Delivered by our team through its European experience (Evolution Agency, France).
Understand the need before talking products, on the website and on the phone.
For Motralec, an industrial equipment distributor, our team built three tools: Dany, the website assistant; a voice agent for after-hours calls; and automatic reading of supplier product sheets.
3
tools delivered: website assistant, voice agent, product sheet reading
7 days
to deliver the after-hours voice agent
0
prices quoted and no invented links during testing: pricing stays with the sales team
Methodology: the results on this page are qualitative. They come from operating the tools and from the testing done before each launch. Every request sent is checked in the tool's log, not taken on the model's word. We publish client-specific percentages only once they are measured and approved.
Context
Motralec is a distributor of industrial equipment (hydraulics, air handling, electrical) based in Boissy-sous-Saint-Yon, France. Its catalog covers thousands of products from dozens of brands: pumps, compressors, valves, fans.
The requests are technical: replace a specific pump, find a part, get a quote. Motralec did not want a bot that pushes quotes. It wanted an advisor that first understands the need. The starting point:
- Very different visitors: a professional replacing equipment, an individual looking without knowing the reference
- Calls that come in outside business hours
- Product sheets to build from supplier PDFs: catalogs of dozens of pages, pump curve tables, specifications in several languages
Challenge
Three different jobs, one shared rule: invent nothing. At a technical distributor, a wrong price, a wrong link or a wrong spec is worse than an unanswered question.
Qualify before suggesting
Understand whether the customer is replacing equipment or looking for a solution before citing a product. And never quote a price: pricing stays with the sales team.
Handle after-hours calls
When the company is closed, the caller still needs to leave a clear request, with contact details that are actually right.
Accurate product sheets
Pull the specs of one specific product out of a supplier PDF, with the right units. On a pump curve, mixing up flow and head makes the sheet wrong.
Solution: three tools, each with its safeguards
Three tools built and run by our team. What they share: they answer from the company's own data, and they flag what they cannot do instead of improvising.
- 1
Dany, the website assistant
Dany answers on motralec.com from four knowledge bases: practical information, technical questions, product sheets and brand pages. It asks questions first to understand the need, then cites catalog sheets with their links. It gives no prices and no installation instructions: it points to the manual and to an installer.
- 2
Complete contact requests
When a visitor wants a callback or a quote, Dany collects name, email and need (phone is optional) and emails the request to Motralec's quote inbox, with a copy to the visitor. It does not book appointments: it passes on a callback request.
- 3
Tested before going live
Before launch, the assistant was tested with personas: the rushed buyer who wants a price, the lost individual, the professional replacing equipment, a hijacking attempt, the hot prospect who goes all the way. Every link it cited was cross-checked against the knowledge base.
- 4
A voice agent for after-hours calls
We also built a phone version of Dany, for calls outside business hours. It takes the caller's request and contact details, then a summary is emailed to the manager. Phone numbers are read back in pairs and email addresses spelled out, so no contact detail goes out wrong. Delivered in seven days.
- 5
Automatic product sheet reading
The supplier catalog is dropped into a form as a PDF. OCR reads the document, a language model writes the sheet for the target product (the same catalog can be renamed after each product wanted), and the result lands in a spreadsheet. A check tests every spec against its unit: a flow and head swap is fixed automatically, any other inconsistency flags the sheet 'to review' instead of letting it through.
- 6
A pipeline that does not fail silently
If the first OCR engine fails, the document is automatically sent to a second one. A failure triggers an email alert, and usage is only counted once a sheet has succeeded.
Results
No percentages on this page: we only publish a client rate once it is measured and approved. The results below are observed in operation and during testing.
No price quoted and no link invented during persona testing
Contact requests arrive complete (name, email, need) in the quote inbox
After-hours voice agent delivered in seven days
Product sheets checked unit by unit, with automatic OCR fallback when the first engine fails
The honest framing: Dany does not replace the sales rep. It qualifies upstream and hands over requests that are already clearly stated.
Stack
Your catalog is technical. Your assistant should be too.
Evaluate your project free with our interactive tools, or book a call: we look at your real requests and tell you what an agent can handle, and what it should leave to your team.