Roami
Due diligence. AI enablement. Fractional leadership. Board and coaching.
The premise
I led AI adoption across an 80-person product, design and engineering organisation before I advised anyone else on doing it. That was MOO, where I spent eight years and finished as CPTO with a seat on the executive team, built the product management function from scratch, and cut turnover by 40%.
Before that, the UK Civil Service, where I took release cycles from eight months to eleven days across a 200-person group. Consulting, Fujitsu and HP before that. And nine years in the British Army as an Information Systems Engineer, running classified infrastructure for multinational headquarters.
I still build. The tools I use in the work are public. I speak regularly at conferences, roundtables and panels, and serve as a trustee on a charity board.
Recent work
Diligence audit on a construction tech SaaS business for a European venture investor. Architecture, team shape, delivery capability, and the risks that do not surface in a data room.
Discovery across the whole business, then a bounded plan to act on it. I mapped the opportunities against end-to-end value streams, separated what was ready from what was blocked, and named the dependencies that had to be owned before anything else could move. It landed as an eight-week programme with named owners and finishable outcomes.
Cross-functional AI training. What capable actually looks like at each level, how to assess it honestly, and how to make it stick outside the engineering team.
Enabling the parts of the business that engineering-led AI programmes tend to skip, including training for product managers and designers.
Built in the open
A team of specialist AI agents for Claude Code. The current version trades surface area for gate discipline and a learning loop that actually fires, enforced by hooks and CI rather than prose.
Run AI coding agents against a local model inside a kernel-level sandbox. Private, offline, no API cost. Built for the first question a regulated client asks.
The shared context layer a team’s AI runs on, as a worked and forkable bundle for a synthetic open-finance company.
Side-by-side model and prompt comparison with auto-scored quality dimensions. Built because most teams put AI output in front of real decisions with no shared way to evaluate it.
The real cost of an AI feature: model comparison, margin analysis, and pricing. The question that arrives a month after the pilot everyone was excited about.
Validate a product idea before building it. Structured discovery, confidence scoring, a research swarm, and a technical handoff deck at the end.
What I offer
Two or three days a week, embedded, in the exec team. I make the calls, not the recommendations.
Discovery across the business, then a bounded plan with named owners. Training for the people who aren’t engineers.
For investors: what you’re buying, what it costs to fix, and what the team can actually ship.
Monthly calls and async in between, for whoever is holding product and engineering.
Hire a tech leader with someone who’s been one.
Quarterly board engagement. Technology risk in language the rest of the board can act on.
1:1 for product, engineering and design leaders. I’ve held the job they’re in.
Contact
If any of the above sounds like the conversation you need to have, email me and we’ll set up a call.
claire@roami.group