All work

Quote Agent

A guided bathroom estimate, with a traceable costing engine behind every line item.

AI tool Trades 2026 prototype
Bathroom estimating workspace — interface redesign concept with sample data

The problem

A bathroom fitter takes a call, and the caller wants a number. The honest answer is "I need to come and look at it," which is true and also the moment a good proportion of those callers ring someone else.

The alternative most trades fall back on is a number off the top of their head. That is worse: it is either too high and loses the job, or too low and gets held to it eight weeks later when the materials came in dearer than remembered.

What I built

A guided questionnaire that asks the fifteen questions a fitter would ask anyway — room size, strip-out or makeover, spec level, tiling height, floor type — and prices the answers against a real materials list and Northwest England day rates.

The architectural decision that matters most: the language model only asks the questions. It never sees a price and never does the arithmetic. Costing runs in a deterministic engine against a priced catalogue, so the same answers always produce the same number and every figure can be traced back to a named supplier listing.

The estimate builds live as the call goes on, broken into labour, materials, plant and contingency, with a "show how this was built" trail behind it. The fitter can read the range out loud with a straight face because they can see exactly where it came from.

FastAPI Claude API htmx Deterministic costing engine WeasyPrint

Project access

A private prototype with no public demo. The image is an interface redesign concept using sample data; the original prototype workflow is described below.

What changed

The caller gets a provisional range before they hang up, and the fitter gets a job sheet already filled in for the site survey. The number is explicitly provisional and subject to survey, which is the honest framing — this replaces the guess, not the visit.

Materials priced
50
LLM does the maths
£0

Next project

Property Maintenance Triage

Tenant reports a leak, the agent triages it, prices it, picks a contractor and puts one decision in front of the landlord.

What are we building?

An app, a better way of working, or an AI integration. Tell me where you’d like to start.