Case study: AI quotation automation for a premium sanitaryware supplier
Managers at a premium sanitaryware supplier spent a lot of time building quotations by hand. We built an AI module that collects requests from every source in one system, picks the products, calculates the order and drafts the quotation. A manager checks every quotation before it goes to the customer. Bills and invoices stay with people: the system does not issue them.
The client and the starting point
We don't name the client. The business, the work and the results are real.
The company supplies premium sanitaryware. Orders come from designers and from customers directly, together with documents and requirements, and every one of them turned into a long stretch of manual work for a manager.
For each request, a manager read the order and the documents and went through the catalogs by hand to find items that met the requirements. Then came copying product photos from websites, assembling schemes into a PDF and putting the commercial proposal together. All of it manually.
Much of that was repetitive copying and formatting, done by hand for every request.
What we built
One intake for all requests
Requests from every source now arrive in one system, wherever the designer or customer sent them.
Product selection
The system reads the requirements and picks matching items from the catalogs.
Calculation and a draft quotation
It calculates the order and drafts the commercial proposal, with product photos and schemes already in place.
Manager approval before sending
Every quotation waits for a manager. The manager checks it, corrects it if needed and approves it. Only then does the system send it to the customer.
Clear limits
The system does not issue bills or invoices. People handle them, along with anything else outside the system's authority.
How it works now
Step by step
- CustomerA designer or customer sends an order with documents, through any channel.
- SystemThe request lands in one system together with its documents.
- AIAI reads the requirements and picks matching items from the catalogs.
- AIIt calculates the order and drafts the commercial proposal.
- TeamA manager checks the draft, corrects it if needed and approves it.
- SystemThe approved quotation goes to the customer.
- TeamWhen the customer confirms, people issue the bill and the invoice.
The system does
- Collects requests from all sources in one place
- Picks catalog items to match the requirements
- Calculates the order
- Drafts the commercial proposal
- Sends it to the customer once a manager approves it
People do
- Check and approve every quotation before it goes out
- Issue bills and invoices: the system does not
- Take over wherever the system is not authorized to decide
Before and after
| Area | Before | After |
|---|---|---|
| Incoming requests | Arrived from designers and customers and were handled by managers | Requests from all sources flow into one system |
| Picking products | Manual search through the catalogs | The system picks items to the requirements |
| Commercial proposal | Photos copied from websites, schemes put into a PDF, proposal written by hand | Drafted by the system, approved by a manager |
| Sending to the customer | Done by the manager | Done by the system after the manager's approval |
| Bills and invoices | Issued by people | Still issued by people |
| Manager's role | Putting together every quotation | Checking and approving a ready draft |
Result
Managers no longer build quotations from scratch. They check and approve a ready draft instead of putting it together, and nothing goes to a customer without that approval. Bills and invoices stay with people. Every request, whatever channel it came from, is in one place.
Results in numbers
- About 2 hours → 15–20 minutes of checking
- Around 150
- Roughly 7 in 10 approved without changes
- Roughly 250 hours a month
Project snapshot
- Scope
- AI module, process automation
- Client size
- About 6 sales managers; around 20 brands and several thousand catalog items
- Timeline
- About 3 months from start to launch
- Period
- Launched in 2025; we still support it
- Team
- Two senior developers: architecture and AI, integrations and interface
- Stack
- Next.js, Node.js, PostgreSQL, n8n, LLM API
- Request channels connected
- 4: email, website form and two messengers
Next step
Describe the situation in a few lines. One of the founders replies and says whether we can help and what the first step would be.
Discuss your project