What field service taught us about logistics AI

When we tell logistics people that Vera grew up in field service — technicians, work orders, service vans — we sometimes get a polite frown. Different industry, surely?

Structurally, it's the same problem. A field-service operation is a routing-and-assignment problem under hard constraints: certifications instead of ADR classes, SLA response windows instead of delivery windows, working-time limits identical in both. Disruptions arrive the same way too — a person calls in sick, a vehicle dies, a customer escalates — and the resolution requires applying a dense rulebook in minutes.

We started there deliberately, for three reasons.

The feedback loop is faster. A field-service operation makes dozens of consequential dispatch decisions a day with a tight loop between decision and outcome. For a young product, that's a brutal and wonderful teacher. Vera's reasoning pipeline has been corrected by reality at a pace a slower-moving domain would never provide.

The rules are richer than they look. Our pilot customer's rulebook — captured in workshops from people who'd never written it down — turned out to have the same texture we see in logistics: regulatory constraints (gas-permit requirements under Tukes rules), commercial ones (Gold-tier response times), and human ones (maximum working days). If the architecture handles that mix, the constraint content is just data.

Mid-market field service was underserved. The big AI money chases the big logistics brands. Meanwhile, a 50-technician operation has the same disruption chaos and none of the tooling. They became paying customers because the product solved a Tuesday-morning problem, not because AI was on a strategy slide.

Six months into supervised production, the transfer thesis looks right: nothing in the engine knows it's doing field service. The rules are the customer's; the solver doesn't care whether the entity is a service van or a 40-tonne truck. When we take Vera into third-party logistics, we change the rulebook, not the architecture.

That's the plan for the year ahead. Same engine, heavier vehicles.