This is a shortened report from a real project. The company's name and its systems are replaced with generic descriptions; the numbers and conclusions are original.
A B2B distribution company, around 30 people, over 25 years on the market. Quotes keep the company alive - and every quote starts as an inquiry that passes through four pairs of hands and four programs:
The customer sends an email, often with an attachment. The inquiry is registered in an internal quoting application: deadline, quantities, contacts.
The inquiry reaches the person who calculates the price. They email the suppliers.
Each supplier replies differently: email, PDF, sometimes even SMS. Someone retypes it all into a costing spreadsheet in Excel, where purchase prices become the selling price.
The price is transferred from Excel back into the internal application and the quote goes out to the customer.
The owners suspected it was slow, error-prone and eating people's time. They wanted to know where to start.
The workshop revealed three places where the process hurts most:
Supplier replies arrive as emails, PDFs and SMS, and someone retypes them into the costing spreadsheet by hand. The retyping and filling-in takes more time than the price calculation itself. People jump between four programs: email, Excel, the internal quoting application and an older company system.
No single template; the structure differs file by file. Nothing can be automated on top of them in their current form; the spreadsheets would have to be unified first. The count is a workshop estimate; I did not count the files myself.
Nowhere is it written down who can manufacture what or how fast each supplier responds. A few people carry it all in their heads - and when one of them leaves, the company loses it.
The client had no savings figure of their own - and I will not pretend a measurement that never happened. For the biggest loss, manual price retyping, the estimate comes out like this:
| ~10 supplier replies a day, ~5 minutes of retyping each | ~50 min |
| Transferring each price into the costing spreadsheet, 2-3 minutes | ~25 min |
| Manual retyping in total | 70-80 min a day |
|---|---|
| Automation reliably reads half to most of the replies | saves 30-60 min a day |
An estimate based on what they told me at the workshop, not a measurement. Your number will be different and we will calculate it together.
The time saved is extra capacity: the same people can handle more inquiries. The left column shows model assumptions - for you, we plug in your numbers.
| Saving 60 minutes a day, 2 people, 21 working days | ~42 extra hours a month |
| A quote takes about 2 hours of work | +20 quotes a month |
| About 30 % of quotes succeed | +6 orders a month |
| Average order 1,700 EUR | turnover +10,200 EUR a month |
| 25 % margin | benefit +2,550 EUR a month |
| Assistant: setup ~4,100 EUR, running costs ~120 EUR a month | cost |
| Assistant payback | under 2 months |
|---|
And even if the extra capacity brought no additional quote at all: 42 hours a month is work worth well over a thousand euros.
The ~4,100 EUR setup estimate is the original number from the report. Today I build and deliver similar assistants myself, for approx. 2,100 EUR - payback then comes out at roughly a month.
More opportunities came up than it makes sense to do at once. The report ranks them by two questions: how much it brings the company and how much work it costs. Here are the five most important:
Nothing. Whatever comes out here does not belong in the plan.
I said no in two places. Precisely these two "no"s protected the client from the most expensive mistakes:
The price here is not a mechanical calculation. Every inquiry is different, and during costing a business decision - and often a new business opportunity - arises. The people who calculate prices protected that space themselves: if a machine built the price, it would disappear. AI may at most suggest from history ("you made a similar thing last year for this much").
Every customer inquires differently: free text, their own spreadsheet, a PDF with pictures. And an email does not reliably reveal what form of quote the customer expects: one large customer wants prices filled into their own spreadsheet, and another with a different one can appear at any time. A mistake here means a lost order. I therefore ruled out two automation ideas entirely: having AI build the quote straight from the inquiry, and having it fill in customer spreadsheets on its own. AI may read and pre-fill - but it must not decide for a person what the quote should look like, and it must not send anything without review.
I would start by extracting data from emails straight into the costing - manual retyping currently runs at both ends of the process.
An incoming customer inquiry as well as a supplier reply.
Who is inquiring, which items, which prices.
A person reviews and approves. Nothing goes out on its own.
You can start right away and it does not touch the existing systems. If it does not prove itself, you switch it off and the process runs as before.
The first step saves time immediately, but the data still lives in four places: email, Excel, the quoting application and the older company system. The long-term direction is one place - a single database of inquiries, quotes and price history.
There is more than one way to get there, and money and pace decide. The first step makes sense precisely because it works even if the big rebuild never happens.
The client chose that first step and left the rollout to me. The assistant is now being built under my direction - it reads incoming inquiries and prepares an input that a person only reviews and approves. The report did not end up in a drawer.
The pilot is already running on real data: from an actual inquiry the assistant extracted all 47 items including specifications and prepared the costing spreadsheet. One pilot case proves nothing yet, but this is exactly what the first step is supposed to do.
In 30 minutes I will tell you whether it makes sense for your company. And if you decide to go ahead based on the plan, I will see the rollout through with you - just like with this company.