An AI agent that reads clients' requests for quotation and finds each line in the company's price list, leaving the decision to the operator.
The challenge
To ask for a quote, a building-services company's clients send a bill of quantities: the list of works, written with their own abbreviations and descriptions. And a line in the bill almost never matches a single item in the price list. 'Lighting point with switch', for example, becomes four items in the price list: the conduit, the cable, the switch with its faceplate, and the installation.
So every request has to be interpreted. You need to know the price list, which runs to thousands of items, and the trade, to know what goes into each job. Doing it line by line is slow and tedious. And because of the time it took, some requests went unanswered: quotes that never went out.
In simple terms
A customer walks into a garage and asks for 'a service'. The mechanic can price it at once, knowing it means oil, filters and labour. In a bill of quantities every line is like that 'service': it has to be broken down into its parts, and each part found in the price list. The Quote Builder does it instead of someone spending whole days on it. So quotes go out sooner and no request gets left behind.
What it does
The Quote Builder is a RAG agent: before suggesting an item, it looks it up in the company's price list, searching in two ways at once. For each line it suggests the matching item and explains why it chose it; the operator confirms or corrects. At the end, the quote goes back into the format the company already uses, its Excel cost file, with the formulas intact.
Three terms, explained simply
- RAG agent
- Before answering, it checks the company's documents instead of going from memory, like a colleague who opens the price list before giving you a price.
- Search by meaning (semantic search)
- It understands that 'S&F mltlyr pipe Ø20' and 'Supply and fix multilayer pipe, 20 mm diameter' ask for the same thing: different abbreviations, same job, same price-list item.
- Exact-match search
- When a line contains a code or a precise wording, it searches for it character by character, like the barcode scanner at the till.
It learns from corrections
When the operator corrects an item and saves, the choice is remembered. The next time the same line comes up, the agent suggests it straight away; it still needs confirming, because the final word always belongs to a person.
Who it's for
For anyone who prepares quotes from clients' requests and their own price list, in any sector. It isn't a one-size-fits-all program: first we understand what your business needs, then we customise it and install it.
How you use it
- You upload the request received from the client, as a PDF or Excel file, and enter the client and the job.
- The agent reads every line and suggests the price-list item, with the reason for its choice. Identical lines appear only once.
- The operator confirms or corrects. Lines with no match in the price list are the operator's call: a new price, several items combined, or leaving it out.
- Save, and the cost file is filled in. Saved corrections stay, ready for the next request.
The resultIn use since June 2026. According to the company, a quote that used to take about 10 working days now takes 2.
What the company keeps. Your most experienced employee no longer spends their time searching line by line: their skills go into more productive work, such as special jobs and the final check of the quote. Every correction they make teaches the agent, which applies it on its own next time. Price list, corrections and documentation stay in the company's records.
Works with: requests for quotation in PDF or Excel · the company price list · the Excel cost file.