Your quotes are slow because the request arrives as a PDF.
We build workflow-specific RFQ and quotation systems around the tools your team already uses — your catalogue, your pricing rules, your ERP.
Built per business · Human approval on every quotation
A quotation is not a hard document to produce.
It is hard to produce quickly and consistently, because the information arrives in whatever shape the customer felt like sending it, and the answer depends on things that live in four different places: the catalogue, the price list, this customer’s agreed terms, and what your most experienced person knows about them.
So the work goes to whoever holds all four in their head, and that person becomes the bottleneck. Enquiries wait a day. Some go cold. A few get quoted at the wrong price, and nobody notices until the margin report.
Who this is for.
Likely a fit
- You quote from a catalogue of thousands of SKUs.
- Enquiries arrive by email as PDFs, spreadsheets or free text.
- Price varies by customer — agreed terms, price files or negotiated discounts.
- Quoting depends on two or three people who are hard to replace.
- You already have an ERP or CRM where quotes are recorded.
Probably not yet
- You sell a small number of configurable products where every quote is bespoke engineering.
- Enquiries already arrive structured, through a portal or EDI.
- You quote fewer than a handful of enquiries a week.
- Your catalogue and pricing are not recorded anywhere a system could read.
The work is all upstream of the quotation.
Your ERP produces a quotation perfectly well once somebody has told it what to quote. The effort is entirely before that point: reading an unstructured request, working out which catalogue item the customer’s description refers to, and determining the right price for this particular customer.
That is not a gap in your ERP. It is outside its job. Which is why adding modules to it does not help, and why the work has stayed manual.
What the system actually does.
This is the shape of a typical implementation. The specifics — which formats, which rules, which system to write back to — are set by the business it is built for.
- 01Email / PDF RFQ arrivesIngested in whatever format the customer sent — PDF including scans, spreadsheet, email body text, or a photograph of a printed enquiry.
- 02Extract line itemsDescriptions, quantities, units and any reference numbers are pulled out, each one keeping the source text it came from.
- 03Identify productsExtracted descriptions are resolved to candidate products, with a score rather than a single forced answer.
- 04Match catalogue / SKUCandidates are matched against your catalogue. Lines the system is not confident about are marked rather than guessed.
- 05Retrieve pricingApplicable pricing is looked up for the matched item and this specific customer — list, agreed terms, or quantity-break tier.
- 06Apply business rulesYour rules are applied and anything outside them is flagged: margin below an agreed floor, restricted items, minimum order quantities.
- 07Human reviewA person sees the draft with every uncertain line marked, the source text beside it, and the rule that produced each price. They confirm or correct.
- 08Generate quotationThe quotation is produced in the format your team already uses, after a person has approved it.
- 09Push / update relevant systemThe approved quote is written back to your ERP, CRM or wherever quotes are recorded today.
Where the human stays.
Nothing is sent to a customer automatically. No price leaves the system without a person approving it. Every line shows the source text it was read from and the rule that priced it, so a reviewer can check a number without reopening the original document.
The thresholds are yours: which categories always require review, which customers are never auto-priced, and what confidence level is acceptable before a line can pass without being looked at.
What we don't claim
We do not claim autonomous quoting. Matching a customer’s description to the right item in a catalogue of thousands is genuinely hard, and on messy real-world documents no system gets it right every time — including ours.
What we build is right on the routine majority, honest about the rest, and faster to correct than to do from scratch. Any vendor claiming full automation on unstructured RFQs is either testing on clean documents or not testing.
Our extraction accuracy varies substantially by document type. We publish that work, including the failure cases, in Labs.
Send us thirty real enquiries.
Before anyone commits to building, we measure. Send thirty to fifty genuine enquiries from the last quarter, exactly as they arrived — including the bad scans, because those are the ones that decide whether this works.
We run them through extraction and matching and report back: how much of your quoting volume is mechanical, how much genuinely needs judgement, and what a system would realistically save. If too much of the work needs judgement, we tell you, and you have not committed to a build.
Common questions.
Is this an existing product I can buy?
No. This is a description of what we build for a specific business. There is no Probatus RFQ product to log into. Each implementation is built around one company's catalogue, pricing rules and systems, because those are the parts that determine whether it works.
Does the system send quotes to customers automatically?
No. Nothing is priced or sent to a customer without a person approving it. The system prepares a draft and marks what it is unsure about; the decision to send stays with your team.
What accuracy can you promise on extraction?
None, before we have seen your documents. Extraction quality varies substantially by document type — a native digital PDF and a photograph of a stamped printout are very different problems. We publish our own benchmark work in Labs, including the failure cases, and we measure against a sample of your real documents before quoting.
What do you need to start?
Thirty to fifty real enquiries from the last quarter, exactly as they arrived. Plus a catalogue export, price lists and customer-specific terms. Twelve months of historical quotes helps considerably, because it encodes pricing decisions nobody wrote down.
Does this replace our ERP or CRM?
No. It sits between the inbox and the systems you already run, and writes into them. Nothing is migrated or replaced.
Send us thirty real enquiries.
We will tell you what proportion resolve cleanly, where they break, and whether a system is worth building for your documents.
20-minute conversation · Human approval on every quotation