Interactive demo
Speak a quick job summary in the van; get back a structured, catalogue-matched report. The AI drafts language and flags what’s uncertain — the technician approves the facts.
Read the full idea page →A field technician for a commercial-kitchen service company, sitting in the van after a job. The same flow fits HVAC, lifts, machinery and medical-device service.
The technician taps “record” and speaks a 30-second summary of the repair they just finished at a bakery in Linz.
Instead of typing a report on a phone with cold hands, the technician just talks for half a minute.
The recording is transcribed and the key facts are pulled into structured fields the office actually needs.
The parts count is genuinely unclear in the audio. The model marks it for confirmation instead of inventing a number that would later be billed.
UncertainHeard “two or three seals” — quantity ambiguous. Parts field flagged for the technician to confirm.
Parts are matched to the van-stock catalogue, and a hard rule stops the report from being exported while a required field is still unconfirmed.
Deterministic checkDeterministic: required-field gate (parts) + each part matched to a real catalogue SKU before it can appear on an invoice.
The AI drafted the wording; the technician owns the facts. They fix the one ambiguous field in a couple of taps.
Confirm parts used: drive belt ×1, bearing grease, and the seal quantity.
One confirmed report fans out into every place the business needs it — no re-typing.