How to put AI to work in your print business
Start with one problem worth solving
By Cole Lundstrom
Andrey Popov & Deemerwha studio – stock.adobe.com
Recently, I told someone on our team to “read the invoice because not everything needs to be AI.” That might sound strange coming from someone who uses it constantly. But if the event time is already on the invoice, then open the invoice to find it. You do not need a robot to turn a simple question into a research project. There are, however, plenty of amazing time saving tasks the AI can accomplish for your shop.
At Merch Troop, we use AI to help with the work surrounding our live merchandise events. Quotes need to be built, customer conversations need follow-up, and the team needs the right information before showing up onsite. These are ordinary business problems. They are also good places to put AI to work.
Put the AI in the middle of everything, everywhere, all at once
Our AI assistant is named Rosie, and our team talks to her in Slack. Someone can ask about an upcoming event or request an estimate in the same place we already chat with each other. That matters. Another tool nobody remembers to open will not do much for your business.
Behind the conversation, Rosie is connecting to GoHighLevel, our customer management system. She checks the business inbox for quote requests, routes items internally, and posts a summary in Slack every two hours. She also pulls together information from meeting notes. This is a game changer for me personally. I simply record my Google meeting with everything I promise the client, and Rosie takes the transcription and makes payable estimates and notes for us. Our OpenClaw setup runs on a VPS, a rented computer in the cloud, so scheduled work does not depend on an open laptop.
You do not need to copy our setup. Start with Claude, Codex, or any assistant you already use. Give it a customer request and your business information, and see whether it is actually useful. I have found myself developing many separate tasks, or crons, as being far more successful than a master instruction for all tasks. So, for example, we have a unique cron for only creating a live printing estimate. The instructions are clear, under 500 words, and the AI has a structure to succeed with.
Give it a specific task not a loose idea
One of our useful workflows starts with a request in Slack and ends with an estimate draft in GoHighLevel. Rosie organizes the scope and prepares the estimate for review. The team can correct it before anything goes to the customer.
Think about the alternative: Somebody reads an email, pulls out the details, checks pricing, opens another program, and types everything again. If AI can help move that request into a usable draft, you have a better starting point for the sales conversation.
But you have to tell it how your business works. Give it your services, pricing rules, examples of good estimates, and the questions it should ask when something is missing. If you do not know your own pricing, AI is not going to magically make you profitable.
One of our estimate drafts priced nine hours when the request already included setup and teardown in seven. That is a real mistake, and a useful reminder. Read the work. Check quantities, labor, shipping, and what you actually promised. A confident answer can still be wrong.
The level of review should depend on what happens next. A summary or a draft is easy to check. Sending a price, making a promise to a customer, or publishing something deserves more care. Give those decisions a clear approval step. Connect only the information the workflow needs, and be clear about what the assistant is allowed to change.
AntonioDiaz – stock.adobe.com
Stop relying on someone remembering
Rosie also posts scheduled summaries and reminders in Slack. Those can bring an upcoming event or an unfinished follow-up back into the conversation. This is useful because the task is easy to describe: look for work that needs attention and put it in front of the person who can do something about it. Those reminders are only as current as the records behind them.
A simple first automation for a print shop could be a weekday review of open quotes. Tell the assistant where to look, which quotes count as needing attention, and what to produce. For example: find quotes with no customer reply for three business days, check for a newer conversation, and prepare a short follow-up draft for the salesperson.
Then decide where the result belongs. It could be a Slack message with a link to the quote, a task in your customer management system, or an email draft. Make one person responsible for reviewing it. Start with a handful of real jobs and check the results before putting it on a schedule.
Scheduling tells the system when to run. AI helps it interpret the request and prepare the response. You can combine those pieces gradually instead of trying to automate your entire business on Monday.
The connection needs testing too. We still check for email inquiries that have not made it into the CRM. An assistant looking only at the CRM will miss those. Agree on where the current information lives and check that the workflow can actually see it. Otherwise, you have automated an incomplete picture.
Get the owner out of every little question
Our team uses Rosie to pull together event details from the information available to her. A summary can show what is known and what still needs to be collected. That gives people a place to start without every question becoming a call to me.
A traditional shop could use the same idea for production handoffs. Have AI organize the approved garment details, decoration locations, artwork status, and deadline. Keep links to the original information so the team can check it. Missing details should come back as questions, not guesses.
If you are the only person who knows what is happening, you have made yourself into a very expensive filing cabinet. Getting that information into a useful form helps the business keep moving while you are doing something else.
Use the time to sell something
The business opportunity goes beyond clearing your inbox. At Merch Troop, customers are buying an experience for their guests. AI can help work through a brief, organize proposal ideas, and explain how a merchandise activation fits the event. You still choose the idea and make sure your team can deliver it.
For your shop, that might mean suggesting a better company apparel program or showing a fundraiser a few options that fit its budget. Give AI the real costs and constraints. A useful alternative gives you more to talk about than another discount.
We also use Rosie to prepare case-study drafts from completed event work and bring them to us for approval. The job is already done. Now its story can help the next customer understand what we offer. The same material could become a social post or a relevant example in a proposal. Review the facts and use photos you have permission to share.
Start with one problem worth solving
Pick something you repeat every week. Define what starts the work, what information it needs, what a good result looks like, and who checks it. Get that working before adding the next connection.
Measure the whole job, including the time spent correcting AI. Are quotes getting reviewed sooner? Are follow-ups happening? Are fewer details being lost? Those answers matter more than a big claim about hours saved.
Then, use the room you created. Call a past customer. Build a better proposal. Go after the kind of work you actually want. The goal is a healthier, more profitable business that depends a little less on you remembering everything.
And when the answer is already on the invoice, you can still just read the invoice.
Cole Lundstrom is the founder of MerchTroop.com, specializing in live event apparel activations for brands and experiential marketing. With 16 years in the print industry, he’s now launching PrintShopCRM.com, a workflow and automation platform built specifically for print shops to scale smarter and operate more profitably.
