How I work
How a project starts, how the work gets checked, and what you’re left with at the end.
Starting
Every piece of work starts with a conversation about what you’re trying to do, not which tool to use. Sometimes the honest answer is to change a process, buy something off the shelf or not build anything yet.
A small first step
Before anything big, I do a small piece of work: a review, a prototype or a written plan. It shows whether the bigger piece is worth doing, and whether we work well together.
Building
I use AI to build faster, and I check what it produces the same way I’d check an engineer’s work: reviewed, tested and understood before it goes anywhere near your customers or data.
Handover
You own what’s built. The code lives in your accounts, with notes your team or another developer can follow.
Principles
- No AI for its own sake.
- Checking is the hard part now.
- Build it so you can leave.
- Say the uncomfortable thing early.
FAQ
Where does our data go?
That’s one of the first things we agree. I’ll tell you which services would see your data, and where it would be stored, before anything is built. If a task can run on models you host yourself, that’s an option too.
Will we be locked in to you?
No. You own the code and the accounts it runs in, and it’s written to be picked up by someone else.
Who owns the code?
You do, once it’s paid for.
Is it safe to let AI agents act in our business?
It can be, if they’re given narrow jobs and limited access, and you keep a person on the decisions that matter. I’d rather start small and widen what an agent can do once you trust it.
What about security?
Agents get the least access they need, credentials stay out of the code, and what they do is logged so you can check it.
What if AI isn’t the answer?
Then I’ll say so. A better process or a simple tool is often the right call.
Can you work with our existing team or supplier?
Yes. I can review their work, help them adopt AI tools or build alongside them.