From AI-cautious to AI-sponsor in five weeks
A senior M&A partner went from watching from the sidelines to leading AI strategy for the practice.
- Services
- Executive AI coaching, one to one
- Period
- Feb – Mar 2026
- Participant
- One person: the Head of Integrated M&A for the firm's advisory practice
A partner who had decided AI mattered, and nothing else
The Head of Integrated M&A for the firm's advisory practice had concluded that AI would change how his team worked. He had not concluded what to do about it. He asked to be put through a foundations course for executives rather than sit through another vendor demonstration.
His starting point was specific. He was, in his own words, struggling to get my head around
server hosting and the role of GitHub. He wanted to understand the tools well enough to make decisions about them for the practice. He did not want to become a developer.
This was Stratum's first client engagement. It was sold as a fixed-fee program of three modules and a checkpoint.
Teach the tools, then build something real
The program was three 90-minute modules a week apart, delivered one to one, followed by a fortnight of independent building and a checkpoint session. We called it AI Foundations.
Two rules shaped it. No module ended in theory: each one set homework that produced something working on his own machine. And the last module was not a demonstration. He scoped and built an application for a live problem in his practice, with us in the session, then kept building alone before the checkpoint.
Every session was recorded. He received the recording, a transcript and a written summary with the homework set out, so the material was still usable a month later when he came back to it.
Three modules and a checkpoint
The prerequisites were a Claude Max subscription and a virtual server of his own, both in place before the first session so that nothing in the program had to be imagined.
- Module 1 — Foundations
- The model landscape, a framework for choosing a model inside a corporate environment, and the administrative work that could be lifted on day one.
- Module 2 — Intermediate
- Settings and memory, connectors to mail and files, reusable skills, chat against agent work, and Claude Code installed on his own laptop. Homework: a first working dashboard.
- Module 3 — Building
- Scoping and building a deterministic application with Claude Code, against a real problem in his practice rather than a sample one.
- Checkpoint
- A fortnight of building on his own, then a session on keeping scope small and the path from a local build to something hosted and shareable.
He builds his own tools now
From the first session to a working prototype of an M&A transformation platform, built with Claude and shared with his own client team.
He agreed to a public testimonial within a month of the first session.
He started publishing his own articles on AI, and sent a draft over for review before posting it.
Introductions he made to people in his own network between February and August, alongside a joint pitch document the two firms took to market together.
The practice followed the individual. Three months later his wider team asked for a session of its own.
The program was written up as a standard offer and has since been delivered to other clients.
I've been blown away by how quickly Stephen has been able to improve my understanding of this rapidly evolving space. I've stopped watching from the sidelines worried I am falling behind and have started making real decisions about how to best implement AI in our practice.
Have a problem shaped like this one?
Fixed-scope discovery to start.
Strata: Ashes Sitoula — Unsplash