A few years ago, while studying how behavior change interventions scale, I ran into a statistic that stuck with me. Executive coaching, the kind backed by decades of research and practiced by credentialed professionals, runs as much as $500 an hour, and a full engagement costs $25,000 or more a year. The math excludes most people before the conversation even begins.
The debate around ai vs human coaching usually assumes you’ve experienced both and are choosing between them. That’s the wrong starting point. Most professionals, especially early-career managers and individual contributors, have never had a coach of any kind. Not a bad one. Not a cheap one. None at all.
So the comparison that matters isn’t “which is better?” It’s “what actually gets coaching to the people who need it, in a form that changes their behavior?”
That’s what I want to work through here, from a behavioral science perspective rather than a vendor pitch.
Why Coaching Works at All (The Part Most Comparisons Skip)
Before comparing delivery methods, it helps to understand what makes coaching effective in the first place. Three mechanisms do most of the work, and none of them are specific to coaching.
Spaced repetition. Practicing a skill across multiple real situations over weeks produces far better retention than cramming concepts into a single session. This is well-established in learning science and applies directly to workplace skills like giving feedback or running difficult conversations.
Feedback loops. You try something, get a response about what worked, adjust your approach, and try again. Without this loop, people either repeat mistakes or default to old habits. The loop needs to be tight. Waiting 30 days between iterations is too long for most skill-building.
Accountability. Having someone (or something) check in on whether you actually did the thing you said you would do. Self-motivation works for about a week. After that, people need an external prompt.
Monthly coaching sessions with a human coach are structurally limited on the first and third mechanisms. Thirty days is a long gap between practice points. And between sessions, there is no one asking whether you tried the approach you discussed. This isn’t a criticism of human coaches. They are often excellent at the second mechanism, providing deep and perceptive feedback. But the format has real constraints that the format itself cannot fix, regardless of how skilled the coach is.
What AI Coaching Actually Changes
Cost gets all the attention when people discuss AI coaching. It’s the obvious difference. But three other shifts matter more for whether coaching actually changes your behavior.
Daily cadence instead of monthly
If you need to have a difficult conversation with a direct report tomorrow morning, a coaching session scheduled for next Thursday doesn’t help. You need to think through your approach now, while the situation is fresh and the stakes feel real.
AI coaching offers that. A conversation when you need it, not when your coach has an opening in their calendar. Before a performance review. After a team meeting that didn’t go well. During lunch when you’re replaying a conversation and wondering what you should have said differently.
The behavioral science is clear on this: for skill development, cadence matters more than the depth of any single session. Ten short coaching conversations spread across two weeks will change your behavior more than one brilliant 90-minute session followed by silence.
The psychological safety paradox
Start with what has actually been measured. The Conference Board reported in October 2025 that AI can provide up to 90% of day-to-day coaching functions, with human expertise still needed for the conversations that are emotionally charged, political, or values-based. What that research does not touch is disclosure, and disclosure is where the more interesting difference sits.
People tell an AI coach things they won’t tell a human one.
The manager who won’t admit to their executive coach that they’re struggling with a direct report will say it to Merlin. The new team lead who doesn’t want to appear weak in front of a senior coach will describe their real confusion to an AI. This isn’t because the AI is better at empathy. It’s because the AI removes the social risk. There’s no judgment, no reputation effect, no worry about what the coach thinks of you.
For skill-building conversations, that honesty gap changes the quality of the entire engagement. A coach working with incomplete information gives incomplete advice. When people share what’s actually happening, the coaching conversation can address the real problem.
Measurable practice, not self-reported progress
Traditional coaching relies heavily on self-report. The coachee says “I tried the feedback technique and it went well.” The coach takes that at face value because there’s no other data source. But self-report is unreliable. People overestimate their progress, forget the details of what they tried, and unconsciously frame things to look good for their coach.
AI coaching can track patterns that self-report misses. Message frequency, topic shifts, which skills a person gravitates toward and which they avoid, the progression of how they describe similar situations over time. None of this requires surveillance, though it is exactly the kind of visibility a responsible AI coaching framework has to govern. It comes from the coaching interactions themselves.
At Risely, we see an average 26% improvement in targeted skills within 12 weeks, measured across 5,000+ users in 40+ organizations. That number comes from the coaching data, not from post-program surveys where everyone checks “satisfied.”
What Human Coaching Still Owns
I want to be straightforward about where AI coaching falls short, because overselling it helps no one.
The moments that need a human
Some situations call for a coach who has lived through something similar and can draw on that experience in ways an AI cannot.
Career inflection points are one. “Should I leave this company?” “Should I take this promotion even though I’m not sure I want it?” These questions involve identity, values, risk tolerance, and personal context that a human coach processes differently than an AI. A good coach who has made a similar career bet can share what they considered, what surprised them, and what they wish someone had told them. That lived wisdom matters.
Interpersonal crises that involve deep emotional processing are another. A conflict with a co-founder. A relationship with a manager that has turned toxic. A layoff you need to lead while managing your own feelings about it. These situations need more than structured skill practice. They need someone who can sit with complexity and hold space for emotions that don’t resolve in a single conversation.
Senior organizational dynamics are a third category. Navigating board relationships, managing up during a restructuring, building political capital in a new executive role. The coach’s network, pattern recognition from similar situations, and understanding of organizational power matter here in ways that AI doesn’t replicate.
Why that matters less than you think, for most people
The gap in the coaching market is not “who coaches the CEO.” Plenty of coaches serve that segment, and the economics work because the client (or their company) can pay $500 an hour.
The gap is the other 95%. The new manager who just got promoted and has never managed anyone. The individual contributor who needs to get better at cross-functional communication. The team lead who keeps avoiding conflict and knows it’s holding them back.
For these people, the question isn’t “is AI coaching as good as human coaching?” It’s “is AI coaching better than no coaching?” Set against a baseline of nothing, that is not a close call.
A person who has never had a coach doesn’t need to start with a $500-an-hour executive engagement. They need structured, accessible practice on the specific skill that’s holding them back. They can get that today.
The differences between AI coaching and human coaching, side by side
Laid against each other, on the dimensions that decide whether a behavior changes rather than the ones that decide the invoice:
| What you are comparing | Human coaching | AI coaching |
|---|---|---|
| Time from decision to first session | Weeks, after matching and scheduling | About 5 minutes from signup with Merlin |
| Session frequency | Typically 2 to 4 sessions a month | On demand, including the evening before a difficult conversation |
| Support between sessions | Usually none until the next appointment | Daily nudges; 73% of Risely users engage with them |
| How progress is measured | Self-report and the coach’s notes | Skill assessments before and after; average 26% improvement in 12 weeks at Risely |
| What the person discloses | Often edited, because the coach is a person whose opinion carries weight | Often more candid, because there is no reputation at stake |
| Who actually gets it | Roughly the top 5% of employees | Everyone in the organization who is invited |
| Typical cost | $500 an hour, $25,000 or more a year per person | $59 per user per month at Risely |
| Strongest on | Succession, high-stakes transitions, board and executive dynamics | Feedback, delegation, difficult conversations, other repeatable workplace skills |
| Weakest on | Long gaps between practice points | Problems that need lived experience at the top of an organization |
Can AI coaching replace traditional executive coaching for workplace development?
No, AI coaching cannot replace traditional executive coaching, because real executive coaching is a small and specific practice: succession, a first C-suite seat, a founder handing over the company, a leader whose next decision affects a few thousand people. The value in that room comes from a coach who has personally operated at that altitude and can say what they got wrong there. Merlin has never sat in a board meeting where a chair was quietly losing confidence in the person being coached.
The more useful observation is about the label rather than the technology. Most of what is sold as executive coaching is not executive coaching. The phrase drifted down the org chart over the last decade until it covered a six-session engagement for a newly promoted team lead, priced at executive rates because the word was on the invoice. That work is worth doing. It is manager development: teaching someone to give feedback, to delegate without taking the task back, to hold a conversation they have been avoiding for three weeks, to run a one-on-one that is not a status update. Whether it lands has very little to do with the coach’s seniority and almost everything to do with how many times the manager practices.
That is the work AI coaching displaces, and the mechanism explains it better than the price does. A new manager delegates something, gives a piece of feedback, or handles a disagreement several times in an ordinary week. A six-month engagement at two sessions a month gives them twelve supported repetitions. The skill either forms in the distance between those two numbers or it does not form.
So the answer splits in two. For the genuine article, hire a person and pay what it costs. For the much larger volume of work that borrows the label, AI coaching is not a cheap version of executive coaching. It is a different intervention aimed at a different mechanism, and it starts this afternoon rather than after a matching process, which for a manager with a difficult conversation scheduled on Thursday is not a small detail.
How to Decide What You Need Right Now
Instead of asking “AI or human?”, ask yourself three questions about the challenge you’re actually facing.
Is your challenge about building a specific skill? Giving better feedback. Delegating without micromanaging. Having difficult conversations. Running productive one-on-ones. If yes, AI coaching is well-suited. These are skills that improve through repetition, and daily practice beats a monthly conversation for building them. You can start with a leadership skills assessment to identify which skill to focus on first.
Is your challenge about processing a major career or personal decision? Whether to change roles, how to handle a relationship with a difficult manager, whether you’re in the right career. If yes, find a human coach. You need someone who has been where you are and can share perspective that only comes from experience. The value here is judgment, not repetition.
Is your challenge about a specific organizational dynamic? Office politics, restructuring, managing stakeholder relationships at the senior level. If yes, find a human coach with relevant experience. Context matters more than frequency here, and the coach’s understanding of how organizations work at that level is the value you’re paying for.
Most people land in that first bucket and don’t realize it. They think they need a mentor or a therapist when what they actually need is structured practice on a specific skill, with feedback, over 8 to 12 weeks. The skill gap is concrete and fixable. The challenge is getting access to something that helps you close it.
“But I’m not sure which bucket I’m in.” That’s normal. Start with the skill. If you spend two weeks working on delegation with an AI coach and realize the real issue is that you don’t trust your team because of a deeper organizational problem, you’ve learned something useful. You now know what kind of help you actually need. That clarity alone is worth the time.
Start With What’s Available to You
If you’ve read this far, you’re probably someone who wants to get better at your work but hasn’t had access to coaching before. That’s exactly who AI coaching was built for.
Try Merlin and see what AI coaching actually feels like. Pick a skill you’re working on, have a few conversations, and decide for yourself whether the daily cadence makes a difference. Merlin works on the web, inside Slack, and inside Microsoft Teams, in 40 languages.
If you’re evaluating coaching options for your team rather than yourself, we’ve built a detailed comparison framework that covers pricing, feature differences, and a decision matrix for organizations.