How to Start Your AI Journey on the Right Foot

There’s been a lot of buzz around artificial intelligence (AI), but using AI in a way that truly makes a positive impact on business has eluded many print service providers (PSPs).

Attendees of the “AI in the Print Shop: Beyond the Hype — A Practitioner's Guide to Real-World Implementation” webinar walked away with a clear, practical framework for identifying where AI belongs in their workflows, helping them see that positive impact.

Amy Servi-Bonner, vice president of PRINTING AI, kicked things off with a crucial reminder: AI is purely a tool, not a replacement for skilled employees with years of print expertise.

“It’s a pattern-matching system that is very fast, very literal, and only as good as what information you put in it,” she said. “… AI does not know your shop — you do. But that combination between your knowledge and AI’s speed is exactly where the value lives.”

She warned that shops that try to use AI as a decision-maker or an employee replacement but don’t keep a human in the loop will fail.

“AI does not make mistakes the way we all do,” Servi-Bonner said. “It makes mistakes confidently, at speed, and it makes the same mistake over and over and over again until a human catches it. In print, some of these mistakes have real dollar amounts attached to them.”

Being Prepared to Use AI

For printing and packaging companies, the holdup on AI ROI often lies in implementation.

“AI is genuinely capable of things that would’ve seemed impossible two, three years ago — even a year ago,” Servi-Bonner said. “It’s not an issue of if AI works, the issue is really the gap between what it does in a demo and what it does in your shop on a real job on a real deadline.”

Assessing AI Readiness

Most PSPs don’t know where the data their employees feed into AI tools goes, according to Servi-Bonner — and that’s a problem.

“This is not a criticism, it’s just a gap that got created because the tools showed up so fast, and nobody had time to build any policies before people started using them,” she said. “... But ignorance of the policy doesn’t protect you from the consequence.”

So, before investing in any new AI tool, she urged attendees to take stock of four key areas:

• Data sovereignty (where data goes).
• Tribal knowledge risk (what information lives in an employee’s head).
• Process flow (how processes actually work).
• Client obligations (e.g., NDAs, SLAs, regulated content).

She also outlined the AI Work Classification Framework, which enables print and packaging providers to evaluate what operational tasks require greater data security. This indicates what type of AI tool is safe to use for it and what sort of governance is necessary.

Evaluating Vendors

Once you’ve gotten a sense of your data situation, it’s time to start looking at AI solutions from different vendors. Dakota Hawkins, founder and lead facilitator at Copilot Courage, emphasized that all vendors are prepared for demos, so you need to put them to the test.

“What I want you guys to start thinking about is how to critically analyze what the vendors are showing you, what the vendors are telling you, and understand what the next step needs to be,” Hawkins said, “whether that’s continuing the relationship, exploring opportunities further, or taking a step back and seeing if the vendor selection itself needs to be reassessed.”

Hawkins proposed three sequential questions that print and packaging providers should ask prospective vendors, plus what constitutes a red-flag answer.

1. Does your tool train on my data, and exactly how is it partitioned? It’s a red flag if the vendor vaguely says the data is secure and can’t provide a written data-handling policy before you sign.
2. Show me a print-specific failure and how your system caught it. It’s a red flag if the vendor says it doesn’t make mistakes. If they can’t point to a print-specific failure mode, it means they’ve never really run the tool in a live print setting.
3. How does this connect to my MIS without creating a new data silo? If the vendor says the integration is seamless but can’t name a connector, an owner, or a maintenance plan, that’s a problem. “Seamless” is an adjective, not data architecture.

Once PSPs have done their due diligence on these pieces, it’s time to put AI to work. Hawkins and Servi-Bonner did several live prompt demonstrations that printing and packaging companies can use to see the biggest impact on their bottom lines.

Didn’t make the session? Be sure to watch out for the next webinar in the series, and subscribe to the PRINTING AI e-newsletter for the latest updates.