AI won’t kill the service desk. It will expose it.

Every executive I talk to asks the same question: “If AI can answer the tickets, why do we need a service desk?”

Wrong question.

AI doesn’t fix a broken service model. It runs it faster. If your catalog is a mess, your knowledge base is stale and every region handles requests its own way, a chatbot just gives the wrong answer to more people in less time.

I learned this the unglamorous way.

At Dolby, I led global IT support for 3,000 users and 15,000 endpoints across 11 countries. Before we leaned on automation, we did the boring work: we rebuilt the service catalog, request fulfillment and knowledge management so every ticket followed one clear model. Then we layered in AI and self-service.

In six months, mean time to resolution dropped from 5 days to 3, ticket volume fell from 60,000 to 42,000, and SLA compliance rose from 85 to 90 percent.

The AI got the credit in the demo. The process work made it possible.

Here’s what I think the next few years of ITSM actually look like:

1. The ticket queue shrinks, and the hard stuff gets harder. Password resets disappear. What’s left is complex, cross-functional and political. You need fewer agents and better ones.
2. Knowledge management becomes a strategic asset. Your AI is only as smart as the articles it learns from. Whoever owns knowledge owns the experience.
3. The service desk becomes the AI supervisor. Someone has to tune it, audit it and catch it when it’s confidently wrong. That someone sits in IT operations.
4. Vendor management gets a new layer. MSPs, SIAM partners and now AI platforms all need SLAs, governance and an exit plan.

The leaders who win won’t be the ones who bought AI first. They’ll be the ones who cleaned house first.

Over the next few weeks I’ll break each of these down with what worked, what didn’t and what I’d do differently.

So, to the CIOs and IT leaders here: what’s the one process you’d never let AI touch until it’s fixed?