Construction · May 28, 2026
Quoting assistant: from 9 days to 2 hours
Czytaj po polsku: wersja polska
This is a demonstration build that shows the pattern we deploy — illustrative numbers, not a specific client’s audited figures. We’ll swap in a named, anonymised case study once the first construction engagements close.
The problem
In most construction and field-service firms, quoting isn’t slow because the maths is hard — it’s slow because of everything around it. A typical path:
- Enquiry sits in an inbox or a notepad in the van.
- Site visit gets scheduled, then re-scheduled.
- Numbers are ready Tuesday; the actual document gets assembled at the weekend.
End to end, a quote that could go out same-day takes around nine days — and a chunk never goes out at all. That’s not a pricing problem. It’s revenue leaking through process gaps.
What we built
A quoting assistant that keeps the human judgement and automates the wrapper:
- One intake. Every enquiry — web form, phone note, referral — lands in one place with the same fields captured.
- Templated estimate. Common job types (bathroom refit, roof repair, X m² of screed) are reusable templates; the owner adjusts quantities, the figures assemble themselves.
- AI-drafted cover. Claude turns the bullet points into a clean scope description, assumptions and terms — in the firm’s voice — for review, never auto-sent.
- Instant, branded send. On approval, the client gets a professional PDF and a short human message in minutes, not days.
The result (modeled)
| Before | After | |
|---|---|---|
| Quote turnaround | ~9 days | ~2 hours |
| Quotes followed up | rarely | every time (automatic) |
| Owner time per quote | scattered across the week | one short review |
The compounding win isn’t just speed — it’s that every quote now goes out and gets a follow-up, instead of only the ones someone remembered.
How it stays trustworthy
- The owner still sets every number and approves every send.
- Templates encode the firm’s real pricing logic, not generic guesses.
- The AI only drafts language; it never invents figures.
This is the same shape as the KSeF invoice auto-assign guide: rules + one AI step for the messy part + a human on approval. Quoting is just the highest-visibility place to start.
Built with
MakeClaudeGoogle Workspace