Get more out of the AI your team already has.
The licences are paid for. The training happened. And the work that would actually benefit from it is still being done by hand.
What I actually do
I sit with the people doing the work and watch what their week is made of. Not a survey — the actual clicking, the re-typing, the file that gets mailed back and forth four times. Somewhere in there is the job worth handing to a machine, and it is almost never the one that got picked in the meeting.
Then I prove it on one workflow, in front of the people who own it. If it holds up, I build the thing that takes it over. If it doesn’t, you’ve lost a week instead of a year, and you know something about your own operation that no vendor was ever going to tell you.
How deep it goes is up to you
Three ways in, and most teams start at the shallow end. You can stop after any of them.
One workflow, looked at properly. That’s the Return Read— the smallest way to find out whether any of this is worth doing.
One real thing, built and put in front of the people who do the job. Not a demo — something that runs on your actual work and either earns its keep or doesn’t.
Some teams want somebody in the room while this becomes normal. A standing day, an open line, and the same person who built the thing still answering for it.
What I won’t do
- I don’t sell a platform, so I’ve got no reason to point you at one.
- I won’t build something you could fix with a conversation and a shared document. The smallest thing that solves it is the right thing, even when it costs me the build.
- I won’t hand you a strategy and leave. Whatever gets recommended, I’m the one who has to make it work.
- I won’t tell your team that AI is coming for their jobs to get a signature.
Six-plus working applications in four months. All still running.