becomes more "exploratory" in its selections

• It's more willing to choose words with lower probabilities

• It's like telling the blender "Feel free to experiment"

• This leads to more varied, sometimes unexpected combinations

So while many AI interfaces label temperature as a "creativity" setting, it's really controlling the randomness in the AI's word selection process. The AI isn't becoming more creative in any human sense - it's simply being allowed to sample from a broader range of possibilities in its statistical model.

Rhetorical Approaches to Temperature

When Ronald talks about style in writing, she describes how writers need to adapt their voice depending on who they're writing for and why. Think about how differently you might write:

• An email to your boss versus a text to a friend

• A formal report versus a social media post

• A technical explanation versus a personal story

Each situation calls for a different kind of "presence" on the page. Ronald explains that good writers develop an instinct for matching their writing style to these different situations. They know when to be formal or casual, detailed or concise, technical or conversational.

This is exactly where temperature settings become valuable when working with AI. Instead of trying to explain all these subtle stylistic choices to the AI, we can use temperature to guide how the AI approaches different writing styles.

Understanding temperature this way helps us see why it can be such a useful tool for controlling style in technical writing or content creation. When we need precise, consistent documentation or instructions, we can use low temperatures to keep the AI focused on the most conventional, clear language patterns. When we want to generate multiple approaches to explaining a concept, we can raise the temperature to explore different phrasings and structures.

But, as we will see later in this lesson, this formula doesn’t always work, especially when you add your on style instructions to your prompt. Just like style is not a static, temperature too can only be used and understood in context.

To help us think through this, let’s take a style heuristic developed by ancient thinkers.

In the ancient world, rhetoricians recognized three distinct levels of style that speakers and writers could employ depending on their purpose. This can be a useful lens through which we think about temperature and style.

Plain Style was used for clear instruction and straightforward communication. This style prioritized precision and clarity over artistry. Think of military commands or legal documents - language that leaves no room for misinterpretation. Writers using plain style wanted their meaning to be immediately clear to their audience.

Middle Style balanced clarity with engagement. It allowed for some artistic flourishes and emotional appeal while maintaining clear communication. This style was perfect for everyday communication that needed to both inform and hold interest - like letters, essays, or public speeches meant to teach and engage.

Grand Style was reserved for moving audiences emotionally and inspiring action. This style employed elaborate language, metaphors, and other artistic devices to create powerful emotional effects. Orators would use grand style when they wanted to stir their audience's hearts as well as their minds.

These ancient insights about style seem to map well onto how we can use temperature settings with AI, but as you will see later in this lesson … it gets a bit more complicated. But it is a good starting point.

Just as ancient rhetoricians chose their style based on their purpose, we can adjust temperature settings to help AI generate content that matches our goals:

Low Temperature (0.2 - 0.4) aligns with Plain Style:

• Prioritizes clarity and precision

• Consistently chooses the most standard, conventional language

• Perfect for technical documentation, instructions, or any writing where accuracy is crucial

• Minimizes artistic flourishes in favor of clear communication

Middle Temperature (0.5 - 0.7) parallels Middle Style:

• Balances clarity with engagement

• Allows for some variation and creativity while maintaining focus

• Ideal for blog posts, user guides, or any content that needs to both inform and interest readers

• Creates a conversational yet professional tone

High Temperature (0.8 - 1.0) corresponds to Grand Style:

• Maximizes creative expression

• Explores more varied and unusual language combinations

• Best for brainstorming, creative writing, or content meant to inspire

• Produces more diverse and exploratory outputs

Note: Dialing temperature at the extreme ends, 0 or 2, usually takes the AI off the rails, though sometimes can producing interesting connections. For 0, you will see a lot of repetition. And for 2, you’ll see a lot of nonsense. Its actually a good way to see how temperature works.

This connection between classical rhetoric and AI temperature settings isn't just theoretical - it gives us practical guidance for choosing the right temperature for different writing tasks.

Just as ancient orators carefully chose their style based on their purpose and audience, we can adjust temperature settings to help AI generate content that establishes the right kind of presence for our purpose.

These ancient insights about style map surprisingly well onto AI writing, though not always in the ways we might expect. Through testing, I discovered that different styles often work best at unexpected temperature ranges

Let's look at some specific examples of how these different temperature ranges work in practice.

➡️ Paid subscribers can also see the prompt blocks I used in the Prompt Lab.

Read more

You can certainly use ChatGPT and other prompting tools to create new content, but AI's lack of creativity is well known.

Because LLMs are a guessing technology, they tend to produce predictable writing.

With work, though, you can get AI to produce more creative content, but that's not really the goal of this particular post. From a workload perspective,…

Read more

This is a free sample lesson from my premium PromptOps course, designed to help you master AI writing systems through structured approaches. The developing course is available exclusively to paid subscribers of Cyborgs Writing on Substack.

You can opt out of PromptOps posts in your Substack subscription settings without affecting your access to other con…

Read more

Welcome to this lesson on tokenization, a crucial concept in AI content operations. In our previous chapter, we explored the importance of prompts in AI-driven content creation. Now, we'll delve deeper into the structural elements that make these prompts effective, focusing on tokenization as the backbone of AI text processing.

Before really digging into…

Read more

In our first lesson, we explored the concept of AI Content Operations, a systematic approach to managing and optimizing the AI content creation process. Today, we're going to apply that same systematic thinking to a specific challenge that many of us face when working with AI for content creation: managing prompts.

If you've been using AI in your content…

Read more

Welcome back to our exploration of AI Content Operations. In our last lesson, we laid the groundwork by defining what AI Content Operations is all about. Now, we're going to dive deeper into one of the key technologies that powers this field: transformer technology.

If you're a tech writer or content developer, you've probably heard of transformer techno…

Read more

This is a free sample lesson for a course on understanding AI writing systems called PromptOps. If you would like to exclude this from your Cyborg’s Writing subscription, you can turn it off in your subscription settings.

How do you keep up with AI developments?

I get asked this a lot. My answer?

I don’t.

Why? It's not about “keeping up with AI.”

It's about strategically organizing your AI workflow, understanding key principles for framing your tasks, and focusing on relevant innovations.

Keeping track of every AI update will lead to information overload and inaction.

Let’s talk grocery shopping. I hate super-big grocery stores. Sure, I can find just about anything, but I am so overwhelmed by the choices. There are so many that I just want to get out fast. I give up and don’t make any choices.

Or let’s think about Netflix. Have you ever logged into Netflix to watch a movie, only to scroll mindlessly for an hour trying to make a choice?

How do you solve these problems? You need a plan before going to the grocery store or logging into Netflix.

It's the same in the world of generative AI.

You need an AI content operations plan.

What is an AI Content Operations Plan?

Let’s not make it complicated. The word operations is about understanding your workflow as a set of networked relationships between people, processes, and technologies. It's sometimes called a writing ecology.

Just as an ecological system is a network of interactions among organisms and their environment, a writing ecology is a complex system of interactions among ideas, prompts, and the content they generate.

Quite simply, thinking about the environments around our writing helps us shape our content in more detailed and intentional ways. We can let our yard grow willy-nilly … that is an ecology. But if we have an idea of what we want our yard to do, we can shape it to a specific purpose.

This is all people mean by operations...shaping the people, processes, and technologies around our content.

Think about grocery shopping in terms of operations. To start, you need to have a clear purpose. What do you want to accomplish? Save money? Prep for the Super Bowl? Cook a fancy dinner for your partner? Prep easy meals for a busy work week?

Then you need something in writing, like a list of what you need, acting as your plan of action or roadmap.

With this list, the number of items becomes manageable. You have a way to measure your trip's success by how much you spend or how prepared you are for your cooking goals.

If your goal is to save money, you focus on the price and ignore more expensive choices. If you're prepping for the Super Bowl, you might focus on the snack aisle or specialized displays.

You don’t have to know about everything in the grocery store.

In AI, a clear purpose and plan help declutter your workflow and intensify focus.

It doesn't matter how many innovations and advancements appear daily. With your roadmap at hand, you can focus on just the pertinent developments. This is the crux of an AI content operations plan.

Whether it's grocery shopping or incorporating AI into your workflow, you need to understand how those choices fit into the relational contexts of your life … or your specific ecology.

AI content operations involves understanding the relationships and dynamics among elements and leveraging them to create effective content.

An AI content operations plan is a structured approach to managing your AI writing workflow and environments. It's your grocery list. Your store map. The sales flier you picked up or coupon page you perused. The items you repeatedly buy every week.

It's about knowing the best way to shop for you.

Whether you're a blogger, content marketer, or technical writer, an AI content operations plan is crucial for understanding how AI does (or does not) fit into your workflow and using that knowledge to more strategically manage your work with AI.

Examples from Everyday Life

The phrase AI content operations might have conjured images of large tech companies with dedicated content creation teams. But the truth is, AI content operations plans are becoming valuable for anyone who regularly creates content, regardless of background or budget.

Here are some examples to give you a clearer idea:

• History Teacher: Mr. Lee wants to create engaging presentations for his middle school class. He uses an AI tool or prompt to find compelling historical images and videos, then another AI tool to generate a basic presentation script based on those visuals. He adds his own insights and humor to make it engaging for his students.

• Technical Writer: Sarah needs to write clear and comprehensive user guides for a new software program. She uses an AI tool or prompt to analyze existing user guides for similar software, identifying structures and best practices. She then uses another AI tool to generate draft content for specific sections. Sarah reviews and refines the AI-generated drafts to ensure accuracy and align them with the company's style guide.

• Social Media Influencer: Emily wants to grow her audience and engagement on Instagram. She uses an AI tool to track her past posts' performance, identifying the most popular content themes and hashtags. Emily then uses another AI prompt to brainstorm new post ideas based on those insights, tailoring the ideas to fit her unique brand and voice.

• Freelance Content Writer: David aims to increase his writing speed and content variety. He uses an AI tool to research keywords and generate outlines for his freelance writing projects. He then uses another AI tool or prompt to suggest different writing styles or tones based on the target audience, refining and expanding the AI-generated content to meet client needs.

• Corporate Blogger: Maria needs to produce high-quality blog content consistently. She uses an AI tool to stay updated on industry trends and competitor content and drafts blog post outlines based on current events or trending topics. Maria fills out the outlines with her own insights and expertise, ensuring the content aligns with her unique voice.

As we've seen with these examples, having a clear purpose is the cornerstone of any effective AI content operations plan. It allows writers to not only generate more useful AI-generated content, but also incorporate their own unique expertise and human voice at strategic points in the workflow.

If you want control over how technology shapes your content or the way you write, then you need a clear purpose and plan.

Remember the grocery store analogy? Don’t let the store designer or display manager dictate how you buy groceries. Go in with a plan.

This will also help with the overwhelm we all feel when trying to keep up AI tools (or brands of yogurt).

Just like having a shopping list keeps you focused on your goals, a clear content creation purpose keeps you from getting overwhelmed by the sheer volume of AI tools and features available.

You don't need to become an AI expert to benefit from AI.

By understanding your goals and implementing manageable AI steps within your workflow, you can unlock significant efficiencies and improvements in your content creation process.

But you have to know what you want to achieve.

Share

Teams Need to Play Together

Consider this from a team perspective too.

If each member of your content team or school is doing their own thing, without a clear collective roadmap, the result would likely be inconsistent content, duplicated work, and lots of missed opportunities.

That’s why an AI content operations plan is particularly crucial for teams and organizations: it ensures everybody moves toward the same goal, maximizing efficiency while also driving consistency throughout your AI-aided content

It also will save you money. Instead of buying unnecessary tools, or not buying any at all, you’ll be able to create the system that matches your specific plan.

That’s what this course is going to be about. PromptOps isn’t just about writing prompts … it’s about managing your AI content operations plan.

Right now, managing your prompts … or prompt operations plan … is the easiest and most accessible way to shape GenAI technologies.

In fact, many “AI tools” out there are just prompts in disguise. So understanding how to manage your prompting workflow saves you from even considering a whole host of AI tools flooding the market.

Why buy an AI tool if you can build it yourself?

So in the next lesson, we are going to dig into what I mean by a prompt operations plan and how you can develop a guiding purpose for how you manage your prompts … and your AI overwhelm.