Inside sales starts the day with quotes that are ready and waiting
A customer sends a message: do you have 250 bags with our logo by the end of the month? The agent reads along, finds the product, calculates the decoration and prepares the quote. A human looks at it and sends it.
The situation
Requests arrive by email, WhatsApp, the form on the site and the phone. They are rarely complete: no quantity, no technique, no date.
So inside sales starts every request by asking back, and then by looking things up. That is the work nobody enjoys and that still determines the lead time: the customer is not waiting for a price break to be calculated, they are waiting for someone to have time to look. And whoever comes back with a price first is in front.
The challenge
A chatbot that only answers moves the work around. We wanted an agent that does the work: search the catalogue, pick the right variant, apply the price breaks, make the proof and put the quote into the system. That only works if the agent can genuinely reach the systems. So it hangs via MCP on the PIM, the quoting module and the mailbox, with what may be read and what may be written fixed per connection.
The second challenge was trust. An agent that sends quotes on its own is one wrong price away from a problem with a customer. That is why the approval step is not a setting but a design decision: it is not on because someone switched it on, and it cannot be switched off.
The line we hold
The agent prepares, the human decides. A quote never goes out automatically. There is always an approval step in between, and it is always possible to see why the agent did something: which product it picked, which technique, which price break.
We also do not use your business data to train third-party models. That is not a promise about good intentions but a property of the setup: what may be read and written is fixed per connection, and that can be checked rather than believed.
How it runs in practice
The agent takes the steps an employee would take, in the same order, and stops where a human has to form a view of the result. What it cannot find, it leaves open instead of inventing it.
So the outcome of a request is not an answer but a draft quote in the system: with product, variant, quantity, technique, price breaks and setup costs, ready to be reviewed.
Read the request
From e-mail or the form on the site. What is not in it — quantity, technique, date — stays an empty field and is not filled with an assumption.
Find the product
In the catalogue, through the PIM, including the right variant. The same source the sales environment uses, so there is no second list that can differ.
Calculate the decoration
The technique, the price breaks, the number of colours and the setup costs, from the technique rules in the system instead of from a spreadsheet.
Prepare the quote
As a draft in the quoting module, with the proof alongside. Then a human looks at it and sends it. The agent sends nothing itself.
The result
At the Brinxx group the agent has been preparing about 380 quotes a month since 2025. Inside sales starts the day with quotes that are ready instead of a full inbox. The work has not disappeared, it has moved: from looking up and retyping to reviewing and deciding.
What is not here is a lead time before and after. Nobody clocked both, so we put down no minutes — not even when an estimate sounds plausible. The figure of about 380 a month also still has to be confirmed, together with the period measured.
What the customer says about it
This quotation is a draft. We wrote it down the way we understood it from the conversations, so that it can be put to the customer. It only goes live once the customer has approved it word for word and the name and role are in place.
“It makes the quote. We look at it and send it. That is exactly the division we wanted.”
Three stories that belong with this one
The cases are connected: the same catalogue, the same technique rules and the same agent come back in several stories. These three sit closest to what you have just read.
Let us run last week's requests
Send us the requests that came in last week. We put them through the agent and show what comes out as draft quotes, and where it leaves a field open because the request was not complete.