An AI receptionist for a print shop: what it should answer, and what it must never touch

Most of the calls and emails a print shop answers every day are the same nine questions. An AI receptionist should take those nine and nothing else. Here is the line, and why crossing it costs you orders.

A print shop does not have a phone problem. It has a repetition problem. The same nine questions arrive every day, by phone, by email, by text and through the contact form, and each one interrupts somebody who was in the middle of production. The interruption costs more than the answer.

That is the job an AI receptionist is genuinely good at. Not selling, not negotiating, not designing. Answering the nine questions that have exactly one correct answer, instantly, at two in the morning, in both languages, without pulling anybody off a press.

The nine questions it should own

Every one of these has a single factual answer that lives in your system already. If the answer is in your order records, your product data or your policy page, an AI receptionist can give it.

• Where is my order. The answer is a status and a date, read from the order record.
• What is your turnaround right now. A number you control and update in one place.
• What file do you need from me. Format, resolution, colour mode, and what happens if the file is wrong.
• What is your minimum on this. Per product, per decoration method.
• Do you have this in stock, in this colour, in this size.
• Where do I order again. A link to their store or their last order, ready to reorder.
• Can I reorder exactly what we did last time. Yes, and here is the link that does it.
• When will it ship and what will shipping cost.
• Are you open, and when can somebody call me back.

Nine questions. In most shops that is the large majority of inbound contact by volume, and close to none of it by value. Handing it to software is not cutting service. It is answering faster than a human could, and freeing the human for the part where the money is.

The four it must never touch

This is the part that gets skipped, and it is the part that decides whether the thing helps you or costs you money. An AI receptionist that oversteps does not fail loudly. It fails by quietly giving away a price you cannot honour.

• A custom price. Decorated goods price on quantity, colours, placements, garment and setup. A number produced without all five is a number you will either eat or argue about.
• A rush promise. Only the person who knows what is on the press this week can promise a date, and a missed rush date costs you the client, not the order.
• Artwork approval. Software can check a file against rules. It cannot decide that a logo will look wrong stitched at that size on that fabric.
• A complaint. Anybody unhappy enough to write in wants a person. Route it to one, fast, with the whole history attached.

What it needs before it can answer anything

An AI receptionist is only as good as what it can read. This is why most of them disappoint: they get installed on top of a shop where the answers do not exist in any system, so the software has nothing to look up and starts improvising.

• Order records it can read, with a real status per order, not a spreadsheet somebody updates on Fridays.
• Product data with sizes, colours, decoration methods and minimums attached.
• Written policies for turnaround, artwork, shipping and returns. One version, in one place.
• A clean handoff path to a named human, with the conversation history attached.
• A log of every question it could not answer. That log is your roadmap.

That last one matters more than the receptionist itself. The questions it cannot answer are the questions your customers keep asking that your business has never written down. Every week you read that list, you either write the answer or fix the thing that causes the question.

Where the receptionist stops and the store starts

Half of the nine questions vanish entirely when the customer has their own store and their own login. They do not ask where to order again, because the link is a bookmark. They do not ask about minimums, because the store only shows them what they can actually buy. They do not ask for their last order, because it is in their history.

That is worth saying plainly: the cheapest AI receptionist is a portal that makes most of the questions unnecessary, with software answering the rest. Automating the answering of a question you should have designed out is just a faster version of the same friction.

Setting one up without losing a quote

Roll it out in this order and nothing breaks:

• Start read only. For two weeks it drafts answers for a human to send. You see every answer before a customer does.
• Turn on the questions that are pure lookups first. Order status and turnaround.
• Add the policy answers next, once the policies are actually written.
• Keep the four forbidden ones routed to a human from day one, permanently.
• Read the unanswered log every week and fix the top item.

Two weeks of drafting gives you something no vendor demo will: the real list of what your customers ask, in their words. Most shops are surprised by it, and it usually changes what they automate first.