From “we need AI” to same-day quotes
A manufacturer of laser-cut, bent and welded metal parts knew AI mattered but had no idea where it would pay off first. We started with a three-week opportunity map — and quoting came out on top.
- Client
- Metal parts manufacturer, Germany & Poland
- Their team
- CEO and sales director (owners), 6 sales engineers, ERP admin
- Our team
- Engagement lead, 2 engineers, data scientist
- Timeline
- 3-week map · 9-week pilot · production in 2 months
The business goal
Answer more requests for quotation, faster — customers often order from whoever replies first — without hiring more sales engineers, who are very hard to find.
Where it was
RFQs arrived by email with PDF and CAD drawings. A sales engineer read each drawing, searched the ERP for similar past jobs and calculated material, machine time and margin by hand. Busy weeks meant many RFQs simply went unanswered.
What we did
- 1Opportunity map · 3 weeksInterviews with 14 people, a look at the data, nine ideas ranked by payback and effort — quoting, production planning, supplier invoices and more.
- 2Outcome brief · 1 weekQuoting chosen. Baseline measured, target and exit rule agreed with the sales director.
- 3Pilot · 9 weeksDrawing and RFQ parsing, search over 18,000 past jobs, a suggested price from ERP history; the engineer adjusts and approves. 83% within ±10% — continue.
- 4Production & runERP integration, quote PDF generation, a monthly review of accuracy and win rate.
What did not work
Complex welded assemblies: estimates were off by up to 40%. They stay fully manual — but the tool still finds similar past jobs, which saves an engineer about an hour each.
Who did what
Before and after
Under the hood
The map changed the conversation. Instead of “we need AI”, we argued about which of nine numbers to move first.