I Built an AI Coach for Leaders Under Pressure. Here Is What I Learned.
At sixty metres underwater, a diver does not get smarter.
The brain gets louder.
Every dive at depth teaches you the same thing: the environment does not care how experienced you are. What matters is whether you can read the signal when the noise is at its worst. Can you act on it? Can you do that before the pressure takes over?
I spent twenty years underwater before I spent twenty years in boardrooms. The parallels are uncomfortable.
Most leadership problems are not intelligence problems
Senior leaders are not short of information. They are short of a place to think clearly when the stakes are real.
The typical response to a hard decision is to get more input. Call another meeting. Commission another report. Bring in a consultant. What that usually produces is more noise dressed up as data.
The real decision — the one the leader already half-knows they need to make — gets buried under the weight of process.
I have watched this happen across 52 countries. A founder who knows they need to let someone go but keeps finding reasons to delay. A managing director who can see the structural problem clearly but cannot say it out loud. A leadership team that is collectively brilliant and collectively unable to move.
This is not a failure of intelligence. It is a failure of environment. Nobody gave them the room to think.
Why I built an AI coach
When I left my corporate role, I had spent years watching pressure distort decision-making at every level of leadership. I wanted to build something that did what the best human coaches do: create a private space where a leader can say the actual thing, hear it reflected back, and find the signal underneath the noise.
I chose AI deliberately. Not because it is clever. Because it is available at 11pm on a Tuesday when the decision cannot wait, it does not have political skin in the game, and it does not need to protect a relationship with the board.
The goal was never to give better advice. The goal was to ask better questions.
What the AI actually does
Most leaders who try The Deep Brief expect it to validate them. It does not.
The coaching model is built on a simple principle: roughly 20% of a session should feel comfortable, 50% should create mild friction, and 30% should be genuinely uncomfortable but necessary. That is not a number I invented. It is what happens in the best human coaching relationships.
The coach who makes you feel good about a bad decision is not coaching you. They are protecting their contract.
The AI asks the question you have been avoiding. It names the assumption you have been protecting. It does not let you use process as a hiding place.
One user described it as "the conversation I needed to have with myself but could not have alone." That is close to what I was aiming for.
What I learned building it
A few things became clear that I did not expect.
Leaders do not want to be told what to do. They want to be heard in a way that helps them hear themselves. The job of the coach is not to provide the answer. It is to create the conditions where the leader can find it.
The moment a leader names the real decision, the pressure usually drops. Not because anything changed in the world. Because clarity is less frightening than ambiguity, even when what becomes clear is difficult.
Trust is the entire product. A leader will not be honest in a coaching session unless they believe nothing they say will be weaponised. That applies to AI too. The architecture has to match the promise. The ethics of confidentiality are not a feature. They are the foundation.
Can AI replace a human coach?
No. And I am a sceptic of anyone who claims otherwise.
A skilled human coach brings something that no AI currently replicates: the accumulated pattern recognition of sitting with hundreds of leaders under pressure, the ability to read the silence, the presence of someone who has been there.
What AI coaching can do is something different and genuinely valuable. It can be there when the human coach is not. It can serve the founder who cannot yet afford senior coaching but who carries decisions that deserve proper attention. It can provide the thinking space between sessions.
The right framing is not replacement. It is access.
Where this is going
The platform is live. The individual product works. The organisational layer — which surfaces decision patterns across leadership teams without exposing individual sessions — is in development.
The early signal is that this matters more than I expected. Not because leaders lack good thinking. Because good thinking requires a specific kind of space. And most organisations do not provide it.
Pressure does not create leaders. It reveals them. The question is whether they have somewhere to think before the moment of truth arrives.