A Complete Playbook for AI Adoption in Maintenance: How One Company Rolled Out AI Across 200 Facilities

AI adoption in maintenance is much more about people than technology. Like any tool, AI is only useful if it’s used consistently and correctly. But if your frontline team doesn’t trust it or know when or how to use it, your AI set-up will gather dust.

Eric Ayanegui, Sr. Director of Operations Engineering at Cintas, knows has been through this process. He and his team rolled out AI to 900 technicians across 200 sites, learning quite a few lessons along the way. We compiled those learnings into a playbook for how to get your maintenance team to adopt AI. This guide breaks down the approach taken at Cintas, from running phased AI rollouts to building dashboards to measure adoption.

Key takeaways

• AI adoption works best when teams start with clean enough data, including consistent asset names, matched OEM manuals, and usable work order history.
• A phased rollout gives maintenance teams room to test AI, build trust, collect feedback, and improve the process before expanding across more sites.
• Power users, required trials, and adoption dashboards help turn AI from a one-time initiative into a tool technicians use consistently and correctly.

Prepare your maintenance data before adopting AI

Teams often delay AI adoption because they think they need perfect data first. The reality is more forgiving. Here is how Eric and team prepared their data for adopting AI:

Standardize asset naming conventions

One of the first things Eric's team did was an audit of their asset naming conventions across their documentation. They discovered that a lot of assets had different names across sites and resources, which they knew they had to fix before rolling out their AI plan. They worked to standardize names, serial numbers, models, and priority levels before moving forward.

Collect equipment manuals

The Cintas team used OEM manuals as the primary knowledge base for their AI tools. These manuals would provide a reliable base for everything from generating procedures to offering troubleshooting help. The team made sure that they had as many manuals in a digital format as possible and that they were correctly matched with the right assets.

Clean up procedures

The AI tools that Eric was helping to set up also rely on existing PM procedures and repair documentation, scanning past work orders to learn everything from common failure modes to successful troubleshooting steps. This required them to standardize fields, ensure work order titles and tags were clear and consistent, and a commitment to more detailed notes in the future.

Create a phased rollout plan for AI adoption

A phased approach to AI adoption can keep your team and organization from feeling overwhelmed, even if you’re not managing an AI roll out across 200 sites. Here’s a look at how Eric and the Cintas team structured their gradual adoption strategy.

Phase 1: Start where you have the best chance at success

Eric’s team started their AI rollout with pilot locations willing to test and provide feedback. They selected these locations based on readiness, willingness, how similar their equipment was, and how close they were to each other. They took the feedback and successes of these pilot sites and used them to develop stronger rollout strategies and to prove to other sites that there was a lot of value in the AI tools.

Lesson: Give your strategy the highest chance of success by starting with a group of people or processes with the clearest path to success instead of trying to apply AI to everything and everyone all at once.

Phase 2: Identify what’s working, what isn’t, and revise your plan accordingly

Eric and his team didn’t forget about sites after the initial AI implementation. Instead, they used regular check-ins to get feedback and wins to strengthen their program and subsequent rollouts.

"One every call with our sites, we’d leave time for them to tell us about their wins, their challenges, and the interesting or valuable ways the team was using AI,” says Eric. He’d then use this information to continually update their strategy for the other sites.

Lesson: Build flexibility into your AI adoption plan. Create a regular feedback loop, like a weekly 15-minute meeting, to gather both critical and positive feedback as well as to share ideas on how to use AI. Use these insights to make changes to your strategy so it gets better and better at each step.

Phase 3: Connect AI workflows

One of the biggest milestones on Cintas’ AI journey was when they were able to connect information and workflows across sites. The AI was creating procedures and sharing it across facilities so it became the new standard for all 200 locations. This connection was important because it connected the work being done across sites and amplified the impact of their AI tools.

Lesson: The early stages of AI adoption are about getting comfortable with the tools and learning how to use them to solve problems. The next step is to tie these disconnected use cases together so entire parts of your operation are powered by AI. Finding these connections means keeping users in the loop as they’re the ones using AI every single day.

Identify and empower AI power users

AI power users are the ones that use the AI tools most, develop new user cases, and get the most from the tech. Tapping into their experience and enthusiasm is key to improving AI adoption. Here’s how Eric and the Cintas team harnessed their power users:

• They used engagement data to spot power users: Look for technicians with high engagement in the first weeks. Power users often explore beyond basic use cases, asking proactive questions rather than waiting for problems.
• They reinforced good habits through recognition: Power users need acknowledgment. They need to know that they are doing the right thing and that experimenting and pushing boundaries is a good thing. Eric did this through Public recognition during team calls.
• They used power users to influence their peers: Peer influence often succeeds where top-down mandates fail. Eric observes that some skeptics converted because they heard from another person on their team that they used it for a new task. Pairing power users with resistant team members during rollout can accomplish more than any training session.

Use required trials to convert AI skeptics

Eric knew that some of the company’s 900 technicians would have an ‘I'll use it when I need it’ mindset to AI. He also knew that this presented a risk that the right time would never arrive and the pressure of using a new tool during an emergency would increase skepticism.

He addresses this concern with a mandatory first-try approach. "We set a minimum requirement that they had to try it,” says Eric about their AI tools. “After that, we could have a further discussion, but you had to try it.”

Here’s how the Cintas team designed their try-first approach:

Build trust with a mandatory experimentation period

Eric framed this step not as a strict mandate with consequences, but as a way for technicians to avoid the stress of figuring out a new tool while trying to get a broken production asset up and running again.

"We told them, ‘We want you to use the AI before you need it so that you believe,’" says Eric. “After you’ve tried it, we can then decide what your opinion is, but that opinion is now informed by actual use.”

Test AI on known problems first

Eric had some experienced technicians test AI by asking about repairs they already knew how to perform. When the AI returned the same answer that the technicians had, they were much more likely to trust the tools. When the AI confirmed what they knew, it built trust and broke down some of the skepticism they had. Even when the AI occasionally gave a different answer, it opened a productive conversation, says Eric.

"Sometimes, it wasn’t exact, and that's great,” says Eric, “because then we identify the gap between the procedure that was provided by the manufacturer and what actually happens on the ground."

Set up dashboards to track AI adoption and value

Without visibility into who's using AI tools, how often, and the results, you can only guess at the success of your initiative. Here is how Eric's team set up reports to show them where adoption was working and identify areas of risk so they could address them while they were small.

User engagement metrics

Key metrics at the individual level include:

• Interaction frequency: How often each technician queries the AI
• Query types: Troubleshooting vs PM preparation vs procedure lookups
• Usage timing: During breakdowns only vs proactively before issues arise

Site-level adoption

Eric's team shares adoption data with site leaders directly so they know if they’re falling behind or leading the pack. This allows them to give recognition to the team or offer additional support and training to bridge the gap.

Qualitative feedback and direct quotes

Numbers only tell part of the story. Eric's team captures exact quotes when talking to technicians so they can use them in training sessions with other technicians. During rollout calls, they leave time for technicians to share wins. A technician who says, "I got the answer in ten seconds instead of walking to the maintenance shop" is a more valuable benchmark than a lot of engagement metrics.

Speed and response time improvements

Eric and his team like to compare the time it takes to complete a step or a process without AI and with AI. This allows them to not only calculate how much it helps reduce downtime and increase technician efficiency, but also gives technicians a tangible way to understand the frustrations and challenges AI can solve for them.

Knowledge capture metrics

The Cintas team is also tracking how much team knowledge AI is helping them preserve through:

• The number of procedures created using AI's procedure generation features
• Troubleshooting scenarios documented in work order comments
• Procedures shared across locations

Cultural transformation indicators

Although this doesn’t show up in any dashboard, Eric observed that after implementing AI, technicians "Started to become more open minded" beyond just the AI tools and the CMMS.

Design AI adoption around your people, not just your tech

AI adoption doesn’t happen because the technology exists. It happens when technicians trust the tool, know when to use it, and see how it helps them solve real problems faster. Cintas’ rollout shows that the strongest AI strategies start small, build feedback loops, and measure adoption as closely as performance. With the right data, power users, and rollout plan, maintenance teams can turn AI into a practical part of daily work instead of another unused system.