What I've built

A small number of jobs, done properly. Here's what the problem was, what I built, and where each one stands — including the ones still in build.

Validera

LiveOutsourced internal audit

A recommendation tracker built around how they actually audit.

The problem
An internal audit practice lives or dies on follow-through. Findings get raised, owners get assigned, dates get agreed — and then it all sits in spreadsheets and email chains, with someone spending their week chasing the status of things rather than doing audit work.
What I built
A tracker built around their methodology rather than a generic project tool: recommendations, owners, due dates and evidence in one place, following the way their engagements actually run.
Where it stands
The chasing becomes a system rather than a person’s memory, and the state of every open recommendation is visible without anyone assembling a report to find out.

Chase Plant Hire

In buildPlant hire

Replacing the off-the-shelf hire software, not decorating it.

The problem
Like most hire firms, they run on plant hire software built for a generic version of the industry — so the yard works around the system rather than in it, the same details get keyed more than once, and a monthly licence goes out for software that fits nobody exactly.
What I built
A bespoke replacement: their processes, their terminology, their rules, with the AI in it from day one rather than bolted on afterwards — bookings taken over the phone, paperwork prepared from the hire itself, and the retyping designed out.
Where it stands
In build. The commercial shape is the point: the build cost is low enough with modern AI tooling to roll into a monthly fee in the region of what they already pay their current vendor.

Tender response agent

In buildBidding

Reads the ITT, drafts the response, checks it against the marking scheme.

The problem
Tender responses eat weeks. Someone reads a two-hundred-page ITT to find every question, works out what the evaluators are actually marking, then rewrites answers the business has already given well a dozen times before.
What I built
An agent that extracts every question and the marking criteria from the ITT, drafts answers grounded in your library of previous responses, and runs them back against the criteria before submission.
Where it stands
In build. A person still owns and signs off every answer — the agent removes the reading, the hunting and the first draft, not the judgement.

Want to see one running?

Here's one of the solutions on its own — the inbox one, reading an enquiry and drafting the reply for someone to check and send. It's a single piece of what goes into a full build, but it's the quickest way to see how the AI behaves.

See it work →

The pattern underneath

None of these started with someone asking for AI. They started with a business paying for software that didn't fit, and a team quietly absorbing the difference — in spreadsheets, in double entry, in one person's head.

The reason I can rebuild rather than patch is cost. Modern AI development tools have changed what a bespoke system takes to produce, which means the build is often small enough to roll into a monthly fee — usually in the region of what you're already paying for the software you're replacing. That's the part most owners don't know has changed.

Is yours the same shape?

Tell me what you're running, what it costs you a month, and where your team works around it. I'll tell you honestly whether replacing it makes sense — and if it doesn't, I'll say so.

Start a conversationHow the pricing works →