AI coaching - why a human who has built something beats a chatbot every time

AI coaching is a human - someone who’s built real products, run real companies, and taught real courses - helping executives figure out how to actually use AI. Not a chatbot. Not a platform. Not an app that asks you how your day went and spits back affirmations. A person who’s been in your shoes and can tell you what works, what doesn’t, and what’s a waste of your time.

If you Google “AI coaching” right now, you’ll mostly find software platforms. BetterUp, Rocky.ai, CoachHub - all selling AI as the coach itself. That’s fine for some things. But if you’re a CEO trying to figure out how AI changes your business, a chatbot isn’t going to help you navigate board expectations, team resistance, or the gap between buying tools and actually using them.

Most AI coaching is a chatbot pretending to care

The search results for “AI coaching” are confusing on purpose. Platforms want you to think AI coaching means AI doing the coaching. There’s a PLOS ONE study comparing AI coaching to human coaching for goal attainment - and the AI chatbot actually performed comparably to human coaches on structured goals over a 10-month trial. Interesting finding. But the study looked at generic goal-setting, not strategic business decisions.

Here’s where it falls apart. A chatbot can’t sit in your leadership meeting and notice that your VP of Operations is terrified of being replaced. It can’t tell you that the AI vendor you’re evaluating has a terrible integration story because it’s seen three other companies try the same thing. It can’t push back when you’re about to spend six figures on a tool your team will never adopt.

It takes a CEO to coach a CEO. Someone who’s never managed board expectations, made payroll decisions under pressure, or shipped a product from a blank screen to paying customers - that person can’t meaningfully coach another executive on AI strategy. They can read you a checklist. That’s not coaching.

When I teach MBA students at the OneDay MBA program or work with executives through WashU’s Skandalaris Center, the questions that matter most aren’t about technology. They’re about judgment. Should we build or buy? How do I know if my team is ready? What if this doesn’t work? Those are questions a chatbot literally can’t answer because they require context, experience, and the willingness to say “I don’t know, but here’s how I’d think about it.”

The coaching industry has grown to over $5 billion globally according to the ICF’s research. That growth is driven by executives who need specialized guidance - not generic advice. And the intersection of coaching and AI is where the biggest gap exists. Plenty of people can teach you prompt engineering. Very few can help you think about how AI changes your competitive position, your team structure, or your product strategy.

Three types of CEO and where they get stuck

After working with manufacturing and technology executives on AI adoption, a clear pattern has emerged. There are really only three types of CEOs when it comes to AI, and coaching looks completely different for each one.

The enthusiast. This CEO uses AI personally all the time. Huge fan. They’ve tried every tool, read every blog post, and can’t understand why the rest of the company doesn’t get it. The problem isn’t that they lack enthusiasm - it’s that they confuse personal productivity gains with organizational transformation. They overestimate what tools can do out of the box and underestimate how much change management is actually required. Coaching for the enthusiast means giving them structure, helping them prioritize, and honestly telling them when they’re moving too fast for their organization to follow.

The skeptic. This CEO hasn’t used AI meaningfully. They’ve heard the hype, they know they “should” be doing something, but they genuinely don’t know where to start. Many are intimidated but won’t admit it - especially in front of their board or leadership team. Coaching for the skeptic is about creating safe, low-stakes wins. Show them one thing that works for their specific job - not a demo, not a pitch deck, an actual use case they can try in the next ten minutes. Once they experience it firsthand, skepticism tends to dissolve pretty quickly.

The delegator. This is maybe the most common and most frustrated type. They see the potential. Maybe they even use AI themselves. But the real problem is the rest of the company. They bought ChatGPT Enterprise licenses for everyone. They ran a workshop. Nothing changed. The delegator doesn’t need more AI knowledge - they need a cascade strategy that reaches middle management, because that’s where adoption actually breaks down.

McKinsey’s State of AI survey found that 53% of C-level executives regularly use gen AI at work, compared with 44% of midlevel managers. That gap might not sound huge, but it’s the difference between a CEO who’s convinced AI matters and a middle management layer that hasn’t caught up. The executives who engage with AI earliest are often the furthest from understanding why their teams haven’t followed. That gap is exactly what coaching is designed to close.

What AI coaching looks like in practice

The reason most AI advice fails is that it’s abstract. “Use AI to be more productive” isn’t actionable. Here are specific use cases that I coach executives through - things that create immediate, visible value.

Competitor research. Not “ask ChatGPT about your competitors.” Structured analysis. Upload your competitor’s last three earnings calls, their job postings from the past six months, their product changelog. Have AI identify patterns - what they’re hiring for tells you what they’re building. What they’re not talking about tells you where they’re struggling. This kind of analysis used to require a consulting engagement. Now a CEO can do a meaningful version of it in an afternoon, once they know how to set it up.

Meeting and board prep. Record your leadership meetings. Transcribe them. Summarize with action items. This alone saves hours per week. But the real value is in board meeting preparation - feed AI your last three board decks, the current financials, and ask it to identify the questions your board is most likely to ask. Then prepare answers. I’ve seen executives go from dreading board meetings to feeling genuinely prepared, which changes the entire dynamic.

The self-interviewing technique. This is the use case I find most underappreciated and it’s something I coach executives through directly. Record yourself answering 20 to 30 questions about your business, your opinions, your management philosophy. Speak naturally - don’t script it. Transcribe those recordings and analyze them for your voice patterns, your vocabulary, the rhythm of how you actually talk.

Then build what amounts to a voice profile. When AI drafts your emails, your LinkedIn posts, your internal memos - it references that profile. The result is communication that sounds like you wrote it at your best. Not like a committee. Not like ChatGPT’s default tone. Like you, on a good day, when you’ve had enough coffee and enough time to think.

This matters enormously for executive credibility. People can tell when something was AI-generated. They can tell when the CEO’s weekly update suddenly sounds nothing like the CEO. The self-interviewing technique solves that problem.

Context is everything. One principle I emphasize in my AI courses and in coaching: the 80/20 rule of AI usage. Eighty percent of getting good output is providing good context. Twenty percent is the actual question. Most executives do the opposite - they ask a brilliant question with zero context and wonder why the answer is generic.

Tools like Claude Projects let you build persistent context about your business. Upload your strategy docs, your org chart, your product roadmap, your competitive analysis. Then every conversation starts from a place of understanding. The difference between a generic AI response and a genuinely useful one is almost always about context, not about which model you’re using.

Working alongside AI, not just prompting it. The most effective pattern I’ve seen isn’t “give AI a task and wait for the result.” It’s collaborative. Think out loud with it. Push back on its suggestions. Ask it to challenge your assumptions. Claude, Gemini, ChatGPT - the specific tool matters less than how you use it. The executives who get the most value treat AI as a thinking partner, not a search engine. That’s a skill, and it’s one that coaching develops much faster than self-study.

For the practical foundations of working with these tools, I wrote a detailed guide on prompt engineering that covers the iterative approach.

Turning coaching into company-wide change

Here’s the biggest failure mode in executive AI coaching: it stays with the executive.

The CEO gets comfortable with AI. Great. They can draft emails faster, analyze competitors better, prepare for board meetings in half the time. But the other 200 people in the company are still doing everything the old way. The AI licenses the company bought - and every mid-size company seems to be buying them - sit unused. Zylo’s SaaS Management Index found that over half of all purchased software licenses go unused across the average organization. AI tools are no exception.

This isn’t a technology problem. It’s a capability building problem. And it’s almost always caused by the same bottleneck: middle management.

Middle managers have the most to lose from AI adoption. Their value often comes from being the person who knows where things are, who summarizes information upward, who translates between the executive team and the front line. AI threatens to automate a meaningful chunk of that. So they resist - not openly, but through inertia. They don’t block AI adoption. They just never quite get around to it.

The cascade model for coaching works like this: CEO learns and builds conviction. CEO coaches their direct reports - not on the technology, but on the strategic thinking behind it. Direct reports establish office hours and support structures for their teams. Middle management sees the executive team actually using AI, not just talking about it, and the social proof starts to work.

This cascade is what turns AI coaching from a personal development exercise into an organizational capability. In building Tallyfy for over ten years, I’ve watched this pattern play out with every type of technology adoption. The fractional AI executive model is one way to formalize this cascade without hiring a full-time executive.

The alternative - and I see this constantly - is the CEO who gets excited, sends a company-wide email about AI, maybe runs a single workshop, and then wonders why nothing changed three months later. That approach fails for the same reason most consulting engagements fail - it treats adoption as an event instead of a process.

What to look for in an AI coach

Not everyone who calls themselves an AI coach is worth your time. Here’s a framework for evaluating whether someone can actually help you.

Technical foundation matters. Can they explain how the tools actually work? Not just “here are ten prompts that work well” but the underlying architecture. Why do large language models hallucinate? What’s a context window and why should you care? What’s the actual difference between models? If your coach can’t explain these things in plain language, they’re going to give you advice that works today and breaks tomorrow. My BSc in Computer Science and years building Tallyfy from the first line of code isn’t a credential I wave around - it’s the reason I can explain why something works, not just that it works.

Business experience is non-negotiable. Has your coach built something? Have they run payroll? Have they sat across from an investor and explained why the numbers aren’t where they should be? The strategic questions that matter in AI coaching - build versus buy, timing of adoption, organizational readiness - are business questions, not technology questions. Someone who’s only ever been a consultant will give you consultant answers. Someone who’s been a founder will tell you what actually happens when theory meets reality.

Teaching ability separates good coaches from mediocre ones. Can they meet you at your level? I teach three structured AI courses - one for founders, one for SMBs, one for schools - and the content is completely different for each audience. A good coach adapts their approach based on whether you’re an enthusiast who needs guardrails, a skeptic who needs proof, or a delegator who needs a cascading strategy. That’s not something you get from a generic AI workshop.

Coaching is ongoing, not a one-time event. The executives who get the most value from AI coaching are the ones who maintain a relationship over time. Not a six-month engagement with deliverables and milestones. More like a flexible advisory where you can call when you have a decision to make, when a vendor is pitching you something you don’t understand, when your team pushes back and you need to figure out why. Monthly or biweekly sessions that evolve as your understanding grows. I offer exactly that kind of flexible advisory - the format adapts to what you need, not the other way around.

The irony of AI coaching is that the technology is the easy part. The models work. The tools are good. They’re getting better every month. The hard part is getting a room full of executives to admit they don’t understand something - and then helping them get past that in a way that actually changes how their company operates. That’s not something you can automate.