AI development consulting
I build software with AI every day, including the products on this site. Plenty of teams now have the tools and a nagging doubt: the agents are fast, the output looks right, and nobody's quite sure what they've signed up for.
That's the gap I work in. Here's what I do, with the caveat that every engagement starts with a free scoping chat, because your codebase isn't the demo.
Who it's for. Teams building software who want AI to do real work in their codebase and their product, without handing it the keys.
What I do
Coding agents that work in your codebase
Choosing the tools, setting up the workflow, and writing down the rules: what the agents may touch, how their work gets planned and reviewed, and what "done" means.
Checking what the agents hand back
A review of AI-written code and the process around it. Agents are confidently wrong in predictable ways (I wrote them up as AI TRAPS), and most of it can be caught before it ships.
LLMs in your product
Working out where a model adds real value to your software, and integrating it so the product still behaves when the model misbehaves.
AI governance
Practical rules for safe and responsible use: data, secrets, review, and who's accountable for what the AI produced.
AI Team Power Up
A structured programme for a development team that knows AI matters but doesn't yet share a view of how to use it.
- How the models actually behave. Enough of the mechanics to predict where they fail, using the TRAPS framework.
- Coding assistants, used well. Workflows that keep code quality and oversight while the team speeds up.
- AI in your products. A practical plan for adding AI features to the software you already have.
- Supervised agents. Where agents earn their keep today, and how to keep a person in the loop.
The evidence
I'd rather show you than tell you:
- Click Dog: open-source ClickHouse observability. It's run in production for months and is now in public beta.
- AsOf: a local archive and debrief of AI coding sessions, in development.
- AI TRAPS: five ways AI goes wrong, and what to do about each.
We enlisted the help of Colt Consulting to conduct thorough benchmarking of our software architecture and practices. Colin's expert advice was highly valuable and provided us with actionable insights to identify both our strengths and areas for improvement. We highly recommend Colt Consulting to anyone seeking a top-tier software coach and advisor.