AI Adoption Is an Operating Model Decision

By Christine Perkins

The company rolls out an AI tool. People attend the training. A few weeks later, the customer service team can draft a response in minutes. Customers are still waiting just as long to hear back.

At the next leadership meeting, someone asks why. The answer becomes clear when the team walks through an actual request. The draft took three minutes. It spent the rest of the afternoon waiting for the same manager who had always approved it.

Follow the work from start to finish

Imagine a customer writes in because they were charged twice. The AI tool helps the employee draft a response. Before sending it, the employee still has to check the account, confirm the duplicate charge, and get approval for a refund. The draft is ready before anyone has decided what the company can do for the customer.

If you’re leading that team, walk through a few recent requests with the people who handle them. Ask what they can verify on their own, what they can decide, and where they have to wait. You may find that the team can safely resolve some requests without another approval. You may also find that the employee needs access to information they currently have to ask someone else to look up. Those are decisions about how the team works. Buying the tool leaves those decisions with you.

Make it clear who can make the call

Say you decide that employees can resolve confirmed duplicate charges up to an agreed amount. Show them what a confirmed charge looks like, what they need to check before issuing a refund, and when to ask a manager for help. Make it clear that the employee is responsible for checking the facts and the resolution before sending a reply.

Then pay attention to what happens the first time someone uses that authority. If a manager asks, “Why didn’t you run that by me?” after the employee followed the new process, the team will learn to keep asking for approval. Leaders have to reinforce the change in the moments when people actually use it.

What to do now

1. In customer service, start with one request the team sees every week.

Pull a few recent duplicate-charge cases and trace them from the first message to the refund. Find out where customers spent the most time waiting. If routine approvals were the bottleneck, agree on which cases employees can resolve and which need a manager. Test the process for two weeks. Look at how long it took customers to get a resolution and whether any refunds had to be corrected.

2. In manufacturing, start with the handoff between shifts.

A plant uses AI to summarize shift notes. The incoming supervisor gets a cleaner report, but still has to call the outgoing supervisor to find out who’s handling an open equipment issue. Pick one production line. Have the outgoing supervisor check the summary against the shift log, then record who owns each unresolved issue and what they need to do next. After a week, ask the incoming supervisors which issues they could act on at the start of the shift and which still took a phone call.

At your next AI review, bring one piece of work the team completed with the tool. Ask where it saved time, where the work still sat, and who can change what happens there. That’s where the next decision belongs.

If your team is piloting AI and work is still getting stuck at the same approvals, start by clarifying who can make each decision. Explore change management support to help your team put those decisions into practice.

Perkin Growth Advisory

Chrstine Perkins

Leadership and Growth Advisory


Let's Connect

Leave your email and I'll reach out to learn how I can help.