Case study
From customer email to quote, with no data entry
A chain of eight services that takes a request for quotation as it arrives, attachments included, and turns it into a quote ready for approval in the ERP. The video shows it on fictitious data: the customer, items and warehouses are made up.
8
automated steps, from email to quote
The starting problem
A request for quotation rarely arrives in a usable form. An email, a PDF, a workshop scan, an Excel sheet: someone has to read it all, find the customer in the ERP, look up every item, check cuts and lengths, check stock, key in the quote, then reply.
Eight tasks, all manual, repeated for every request. It is the most time-consuming flow in sales, and the time it takes is not spent with the customer: it is spent copying.
What was built
Each task became an automated service. The email arrives and the sender is matched to their account in the ERP: customer, delivery address, contact, sales scope and language. A vision model then reads every attachment, whether PDF, scan, photo, Excel, Word or ZIP archive, and extracts the lines of the request.
For each line, AI interprets the free text and the catalogue picks the exact item: grade, dimensions, unit, quantity. The cut is decided from standard lengths, then the line is checked against real stock in each warehouse and against the delivery date. The quote is then created in the ERP, and the reply goes out to the customer in their language, with their original email attached.
The demonstration
The technical choices, and why
Three decisions that keep an automation chain reliable, and that make sales reps trust it.
One service per step
The eight steps are eight separate services: each one drops off its result, the next one picks it up. A step can be improved, or replaced, without touching the others.
AI suggests, the catalogue decides
The language model is there to understand a request written the way you would write to a colleague, "HEA200 in S355, 12 m long". The chosen item always comes from the catalogue: the chain never makes up a reference, it picks from the ones that exist.
The sales rep stays in control
Every quote line carries a confidence icon: ready, stock to confirm, quantity to check, item to verify. The sales rep does not re-read everything; they go straight to the lines that need it, then approve.
What it changes in practice
The most time-consuming flow in sales runs without data entry. By our estimates, the same team can handle up to three times as many requests, which means up to twice the potential revenue, without hiring. These are potential gains, depending on the volume and nature of the requests.
The time freed up does not vanish: it goes back to technical advice, negotiation and customer follow-up, in other words to what actually sells.
The same mechanics, for your flow
Take a request as it arrives, turn it into structured data, check it against your reference data and produce the final document in your own tool: the mechanics apply to quotes as much as to any flow that currently relies on re-keying.
What I deliver is the chain tailored to your tools and business rules, the hosting, and the source code. If your sales reps spend more time typing than selling, that is probably where to start.
- Requests for quotation and quotes
- Orders received by email or as PDFs
- Supplier invoices and delivery notes
- Complaints and after-sales requests
Frequently asked questions
- No. The chain relies on the ERP you already have: it finds the customer there and creates the quote there. What changes is that nobody has to copy the request anymore.
- The model is there to understand the request, not to invent a reference: the chosen item always comes from the catalogue. And every line carries a confidence icon flagging what still needs confirming, whether stock, quantity or item. The sales rep reviews those lines first, then approves.
- The email body and its attachments: PDFs, scans, photos, Excel or Word files, ZIP archives. A vision model reads them line by line, including when the request is a workshop scan.
- No, they are estimates of potential gains. They depend on the volume of requests, their nature, and how much of your sales reps' work is data entry. Scoping is precisely where we estimate what automation changes for your flow.
- The demonstration uses fictitious data: the customer, items and warehouses shown are made up. No production data appears in the video.
- Yes. The terms covering code, documentation and hosting handover are spelled out in the proposal, before any commitment.