A quote from a voice note

The problem

Trade businesses lose jobs because quotes sit for days: after a long day on site, there is no energy left in the evening to type every line item neatly into the software.

The solution

Right after the visit the tradesperson dictates a short note (“full bathroom, 6 m² of tiles, rip out the old ones, new fittings, ~3 days”). A locally running AI structures this into line items, adds standard texts and guide prices from the company’s own price list, and creates a draft quote — which the tradesperson only needs to review and release.

How it works

1. The voice note is transcribed locally (nothing sent to the cloud).
2. A small language model extracts trades, quantities and effort.
3. Matched against the stored price list → line items with guide prices.
4. The draft quote lands in the existing system for release.

What changes

What used to be “an hour of typing in the evening” becomes two minutes of talking plus five minutes of checking. Quotes go out the same day — and faster quotes measurably win more jobs.

Honest take

No frontier model is needed here. A small, locally running language model is entirely sufficient — cheaper, faster and data-frugal. What matters is not “the biggest AI” but the clean link to the company’s own price list.