Writing a quote takes five minutes. That is not the problem.
The problem is that it sits there until the boss has time. How to fix that without replacing a single system, what it costs, and how you can tell whether the numbers in the draft are right.
An industrial firm, five employees, a supplier to the manufacturing industry. Enquiries arrive by email, in German, French and English. The owner writes the quotes himself, and he is quick at it: check stock, run the costing in an Excel with a fixed formula, write the text, send. Five minutes.
Quoting is still his biggest bottleneck. Not because of the five minutes. Because those five minutes are only his, and he spends half the week at a customer site, in an audit, or travelling.
An enquiry that arrives at five on a Friday gets answered on Monday. If a competitor answered at six on Friday, that was not a time problem, it was a lost order. And nobody ever records it, because you count the quotes you sent, not the ones you sent too late.
That is the real reason a quote assistant pays for itself, and it has nothing to do with saving time. It is that an enquiry gets an answer within the hour even while you are standing in a workshop. And that the answer comes back in the language it was asked in.
Why no software solves this
Good quoting software exists. For most small firms it fails not on features but on its precondition: it needs clean, structured data. That is exactly what you do not have.
At the firm above it looks like this, and I suspect yours is not far off:
- The inventory system dates from 1996, runs reliably, and is needed for another two years or so. It can do exactly one thing towards the outside world: print a PDF.
- The costing lives in an Excel with one formula that grew over years, and the owner occasionally lowers the price by hand because he knows the customer.
- Enquiries arrive as prose in emails, in three languages, sometimes with a drawing attached, sometimes with a part number in the subject line.
Every product on the market tells you the same thing at this point: migrate first. That is not wrong, it is just a year-long project you will never run alongside daily business, and at the end of it you have software that does not know your margin.
We go the other way round. Nothing is migrated and nothing is replaced. One mailbox is the only interface.
How you can tell whether the numbers are right
This is where it separates: whether somebody built an assistant or a chatbot with a good cover letter. A language model invents numbers if you let it. Not out of malice, but because a quote does contain a number and it wants to complete the task.
We measured this on our own system, not on a client's: the instructions said, verbatim, that it must not invent numbers. In four out of five runs it did exactly that anyway. A ban in the instructions is not a brake, it is a request.
What works instead is four things that are not language models:
Every figure in the draft has to point at a row in your price list or at the PDF export you sent along. If no source is found, the run stops and you get a note instead of a draft. Better no quote than a wrong one.
Always EXW. Discount capped at this much. Valid for thirty days. Those sentences sit in a file you can read and change, not in an instruction to an AI that gets re-rolled with every model change.
Is there a price? A lead time? The right language? Do the totals add up? That is not checked by a model but by arithmetic. Whatever fails is put in front of you instead of being sent.
The draft sits in your mailbox, customer in the To field, text finished. You read it and send it. That costs you twenty seconds instead of five minutes, and it works from anywhere.
The long version, with the incidents we learned this from and an excerpt from a real rules file, is written up here.
What you have to provide
Everybody asks this first and almost nobody writes it down. For the first stage we need from you:
- Ten to twenty past quotes, exactly as they went out. That is how we learn your tone, your structure and your standard phrases.
- Your price list or costing logic, in whatever form it exists. An Excel is fine. It does not have to be pretty.
- One hour of conversation about your rules: delivery terms, discount limits, what you never commit to.
- A mailbox we are allowed to use. Your IT provider sets it up in ten minutes, and we are happy to work with them directly.
What you do not have to provide: access to your ERP, a cloud migration, clean master data, or a project team.
What it costs
The first stage goes live in three to four weeks. For most firms the fixed price lands between €6,000 and €9,000 net, depending on how many special cases your quotes involve. We agree the number before we start, and it does not move afterwards.
For companies based in Luxembourg the arithmetic is different. The SME Packages AI scheme covers 70 percent of eligible costs for an investment between €3,000 and €25,000 net. A project of this size sits in the middle of that window. That leaves roughly €1,800 to €2,700 with you. You file the application yourself; we supply the documents it requires.
Amounts are orders of magnitude from live and costed projects, not commitments. The funding terms are as of 21 August 2026; check them on guichet.lu before applying.
What this is not
Not a CRM, not a new system, not a replacement for anything. The assistant writes drafts and nothing else. It sends nothing, it changes nothing in your systems, and if you switch it off tomorrow you carry on exactly as before, just more slowly again.
And it is no substitute for your judgement. With a customer you have known for fifteen years, you know whether the price should be different this time. No system can do that, and none should. The assistant takes the repetition off you, not the decision.
If there is a quote on your desk right now that has been waiting since Friday
Then a conversation is worth it. We look at how enquiries reach you, what actually takes the time in a quote, and whether a first stage pays off for you. If it does not, we say so.
The firm described at the top is a live project of ours, reproduced without a name and without identifying details. Figures and processes in this article are orders of magnitude from practice, not commitments for your case.