title: "Site-notes → priced quote for a commercial-trades contractor" dek: "A 2024 implementation that took builder-on-site notes (voice + photos) and produced priced quotes in 22 minutes. Conversion-rate up 24%, quote backlog gone." sector: "Trades & Construction" client: "Commercial-trades contractor · ~80 staff · electrical + HVAC" engagement: "Pilot" duration: "13 weeks" year: "2024" outcome: "Quote turnaround: 4–7 days → 22 min · win-rate +24% · quote backlog 70 → 0" solution: "Site-visit voice + photo capture → AI scope extraction → priced quote against the firm's rate-card with estimator review." timeSaved: "~3.5 hours per quote · A$0.60 per quote drafted" visual: "none" cardFigure: "workflow" timeMetric: "3.5 hrs" timeMetricLabel: "saved / quote" costMetric: "A$0.60" costMetricLabel: "cost per quote" speedMetric: "270×" speedMetricLabel: "faster turnaround" publishedAt: "2024-11-04" keywords:
- construction AI Australia
- trades quote automation
- estimator AI
- HVAC electrical quoting
The problem
A commercial-trades contractor — eighty staff, electrical and HVAC across Melbourne and Geelong, A$26M annual revenue — was losing work to faster competitors. The builder on site would estimate the job in their head, drive back to the office, brief the estimator, the estimator would draft the quote in 4–7 days, and the prospect had already signed with someone else.
The owner had three estimators and a 70-quote backlog. Hiring a fourth estimator wouldn't fix the speed problem — it would just clear the backlog and re-establish the same 4–7 day cycle.
What he wanted was a way to get a defensible priced quote into the prospect's hands the same day. Not an automated quote — the firm's rate-card had subtleties no automated system was going to handle right. But a faster path through the same estimator.
What we did
Three weeks of scoping (we rode along on twelve site visits across both trades), eight weeks of build, two weeks of pilot with three of the senior site supervisors. The deployed system:
- Accepted voice notes and site photos through a phone app the builder used during the walk-through
- Transcribed and extracted scope items — fixtures, fittings, runs, switchboards, ducting, access constraints, programme dependencies
- Cross-referenced extracted scope against the firm's published rate-card and labour-allocation rules
- Drafted a priced quote in the firm's standard format
- Routed the draft into the estimator's queue with confidence scores per line item and photos linked to each scope item the estimator could verify
The estimator reviewed every quote. They corrected anything the system had under- or over-priced. They added the soft cost the system didn't know about (access difficulty, client temperament, urgency premium). They sent.
The outcome — at 6 months in production
| Before (FY24 H1) | After (FY24 H2) | |
|---|---|---|
| Quote turnaround | 4–7 days | 22 minutes (median, including estimator review) |
| Quote backlog | 70 | 0 |
| Win-rate on quoted work | 38% | 47% (+24% relative) |
| Quotes issued per week | 24 | 41 (same headcount) |
| Site-supervisor time on quote prep (per job) | ~45 min after-hours | 0 (capture is done during the walk-through) |
| Cost per quote drafted (model + infra) | n/a | A$0.60 |
The 24% win-rate lift mattered more than the throughput. The firm wasn't winning more bids because the quotes were better — they were winning because the prospect got the quote before the competitor's estimator had picked up the phone.
The biggest change wasn't the AI. It was that my best site guys stopped doing two hours of paperwork after every site visit. They go home to their kids now.
— Owner, commercial-trades contractor
What we'd do differently
Pilot with a single trade first. We piloted across electrical and HVAC concurrently. HVAC's scope-extraction was substantially harder (more variables, more configurations) and slowed the electrical rollout by three weeks. Should have shipped electrical to production first, HVAC second.
Capture programme dependencies more carefully. Trades-sequencing dependencies (must follow plumber, must precede painter) were initially captured as free-text notes. We moved them to structured fields in month three. Should have been week one — they drive material delivery scheduling more than they drive pricing.
What we didn't do
We didn't send any quote without estimator review. We didn't replace the firm's job-management system. We didn't auto-calculate margin — every line stayed at rate-card price, and the estimator continued to apply the margin overlay.
The most important architectural decision was the boring one: capture happens on site, draft happens during the drive back, review happens that afternoon, the prospect has the quote that day. Compress the workflow without removing the estimator.