Contact Centres and AI

Here's what actually happens when a call centre agent can't find an answer.

Put the customer on hold. Ask the people sitting nearby. Flag down a floor walker. Post in the team chat and wait. If none of that works, transfer the call somewhere else.

One agent described exactly that in a thread I analysed for a research review. What struck me is that the same person said their knowledge base was good and they wouldn't change their tools. The information was there. They just couldn't get to it fast enough.

That sequence pulls in three or four people for one unanswered question. Nothing measures it. But it counts against the agent's handle time, which they are judged on.

I spent the last few weeks looking at this properly. 45 agent accounts, two Cornell surveys covering around 5,400 workers in the US and Canada, and a Quarterly Journal of Economics study of a real AI deployment.

The pattern underneath all of it: agents are the last stop for problems that started somewhere else. In 2017, 80% said customers frequently held them responsible for things they couldn't control. By 2023 that was 51%, which is still every second agent.

And the one type of AI workers actually rate positively is the type that helps them find information. Not the type that watches them.

Six years in call centres taught me the problem. The past year building with AI taught me why most tools aimed at it fail.