From voice memo to work order — field engineers documenting interventions hands-free

Field engineers spend their day on-site — diagnosing problems, fixing equipment, talking to customers. The work is hands-on; the documentation is not. Every job ended with an hour at a laptop typing up what happened, what was replaced, what needs follow-up, what to bill. Half the time the notes were thin because the engineer was tired and the day was over. The follow-up tasks fell through.

What we built

A documentation system the engineers never have to type into. They speak a voice memo at the end of each visit — phone in hand, often still on-site. AI transcribes the audio, extracts the structured fields the business needs (customer, equipment, fault, parts used, time spent, follow-up actions), generates the right tasks for whoever owns them, and pushes the lot into the existing business systems. The engineer’s laptop stays closed; the next job is already loaded.

How we built it

A Next.js mobile-first app on Vercel for the engineers, Supabase as the source of truth for visits, customers, equipment, and follow-ups. OpenAI Whisper transcribes the voice memos; OpenAI parses each transcript into structured fields and writes them back. Trigger.dev runs the durable jobs — task generation, business-system handoff, retry on failure. The integrations push into the customer’s existing CRM, dispatch system, and billing — engineers don’t learn a new tool.

The outcome

An hour back per engineer per day. Documentation that actually reflects what happened on-site, because it was captured while it was fresh. Follow-ups that don’t fall through, because the system makes the tasks before the engineer leaves the parking lot.