September 18, 2026

AI for the Over 40 – Week 34: The Four Skills Behind Every AI Conversation

I am designing a hands-on AI workshop for business leaders. Ninety minutes. Most of the audience had never opened an AI tool. The goal: get them from zero to literate in the time it takes to watch a movie.

My first instinct was that 33 weeks of documented AI learning gave me more than enough material. If anything, the problem would be cutting, deciding what to leave out of a series that had covered everything from basic prompting to diagnostic frameworks to building custom tools.

So I sat down with AI and started designing.

The first couple of iterations were fine. Solid structure. Logical flow. Good exercises. The kind of workshop you would expect from someone who has been writing about AI literacy for eight months.

But something felt off. It did not feel like what would have helped me on day one of my journey.

That is when I tried something I probably should have done from the beginning. Instead of designing forward from workshop objectives, I asked AI to go backward through my own journey.

Using AI to audit your own AI journey

I loaded the context of my full series and asked AI to do something specific: identify the skills I had actually picked up along the way, not the concepts I had written about, but the working skills I demonstrated across 33 weeks. Then I asked it to flag gaps, skills that were critical to accomplishing the work I had described but that I had never named, and skills I was not using that could improve my results.

What came back was not a generic checklist of things AI can do. It was a catalog of capabilities traced through my own experience, skills I could point to specific moments where I had learned them, specific articles where I had used them, and specific failures where I had discovered them.

And it was humbling.

A couple of things jumped out immediately.

First, there were skills I use in literally every AI conversation that I had never once named as teachable skills in 33 weeks of writing. I demonstrate iterative refinement constantly, treating every AI output as a first draft I shape through dialogue, but I had never called it out as a distinct skill someone needs to learn. I just did it, and assumed readers would absorb the technique by watching.

Second, and this one stung, there were simple, practical techniques I rarely use myself despite knowing they improve results. Specifying the format you want the output in, for example. I know this works. I have seen how much more effective it makes the output. And I almost never do it. It is the AI equivalent of knowing you should stretch before running and skipping it every time.

The four skills that actually matter

When I sorted everything, 33 weeks of learned skills, documented gaps, and instinctive habits I had never named, the inventory kept consolidating. Dozens of individual techniques, but they clustered into four fundamental capabilities that everything else builds on.

Think of these as layers. Each one depends on the one before it. You cannot evaluate output you have not learned to shape. You cannot manage a working conversation if you accept whatever comes back. And you cannot think strategically with AI until the mechanics of conversation become second nature.

Skill 1: Set the Stage. Everything that happens before AI generates its first response. Context, roles, examples, format, and knowing how to break big problems into smaller ones.

Skill 2: Read It Like a Skeptic. Everything that happens after AI responds. Evaluating output against your own expertise, spotting confident nonsense, recognizing when AI is telling you what you want to hear instead of what you need to hear.

Skill 3: Work the Conversation. The rhythm of real AI work, iterating on output, managing context, knowing when to push further and knowing when to start over.

Skill 4: Think Bigger, Think Better. Using AI for strategic thinking and blind-spot detection while knowing what to protect, what to verify, and when to override your own framing.

That is it. Four skills. Everything I have written about for 33 weeks either fits inside one of these or builds on top of them.

Now let me unpack what I actually found in each one, including the gaps I discovered in my own practice.

Skill 1: Set the Stage

This is where most AI education starts and stops: write better prompts. But when I mapped the full set of skills that actually shape what AI gives you, prompting was only part of it.

The foundational shift, the one everything else depends on, is understanding that you are having a conversation, not issuing a search query. That is what I wrote about in Week 1 when my son’s advice changed everything, and what I expanded in Week 3 with the Master Improv Actor framework. Most people approach AI the way they approach Google: type a question, get an answer, move on. Until that pattern breaks, nothing else in the inventory matters.

But “it’s a conversation” is only the starting line. The skills that actually set the stage go further.

Giving context is the thinking-partner shift: telling AI who you are, what you are working on, and why it matters before asking your question. Not because AI needs your biography, but because who you are shapes what useful advice looks like. I covered this in Weeks 1 and 3, and it is the core of everything I teach about Shift 1. This one I got right.

Assigning roles is the Improv Actor concept from Week 3. Give AI a character to play. “Act as a process improvement consultant” produces fundamentally different output than asking the same question without a role. Covered well.

Then the gaps start.

Providing examples of what good looks like is one of them. I touched on this in Week 28’s production-prompting work, but I never named it a fundamental skill. It is one. When you show AI an example of the writing style you want, the analysis format you need, or the level of detail you are looking for, the output quality jumps immediately. It is the difference between telling someone “write me something professional” and handing them a document and saying “like this.”

Specifying format and constraints is the gap that surprised me most, not because I did not know about it, but because I know it works and I still skip it. Tell AI you want a bulleted list, a three-paragraph summary, a table with specific columns, a response under 200 words. The output improves dramatically. I have scattered this technique across the series without ever elevating it to the fundamental skill it is. It might be the single most practical thing you can teach someone in their first hour with AI, and I had been treating it as too simple to be worth naming.

Breaking big problems into smaller ones is the last gap in this cluster. My diagnostic framework from Week 16 does this implicitly; the whole “Review of Systems” approach is about decomposing vague frustration into examinable components. But the general technique of breaking a complex request into steps before asking AI to execute is a transferable skill I never taught as such. Your first instinct is to dump a massive question on AI and hope for the best. The skill is learning to sequence: first help me understand the problem, then help me evaluate options, then help me draft a solution.

Skill 2: Read It Like a Skeptic

This is the skill that separates people who use AI from people who use AI well. And it is where I discovered the gap between what I covered conceptually and what I practice daily.

Week 4 addressed AI limitations. Week 3’s “Yes, And” principle explained why AI tends to agree with you. Week 31 explored what happens when human experts hallucinate with the same confidence as AI.

The concepts are there. The practical skills need work.

Spotting confident nonsense is the first one. AI does not hedge when it is wrong. It delivers fabricated information with the same tone and structure as accurate information. Week 31 addressed this philosophically, the idea that humans hallucinate, too. But the practical skill of catching fabrication needs more attention than I gave it. There is a specific list of things AI is reliably bad at: basic math, counting items in a list, knowing about events after its training data ends, and, critically, admitting when it does not know something. I never compiled that list in the series. I worked around these limitations instinctively. I did not teach anyone else to do the same.

Detecting sycophancy goes beyond knowing that AI agrees with you. It is the active skill of noticing when AI is flattering your ideas instead of examining them. Given everything I wrote in Week 30 about the Knowledge Trap, how having more information does not translate to persuasive power, and Week 31 about the Expertise Paradox, this skill connects directly to the most recent arc of the series. If AI confirms your assumptions because it is architecturally inclined to agree with you, and you accept that confirmation because it feels like validation of your expertise, you have built a feedback loop that makes you more confident and less accurate at the same time.

The practical counter-skill is to make it a habit to ask AI to argue the other side, find weaknesses in your reasoning, and tell you what you might be missing. I do this reflexively now. If someone had taught me to do it explicitly in my first week, it would have saved me months of accepting mediocre output because AI told me it was great.

Checking your own framing is the deepest version of this skill, and it is the one I had not fully articulated until now. AI reflects your framing back to you. If you describe a problem as a technology problem, AI gives you technology solutions. Describe the same problem as a people problem, and you will get entirely different recommendations. The pushback exercises from earlier weeks address this partially, but the core skill, pausing to ask whether your framing might be shaping the answers you are getting, is distinct from just pushing back on AI’s output. You are pushing back on your own input.

Skill 3: Work the Conversation

If Skill 1 is how you start and Skill 2 is how you evaluate, Skill 3 is how you work, the rhythm and mechanics of real AI collaboration over the course of a conversation, across conversations, and across platforms.

This cluster contains the skill I have most consistently demonstrated without ever naming it.

Iterative refinement might be the most important unnamed skill in the entire series. Every experienced AI user does it: treat the first output as raw material, then shape it through follow-up instructions, corrections, and redirections. “That is good but make it more concise.” “Keep the structure but change the tone.” “The second section is strong, rebuild the first to match it.” I do this in every article I write, every framework I develop, every client deliverable I produce. Thirty-three weeks of demonstrating it. Zero weeks of calling it out as a skill someone needs to learn.

Context management, summarizing where you are mid-conversation, resetting when things drift, I covered in Week 8’s second pillar. Carrying context between conversations through copy-paste workflows and cross-platform transfer has been a recurring theme. Those parts of the map had labels.

Knowing when to start over did not. This is a real gap. I never addressed the skill of recognizing that a conversation has gone sideways and that a fresh start will be more productive than ten more rounds of refinement. There is a moment, experienced users know it, where the AI has accumulated enough confused context that it is working against you rather than with you. The right move is to open a new conversation, paste in a clean summary of what you actually need, and start fresh. I do this regularly. I never mentioned it.

Knowing when to stop is the twin of knowing when to start over. Diminishing returns are real. I have burned more time than I would like to admit refining output past the point of meaningful improvement, running a piece of writing through too many rounds when a few rounds was probably ninety percent of the way there. The skill of recognizing “good enough” is particularly important for busy leaders who could easily spend two hours polishing something that needed twenty minutes. Nobody teaches this because it is not exciting. But it might save more time than any other skill on the list.

Skill 4: Think Bigger, Think Better

The first three skills are about working with AI. Skill 4 is about working with AI, using it as a lens to see things you could not see before, while maintaining the judgment to protect yourself and your organization.

Using AI to find problems you have normalized is the diagnostic work from Weeks 15 and 16, the unlocking questions, the seven barriers framework, and the “Review of Systems” approach. It is the category of AI use that I believe has the most transformative potential, not solving the problems you already know about, but revealing the ones you have stopped noticing.

Using AI to pressure-test your own ideas is one I use constantly, but never named as a transferable technique. Before I present a proposal, share a recommendation, or publish an article, I run it past AI as a skeptical audience. “What is the weakest argument here? What am I assuming that might not be true? If someone wanted to tear this apart, where would they start?” It is a rehearsal for every difficult conversation and is available to anyone with a subscription and the habit of asking.

Think Before You Research is the technique from Week 26 where you design your approach collaboratively with AI before deploying it to expensive deep research modes. This is a strategic skill that saves both time and money and represents the kind of second-order thinking that develops naturally after the first three skills become habitual.

Then the safety layer, which I would argue should be taught on day one, even though it only fully makes sense after you have been working with AI long enough to feel tempted to skip it.

Knowing what not to share comes from the security tiers in Week 2. This is non-negotiable and straightforward, but people need to hear it early and often: not all AI platforms protect your data equally, and what you share in a free tier conversation might be used for training.

Understanding that AI output is a draft, not a deliverable, is the editorial board concept from Week 5. AI gives you raw material. You give it judgment, verification, and accountability. The moment you submit AI output without review is the moment you have delegated your professional reputation to a system that cannot be held responsible for what it produces.

Your Week 34 Challenge: Audit Your Own Skill Gaps

1. Rate yourself honestly against the four skills. Not “have I heard of this,” that is awareness, not literacy. The question is: do you do this consistently, without thinking about it, in your actual work? Set the Stage, Read It Like a Skeptic, Work the Conversation, Think Bigger, and Think Better. Where is your biggest gap between knowing and doing?
2. Try one unnamed skill this week. Pick the simplest one: tell AI exactly what format you want before you ask your question. “Give me this as a three-paragraph summary” or “Respond in a table with columns for pros, cons, and my suggested action.” Notice what happens to the output quality.
3. Ask AI to audit you. Share a few recent AI conversations and ask: “What skills am I using well? What am I not doing that would improve my results?” It is the same exercise I did with my series, scaled down to your own practice. The answers might surprise you.
4. Name one skill you use but have never articulated. Everyone who works with AI regularly has at least one technique they have developed instinctively but could not explain to a colleague. Name it. The act of naming makes it transferable to your team, your organization, or your own intentional practice.

The Bottom Line

Thirty-three weeks into this series, I thought I had a pretty good map of AI literacy. Building a workshop that had to fit into 90 minutes revealed that my map had skills I had never labeled and routes I had never drawn, not because I did not travel them, but because I had stopped noticing the turns.

That is not a failure of the series. It is how practitioners actually learn, by watching someone do it before anyone names what is happening. But demonstrating a skill and teaching a skill are not the same thing. The workshop forced that conversion: you cannot teach someone a skill in 90 minutes if you cannot name what the skill is.

Everything I have learned clusters into four capabilities: Set the Stage, Read It Like a Skeptic, Work the Conversation, and Think Bigger and Think Better. They build on each other in that order. And the gaps I found were not in the advanced skills, they were in the simple, practical habits that experienced users do without thinking. Specifying format. Providing examples. Knowing when to stop. Knowing when to start over.

Simple enough that they feel unworthy of an article. Foundational enough that everything else builds on top of them.

If you have been following this series, you have been watching many of these skills in action for months. Now you have names for them. And naming something is the first step to teaching it, to your team, to your organization, or to yourself on the days when you know better but skip the stretch anyway.

This post is part of my “AI Over 40” series. It first appeared on LinkedIn: The Four Skills Behind Every AI Conversation

---

Read more AI and Copilot blogs.

Trending Posts