A coach built an AI agent in an hour. His client used it alone at 3am — and got his life back within a week. Two sides of the same AI coaching story.
Most stories about AI coaching come from one side of the table: the founder explaining what their product does, or the researcher explaining what the data shows. This one is different. It comes from both sides of the same interaction, recorded independently.
Maxime Bui is an organizational coach. Two years ago, working with a small company, he noticed a pattern repeating across very different people: a decision paralysis that had nothing to do with lacking options, and everything to do with not being able to weigh them. He had a method for this already, a clarification tool he'd used for years in career assessments. So he did something most coaches haven't yet done: he built it into an AI agent, and put it directly in his clients' hands.
One of those clients, a 26-year-old engineer, later sat down separately to describe what it was actually like to use it, on a Sunday night, alone, working through a decision that had quietly stalled his entire life for months.
Below are both stories, in their own words, and what I think they teach us, together, about what AI coaching is actually for.
Maxime's account is, in effect, a short course in the discipline required to turn a coaching method into something that works without you in the room.
Confidentiality can stay completely intact. One of Maxime's clients used the agent to work through a purely personal issue, one Maxime still doesn't know the content of. The client's manager independently confirmed the impact: productivity and engagement fully restored. Maxime's own framing stuck with me: "I make tools available, but I don't know what the topic was. The coaching contract is respected." The coach guarantees the process. He was never entitled to the content.
A powerful agent takes an hour to build, not a month. Once the intention behind the tool is genuinely clear, the technical build is fast, sometimes under an hour with experience. The real investment isn't in the technology. It's in knowing precisely what you're trying to do and why.
A well-designed agent makes clients autonomous. Before the agent existed, clients who hit a familiar wall had to call Maxime back and re-explain. Now they can revisit the same structured process on their own, whenever the situation recurs. Autonomy, not availability, is the actual measure of a coach's impact.
Clarifying a decision usually means removing noise, not adding options. In one case, a manager convinced he faced a hard choice realized, a few questions in, that the decision had already been made. The other "options" weren't real alternatives at all. Clarity rarely comes from more information. It comes from identifying what was never truly relevant to begin with.
The agent isn't the one doing the work. The client is. Maxime is blunt about this: "It's the client who works, not us. We offer a process, we guarantee the process." This is also why he warns against over-generalizing: a tool built for a broad, generic use case adds little value. A tool built for a specific, well-mastered problem does the opposite.
"Choosing means giving something up" can be turned into motivation instead of paralysis. Maxime's method never compares option A to option B in isolation. It always adds a third comparison: staying exactly where you are, weighed honestly. Making the cost of inaction visible is often what finally creates enough tension to move.
Building an agent is, first, an act of reflexivity for the coach. You cannot encode a vague instinct. You have to name the trigger, the transition, the intention behind every question you normally ask on autopilot. The side effect of building the tool is a forced audit of your own practice.
Purpose-built AI coaching tools produce real mastery, not generic answers. Maxime tested this directly: he reviewed clients' prior conversations with generic AI tools and found the underlying process was never actually controlled, only reconstructed after the fact. A method-driven agent, by contrast, knows where it's going before the conversation starts.
A well-structured agent can put a first step of coaching within reach, 24 hours a day, for anyone. Access has always been coaching's structural bottleneck: budget, scheduling, proximity to a coach. Availability alone isn't enough, but paired with a real process, it changes who gets to take that first step at all.
In the end, technique is only ever in service of intention. This is the thread running through everything Maxime describes. The technology never does the thinking for you. It executes whatever intention you gave it, clearly or vaguely, well or badly.
The client's account, recorded separately, describes the same tool from the inside, and confirms almost every one of Maxime's points without having heard them.
You don't always talk to your friends about your real problems, and it isn't distrust. "There are things so intimate, so complex, that aren't even clear to me yet. I'm not going to call my friend, because the thing isn't even clear to me." The block wasn't social. It was that the thought had no shape yet to be shared.
At 3am, when overthinking loops the hardest, no human coach is available. The tool is. "Try finding a coach at 3:30am. Every coach on the planet is asleep." This isn't a knock on human coaching. It's a real structural gap that availability alone can close.
Generic AI tells you what you want to hear, and builds nothing. "It just throws the answer you want to hear right at you. Three messages later, it's already dropped everything, because in that moment you just wanted to hear that, so it follows you. There's nothing built. No depth, no foundation." He'd used ChatGPT before. The comparison, made from direct experience, is exactly what distinguishes a model-based coaching process from an unstructured chat.
We live in a world where it's harder and harder to feel understood, simply because everything moves too fast. The same critical thinking that makes someone sharp at work gets carried into personal life too, useful for decisions, exhausting for keeping pace with everything at once.
A conclusion can be shattering and calming at the same time. "You really feel like you're learning something about yourself, but also a kind of relief, because it's like your unconscious has been sounding an alarm about everything you were taking for granted, without you consciously knowing it." Whatever you call the mechanism, something that was already true, but never named, finally surfaces.
"Making the implicit explicit" is the actual mechanism, and a single number can trigger it. Asked to weigh a factor, "does it weigh 10 kilos, 20, 50?", he froze completely. Unable to produce a number. The agent offered a hypothesis instead: "let's say, based on what you've told me, it's 7 kilos." From that single placeholder, every following number came naturally. He wasn't given an answer. He was given something to react to, and that was enough.
The impact was immediate and verifiable. He'd skipped three months of German classes. The week after his conversation with the agent, he showed up again. He started going out more. "The change in my quality of life was immediate." Not a feeling reported days later. A behavior, resumed within a week, confirmed independently by his manager.
What looks like procrastination is often a missing structure, not a missing will. "It felt impossible. I had the feeling of attacking the question and never reaching a satisfying answer. So in the back of your mind, you tell yourself you're procrastinating." Too many criteria held in the head at once isn't laziness. It's an information-processing problem waiting for a framework.
The real luxury isn't having an answer. It's not being forced to have one right away. "There's really no pressure. You feel like you're moving forward, and there's also this calm of knowing that at any moment in my life, I can pick this back up. And even if that moment never comes, I still feel good about my life." AI coaching's most underrated feature might not be intelligence. It might be the absence of a deadline.
Video of Maxime's coaching client sharing his perspectives.
Read on their own, either account is a good testimonial. Read together, they describe a single, closed loop, and that's what makes this case study worth more than the sum of its parts.
Maxime's discipline on the build side (guaranteeing process without demanding content, designing for a specific problem instead of a generic one, treating the agent's design as an act of reflexivity rather than a technical task) is precisely what makes the client's experience on the other side possible: the freedom to speak without managing anyone's reaction, the absence of pressure to perform, the sense that a real method, not an algorithm guessing what feels good, is holding the structure.
My own research on AI coaching keeps landing on the same pattern from a different angle: participants describe being able to "speak about emotion without the emotional background of the other person," and that absence of judgment, engineered deliberately rather than accidental, is often what lets people go further than they expected. What Maxime built and what his client experienced is that same finding, lived out in a single real-world case, end to end.
Thanks to Maxime Bui for building this with the rigor he describes, and for letting both sides of this story be told.
