Automating quotes and orders with AI in a small business.
What an AI agent does with the quotes of an online print shop, what stays with the team, and why.
Automating quotes with artificial intelligence means an AI agent reads every request that comes in, works out what the customer wants, proposes a price and leaves the reply ready. A person reviews it and sends it. We are rolling this out now at an online print shop owned by about twenty printing companies: the website takes the orders, and one of the owners prints each job. Here is what the agent already does, what stays with the team, and where we don't use artificial intelligence yet.
Where do quote requests come from?
Almost never from the website form. At this print shop, nine out of ten arrive by email, and each customer writes it their own way: which product, how many, sometimes with a file attached and sometimes without the size. So the automation starts in the inbox.
The agent reads every incoming email and decides whether the customer is asking for a price. Someone asking where their order is isn't, and that email stays off the list of quotes to do. If the agent gets it wrong, a person moves it onto the list with one click.
Where does the price come from?
The agent doesn't make up prices. It reads the request (which product, how many copies, which paper, which finish) and looks for the price in three places, in this order:
- The shop's own price calculator, the same one customers use on each product page. The agent can only pick the options that page offers, so it can't make up a finish either.
- The cost sheet of the printing company that prints the job, plus the shop's margin. For now, only for books.
- Past quotes for similar jobs.
Next to every price, the agent shows which of the three it came from. If none of them gives a price, a person works it out, and the agent notes why. And when it doesn't understand the request, it says what it couldn't read and doesn't guess.
What stays with the team?
The part that needs judgement. When a request is missing something (the size, the paper, the delivery date), the agent spots it and drafts an email with the questions. A person reads it, edits it if needed and sends it. The send button asks for confirmation: two clicks, so no email reaches a customer by mistake.
Then comes the follow-up. Whoever does it sees the calls, the emails and whether the customer opened the quote, all on one screen. For each quote, the agent writes one line on where it stands.
One mechanical step remains. The quote the customer receives goes out from the online shop's back office, and it has to be created there by hand: each detail is copied from the agent's screen with a button and pasted into the back office. The next step is for the agent to create it on its own.
And all of this is new. The agent already reads the inbox. The team starts working with the quote and follow-up screens in November 2026.
What about orders?
There was less to do there. The online shop already handled almost everything: each order goes to the printing company that prints it, the customer gets an email at every status change, and the printer marks when the shipment leaves. We left that alone. What we did was translate the payment reminders into Spanish and switch them on. Bank transfers are still marked as paid by hand.
Before adding artificial intelligence, check what your current software already does. Sometimes it already does the job, and the feature is switched off… or in the wrong language.
Where we don't use artificial intelligence yet
File checks. Every order stops until a person confirms that the file matches what was ordered, and the color profile is applied by hand to every one. It's repetitive, exactly the kind of work you want to automate. But a mistake there can't be fixed with another email: it comes out printed.
So it comes after quotes. It's the rule we always follow, and we explain it in “AI virtual employees: what they are and when they pay off”. Start with high-volume, low-risk work, and give the agent more as it proves it gets things right.
Is it worth it for your business?
Four things to look at:
- How do requests reach you? If they come by email, each in its own words, that's where the agent takes a lot of work off you. If they come through a form with fixed fields, you're halfway there without any agent.
- Is there a price list, a calculator or a cost sheet to get the price from? If only one person knows how to work out the price, the agent has nothing to go on.
- Do you keep the quotes you've already sent? When a job isn't on the price list, that's where the price comes from.
- Who signs off? Someone has to read every quote before it goes out, and at first, every single one.
Where to start
With the inbox. Have the agent separate what asks for a quote from what doesn't, and say what each request is missing. None of that reaches the customer without a person seeing it first, so a mistake there costs little.
The price comes next. Letting the agent send quotes nobody has read comes last of all. If it ever gets there.
If quote requests reach your business by email and get answered one by one, tell us about your case on a 30-minute call. We'll tell you plainly which part we'd automate and which we'd leave as it is.
Shall we talk about your quotes?
The first call is free and takes thirty minutes.
Let’s talk →