Why the best AI coaching
has a human in the loop.

The strongest version of the practice isn’t either-or — it’s both, layered deliberately.

By Dennis Willis — 20-year executive coach, ActionCOACH · 30-year software founder and CIO · Builder of Wayfinder AI

There’s a frame most writing about AI coaching gets wrong in the first paragraph.

The frame is that this is a horse race. AI coaching versus human coaching. Which one will win, which one will replace the other, which side you’re on. The cheerleaders say the AI is just as good already. The skeptics say nothing replaces a real person. Both camps spend most of their energy arguing about the wrong question.

I’ve spent twenty years as an executive coach. Before that, thirty years building software companies. I’ve spent the last two years building an AI coach. So let me be direct about what I’ve found.

The AI isn’t replacing the human. The human isn’t making the AI obsolete. What’s actually happening — and what the most serious practitioners on both sides are quietly discovering — is that these two things are different tools for different jobs inside the same work. And the strongest version of the practice uses both.

Here’s how that works, and why it matters.

What the AI is genuinely good at.

Three things, and it’s not close.

One. Engagement depth.

A working executive can afford a human coach for maybe four hours a month. On a good month. The rest of the time — the 716 hours between sessions when the actual situations are happening — they’re on their own. The hard conversation at 3 PM. The decision that has to be made by morning. The spiral that starts Sunday at 9 and doesn’t resolve until they fall asleep exhausted.

That’s 99.4% of their life where the coaching isn’t there. Most of which is where the coaching would actually matter.

An AI coach available by voice, chat, text, or email — on whatever cadence fits the person’s life — fills that 99.4%. Not as a replacement for the four hours with the human. As the other 716.

The depth of engagement this produces is beyond what was possible before. A person talking to Wayfinder for as little as five minutes at a time — or as long as the moment requires — is accumulating structured coaching conversation across the week. Someone averaging thirty minutes a week builds 26 hours over a year. Nearly 80 over three. A human coach delivering that at $600 an hour would cost $15,600 annually. At $49.00 a month, the AI delivers the same volume for a fraction of the price.

The AI isn’t cheaper coaching. It’s more coaching per dollar — by a wide margin.

Two. Continuity and memory.

A human coach remembers what you told them last session. Maybe. If they’re good and they took notes.

An AI with a well-built memory architecture remembers every conversation you’ve ever had with it. Your quest from six months ago. The ally you named a year ago and haven’t mentioned since. The pattern you were working on last quarter and walked away from. The exact phrase you used three weeks back that you’ve been quietly avoiding ever since.

A human coach can catch up. An AI never forgot.

This isn’t a replacement for human relationship. It’s a thing that has never existed before: a coaching presence that holds the whole arc of your journey, across years, in specific detail, always available to be queried.

Over time this becomes something radically different from what a human coach can offer, because no human holds that much context about another human in working memory. Your therapist doesn’t. Your spouse doesn’t. You don’t, about yourself.

Three. Cost structure.

Coaching at the executive level has historically been rationed by price. $300 to $800 an hour means coaching is available to the top couple percent of professionals. Everyone else gets books and podcasts and whatever peer support they can cobble together.

AI coaching at $49.00 to $349.00 a month moves the price point from thousands per year to hundreds. That means a middle manager, a founder pre-Series A, a parent trying to do right by their kids, a teacher, a nurse — all of them now have access to a coaching practice that used to require a C-suite title or enough savings to fund it out of pocket.

This matters because the research on meaning in life — Rogers et al. 2024 in the Journal of Personality and Social Psychology, Frankl sixty years before that, all the altruism and flourishing research in between — doesn’t apply only to executives. It applies to humans. A coaching approach that was previously available only to people who could afford $10,000 a year is now available for less than what most people spend on streaming services.

That’s not a small thing. For the first time, people who could never afford a coach have a real one — not a chatbot, not a self-help app, but a structured practice with measurable effects on how they experience their own lives.

What the AI is not good at — and what the human coach does that it can’t.

This is the part most AI coaching companies skip, and it’s why they lose trust with serious readers.

The AI cannot see you.

Not in the embodied sense. A human coach sitting across from you can see your shoulders drop when you mention your father. The AI doesn’t know your shoulders exist. The AI doesn’t know that your voice got smaller when you talked about the promotion. It doesn’t catch the micro-hesitation before you said “I’m fine.” The pattern-recognition an experienced human has for the difference between what someone is saying and what they’re actually communicating is not in the AI. Not yet, and possibly not ever in the way it matters.

The AI doesn’t catch what happens in your life.

A human coach seeing you across years notices that you’re handling the same situation differently now than you did in March. They have the whole arc of you in their memory and can pull back to see the shape. The AI has all the data — arguably more — but pattern recognition at the life-arc level is still better done by a human with judgment than by an algorithm with memory. The AI can tell you what you said. The human can tell you what’s happening to you.

The AI is not in the room.

Being in the actual room with another person who is committed to your growth is doing something. We don’t fully understand what that something is — there’s research on mirror neurons, on co-regulation, on the simple fact that humans evolved to be present with other humans — but the effect is real. A human coach across a table from you is contributing something that isn’t in the transcript. An AI on a phone, however good, is not in the room.

And the default AI will flatter you — unless it has an external standard to hold you to.

Every large language model in 2026 is sycophantic by default. The training process that makes them useful also makes them accommodating, eager to please, prone to tell you what you want to hear about yourself. A user asking “how am I doing?” gets affirmation. A user describing a bad decision gets it reframed charitably. This is the single biggest failure mode of generic AI coaching, and it’s why most AI “coaches” on the market today are flattering chatbots dressed up.

The way Wayfinder sidesteps this isn’t through better prompting alone. It’s structural. The Hero’s Journey framework gives the AI an external standard that doesn’t come from the user’s self-report. Wayfinder isn’t evaluating whether you feel good. It’s evaluating your actions toward other people through a defined lens — are you serving or harvesting, are you building allies or collecting followers, are you crossing thresholds or retreating to comfort, are you helping someone else become a hero or performing one yourself. Those questions have answers that don’t depend on whether you want to hear them.

This is the architectural reason the framework matters. Without it, even a well-prompted AI drifts toward agreement with the user over time. With it, the AI has something to hold you to that isn’t your own mood.

A built-in mechanism for confrontation: the Hard Questions.

Wayfinder ships with thirteen hard-question prompts you can fire any time, as often as you want. “Tell me what I’m avoiding.” “Name the gap between what I say matters and how I’m actually showing up.” “Where am I performing instead of serving.” One click. The AI answers using everything it knows about your journey, held to the framework.

Here’s what that looks like in practice. A real exchange — me, in my own product, this week:

Notice what’s happening there. The AI isn’t being “direct” in the abstract. It’s directing the confrontation through the framework’s own language — Inmost Cave, performer loop, hero loop — terms I’ve been working with for years and that the AI knows I know. A generic chatbot saying “you seem to be avoiding something” is noise. A framework-trained AI saying “you’re at the Inmost Cave and pretending it’s a scheduling problem” is coaching.

This is what the framework buys, structurally. Not just an external standard — a shared vocabulary for confrontation that the user has already consented to and learned. The standard is external. The AI is just applying it. The user has already agreed it’s the standard worth applying.

A human coach can refuse to be nice to you when being nice isn’t what you need. The Hard Questions feature gives the AI a principled way to do something similar, on demand, repeatedly, at 2 AM if that’s when you need it. Not as good as the human confrontation. Better than the flattery. And importantly, it’s the reason a human coach working with the same framework amplifies the AI’s work rather than contradicting it — you’re both pointing at the same external standard, applied to the same life.

A human coach has timing, mood, fee structure, scheduling. The Hard Questions are there when you’re ready, not when the calendar says it’s time.

These are not small things. They’re the things that make the combination work — and they’re the reason the framework matters as much as the AI does.

Why the combination is the point.

Now put the two together and notice what happens.

The AI handles the detail, the continuity, the daily practice, the availability at 2 AM and 3 PM and the moment before the hard conversation. It holds your whole journey in structured memory and can speak to you as someone who knows where you’ve been and where you’re going. The Hero’s Journey framework keeps it honest. The Hard Questions let you summon confrontation on demand.

The human pulls back periodically — quarterly, monthly, whatever cadence works — and does the things the AI can’t. Sees you. Catches the lie you don’t know you’re telling. Names the pattern across years. Sits in the room. Refuses to flatter when flattery is what you’re asking for — and reads when you’re asking for it without saying so.

Neither one is doing the other’s job. Each is doing what it’s uniquely good at.

That combination isn’t a compromise between two inferior options. It’s a better version of coaching than either alone, delivered at a price point most professionals can actually afford.

What this looks like in practice.

A few patterns we’ve seen or recommend.

Pattern one. AI daily, human quarterly. Wayfinder picks up a few minutes here, a longer session there, across the week — as little or as much as the moment calls for. A human coach — ours or someone you already work with — meets with you every quarter for a real session. The human session works better than it otherwise would because you’ve been doing the work between sessions, and the AI sessions work better because the human has helped you name what to pay attention to.

Pattern two. Keep your current coach, add the AI. Many of our executive users already have a human coach. They don’t fire them. They layer Wayfinder in for everything between sessions. The coaching conversations they have with their human get richer because they arrive with more groundwork done. The human can focus on what they’re uniquely good at. The coaching investment works harder.

Pattern three. AI only, for now. For people who can’t access a human coach — because of cost, geography, availability, or readiness — Wayfinder alone is a serious tool. Not everything a full combination can be, but a real coaching practice available to a population that has never had one. If and when human coaching becomes available, the foundation is already built.

All three patterns are legitimate. None is the “right” answer; each fits different lives, different budgets, different seasons.

The question this actually answers.

The writing about AI coaching in 2026 is mostly stuck arguing whether AI is as good as humans. That framing misses the point.

The better question is: what does a coaching practice look like when the tools of 2026 are used well, honestly, and at their highest leverage?

And the answer, as far as I’ve been able to find it after twenty years of human coaching, thirty years of building companies, and two years of building the AI, is: you use both. You use the AI for what the AI is uniquely good at — depth, continuity, availability, cost, confrontation on demand. You use the human for what the human is uniquely good at — embodied presence, reading you when you can’t read yourself, pattern recognition across years, being in the room. And underneath both, you use a framework that gives the coaching something to point at beyond the user’s own mood. You don’t pretend any one of them is the whole answer.

The framework keeps the practice honest. The AI makes it available. The human makes it embodied.

Together they’re more than any one alone. That’s the thesis. That’s what we’ve built. And it’s why Wayfinder’s strongest tier isn’t the AI-only one — it’s the one that includes human coaching touchpoints, because that’s where the research and the decades of practice point.

We’re trying to be better humans. We do need to rub up against other humans in the process — even if the tool we use daily is one modeled on being a better human, researched carefully, and applied well.

The framework is real. The tool is real. So is the need for the human. Use all three.