There’s an enormous amount AI could do for small firms, and almost none of it is reaching them.
Nearly everything written about this assumes you’ve got an AI team. Most firms haven’t, and won’t. That’s the gap I wanted to work in.
Three things I believe
Teach-first, not built for you.
Value shows up when a whole team can use AI well, not when one enthusiast can, and not when an agency has built something nobody in the building can change. Your team gets the skills to build and adapt it themselves. This is about getting more out of the AI you already pay for, not learning to code.
Aim before you automate.
Most people automate the nearest thing rather than the right thing, and end up running a bad process faster. So we work out where AI actually counts, redesign the work where it needs it, and keep the parts that should stay human.
Measurement and memory are both mandatory, and they’re different things.
Measurement is being able to see and steer it: what it costs, whether the quality holds, whether it’s genuinely working. Memory is the thing remembering — corrections from your team, and real signals from the business about what actually happened — so the work gets better on its own rather than staying where you left it.
How I work
I use this in my own business first.
I’m deep in these tools daily, and my own AI work is instrumented and measured, so when I say measurement matters it’s because I do it rather than because it sounds good.
I keep up with more than the tools.
What’s worth knowing isn’t the next feature release, it’s how AI-native companies are actually running: how they’re structured, how they sell, how the work gets done differently. That’s what filters down to firms like yours next.
Every session changes the next one.
The questions people ask and the parts that don’t land are how the material improves, which matters when there’s only one of me.
And if I don’t think a day will pay for itself, I’ll tell you before you book it.
Background
AI work since 2018
Starting at Capgemini: a machine learning stress-detection system for air traffic control rooms, and a model tracking play from live rugby broadcast.
Improving services at HMRC
Helped establish a cross-department innovation capability and led teams improving live services. One project halved case-processing time from 28 to 14 days.
Built AI capability in-house
Then innovation director at an Oxford algorithms company, where I set up its AI consultancy arm and ran the internal innovation work on the company’s own processes.
Master’s in Computer Science and AI
With distinction.