Jeff Bradshaw

CTO, board director, and someone who still ships code.

I founded and sold an integration company to RELX, then ran global technology at Proagrica and Telus Agriculture — 60 and 80 product portfolios, teams of 500 and 1,000+. I now build AI systems inside an FCA‑regulated bank, and chair Fotenix, an AI agri‑tech business.

I write here about the problems I actually solve — mostly where messy data, applied AI and commercial reality collide.

See the work Read the writing

ROUNDUP 5L resolves to GLY‑540 ×5 ✓ matched
24 hours becomes under 5 minutes
£60.00 per task becomes £0.10 per task
14 disconnected systems becomes one reconciled ledger

Thirty years of getting two systems to agree on the same fact — through EDI, through SOA, through the cloud, and now through AI.

Track record Perwill · DHL · Equitas Adaptris — founded, sold to RELX Proagrica / RELX Telus Agriculture Oxbury Fotenix AgGateway Forbes Technology Council

Selected work

Problems, and what happened to them

Three that show the range: agentic AI in production, the unglamorous data work underneath it, and taking a company from founded to acquired.

£60 → 10p Automated financial statement spreading for lending decisions. A day of accountant time became five minutes — and the back book now re-reviews itself as new filings appear. Oxbury Bank Read the case study → Same product, seven names Entity resolution and unit-of-measure normalisation behind the rebate platform — the boring layer that decides whether anything above it works. Oxbury Earth Read the case study → Founded → acquired Built a hybrid integration platform onto the Gartner Magic Quadrant for iPaaS, scaled it to ~£20m turnover at 73% recurring, and sold to RELX on a full earn-out. Adaptris Read the case study →

Writing

Notes from the work

Written for people who have to make these decisions, not for a search engine.

Aug 2026 Everyone has an API now. That's the accelerant, not the cure EDI solved semantics, delivery guarantees and non-repudiation by about 1995. We threw all of it away for REST, and spent twenty years rebuilding it badly. Aug 2026 Offshore was never about the hourly rate Labour arbitrage worked because coordination cost less than the wage gap. AI has collapsed one side of that equation. Jun 2026 Your data cleaner will rot, and it won't tell you Models degrade quietly. A practical guide to the training and retraining lifecycle, using agricultural product matching as the worked example.

All writing →

Working together

Three ways I take on work

Alongside a full-time role, so engagements are deliberately small and sharply scoped.

Non-executive & board

I chair Fotenix, an AI agri-tech company. Open to one or two further board seats where the technology is the hard part and the commercial model still has to work.

Technical due diligence

I've been on both sides of an acquisition. Give me two days with an engineering team and I'll tell you whether their problems are normal or fatal.

AI-accelerated delivery

Small, tightly scoped builds. One senior engineer with current tooling delivers in weeks what a team quotes in months — prototypes, data tooling, proofs of concept.

Engagements are contracted through Integration Guys Ltd, registered in England & Wales.

Currently

Building AI systems at Oxbury. Chairing Fotenix.

Open to board conversations and short consulting engagements. If you've got a problem where the data is messier than anyone will admit, that's usually the interesting one.