Quoting at a manufacturer: the numbers on 380 quote requests a month
Take a component manufacturer with a four-person technical team and around 380 quote requests a month. Each one means reading the request — often with a drawing or a specification attached — picking items from the catalogue, checking current material costs, running the calculation and putting the quote together. Below we break that down into numbers and show where the value in this process sits that the hours saved do not capture. As always, the caveat: this is a worked example on typical values, not a report from a specific implementation.
- A worked example on typical values, not a write-up of one particular manufacturer.
- At 380 quote requests at 42 minutes each, manual quoting costs about PLN 14,630 a month.
- An agent preparing 71% of quotes for approval cuts the monthly effort from 266 h to 95 h.
Starting point: 380 quote requests a month
Three numbers: 380 requests, 42 minutes each, PLN 55/h for the technical team. That is 266 hours and PLN 14,630 a month. Significant, but the real cost sits elsewhere: at that workload a quote goes out after two days on average, and some requests — usually the smallest — never get an answer at all, because the team prioritises the bigger ones.
How the agent prepares a quote
The agent does not send quotes. It prepares them for approval — and that distinction is the key one in this process, because price is a commercial decision, not the output of a calculation.
Reading the request. Pulling parameters out of the email body and attachments: dimensions, material, quantities, deadline, standard. Missing parameters are listed explicitly rather than assumed.
Matching to the catalogue. Finding the items that match the parameters, plus the nearest variants. Unusual requests outside the catalogue go straight to a design engineer.
Calculation on current data. Pulling current material costs, availability and the customer's commercial terms from the ERP and CRM, then running the calculation against the margin rules in force.
Draft quote for approval. A finished document with the calculation broken out and the assumptions stated goes to the salesperson. They approve it, adjust the margin or reject it — the pricing decision stays with a human.
The technical team's monthly workload before and after implementation, at 71% of requests with a quote prepared automatically and approval taking about 4 minutes. The remaining 29% is handled as it is today.
What changes in the numbers
With 71% of requests prepared automatically, the team approves 270 quotes at about 4 minutes each, which is 18 hours, and handles 110 unusual requests manually, which is 77 hours. That is 95 hours in total instead of 266. The cost per request falls from PLN 38.50 to about PLN 16, and the monthly cost of the process from PLN 14,630 to about PLN 6,100 including the cost of the system — roughly PLN 102,000 over a year.
The second change often matters more. A quote on a catalogue request goes out the same day instead of two days later, and the smallest requests stop dropping out of the queue, because preparing them now costs nothing. In buying processes where the order quotes arrive in matters, that translates into revenue — except you cannot honestly write that value into a savings calculation, because it depends on the market rather than on the implementation. So we keep it separate.
Where this calculation does not hold
The first place: the share of unusual requests. At a manufacturer working mainly to individual orders, where every quote needs a design engineer, 71% is out of reach and 25–35% is the realistic figure. The calculation still comes out positive, but the payback period stretches noticeably. The second: how current the material cost data is. If prices in the ERP are sometimes out of date, the agent will quickly produce a quote on bad data — and that is worse than producing it slowly.
That is why the first step in an implementation like this is not the agent, it is checking how often and from where material costs get updated. It is usually the shortest part of the audit and the one that most often changes the scope of the pilot.
Frequently asked questions
Does the agent send quotes on its own?
Not in this process. It prepares a complete draft quote with the calculation and the assumptions, and a human sends it after approval. Price is a commercial decision and we treat it as one — we automate the work leading up to the decision, not the decision itself.
What about requests with technical drawings?
The agent reads parameters from attachments where they are stated, and lists explicitly the ones it could not establish. Requests that need the drawing interpreted go to a design engineer — with part of the data already filled in, so their work starts later than it does today.
Can we start with part of the product range?
Yes, and that is usually how we start. The pilot covers one catalogue group with the highest request volume, which lets you measure the real share of automated cases before the scope extends to the rest of the range.