Keith Andes, here. I’m a former Gartner Analyst turned product marketer, and I’ve been researching AI impacts in ITSM. And I’m sorry, but we need to talk.

Most IT organizations today are eager to chase Artificial Intelligence (AI). On the other hand, while the pressure to be “AI first” continues to grow, IT leaders are still struggling to answer a simple question: Where do we actually start? We’re using it, generally. But how do we get value?

Interest in AI is high, executives are asking about it, and teams are experimenting with tools like ChatGPT. But few have a clear strategy for where and how to apply it in ways that actually create business value.

The result is what happens when a new tool arrives before a real problem. Everyone’s experimenting, but few know what they’re trying to fix. AI becomes a solution in search of a use case. It’s busy, exciting, and completely directionless. And even in pockets of useful outcomes, what works for one person using ChatGPT in isolation rarely scales across the organization.

The truth is, you don’t need to be AI first. You need to be AI last. I’ll explain why and how below.

(By the way, I’m AI bullish. I’m firmly in the “AI will define the future” camp, but I reject the fantasy that it will be easy.)

The AI-First Problem

“AI first” initiatives usually start from the top down: We need to use AI. The intent is good because no leader wants to miss the next wave of innovation, but the directive often lacks clarity. Teams are told to “do AI,” not to solve specific problems. That’s how organizations end up with scattered pilots, redundant projects, and employees burning through AI-compute like free candy.

According to recent EasyVista research, 95% of enterprise organizations say they’re already using AI in ITSM in some way (and for SMBs the adoption was still 90%). But usage doesn’t equal value. Without structure, alignment, and measurable outcomes, experimentation rarely scales, and teams walk away thinking AI was overhyped.

The real challenge isn’t whether you use AI. It’s whether you apply it where it actually makes a difference.

Fix What You’re Scaling

AI doesn’t fix broken processes (the opposite, actually: it magnifies them). So, before layering AI onto your ITSM stack, focus on the areas that already run well enough to benefit from acceleration.

If your incident routing is inconsistent today, AI will only misroute faster. If your CMDB is messy, AI-driven insights will just mirror the noise. Garbage in, garbage out.

That doesn’t mean you need perfect processes before you start, by the way. That’s not where I’m going with this. It just means you should begin in the parts of your operation that already work. Where you have reliable data, clear workflows, and consistent outcomes, that’s where AI will make the biggest impact. In messy areas, it will only add more noise (how will you even know the AI got it right?).

Think of AI as a force multiplier for competence, not a cure for dysfunction. It performs best in mature parts of your environment, where automation and human judgment already work hand in hand. Start there, prove value, and expand outward.

Build Capability Through Shared Learning

Every successful AI strategy starts with learning. Encourage experimentation, but make the learning collective. Let your teams play with ChatGPT or other AI tools, but don’t let that learning stay siloed.

Host a short monthly “AI Hacks” session (it helps if each session focuses on a single role) where team members share how they used AI that month:

• Where did it save them the most time?
• How did they apply it to a workflow?
• Was there a creative use case others could replicate?

These stories become your internal playbook. Over time, you’ll see patterns where AI consistently saves time, improves experience, or reduces manual effort. Those are the use cases worth scaling, so document and build from them.

As your teams build confidence and internal knowledge, the next step is to connect that learning to actual business processes. That’s where being “AI last” comes in—deciding where AI truly belongs in your workflows and where simpler automation will do.

Be AI Last (Ask Smarter Questions)

Once you have a foundation of shared learning, start translating those insights into real process improvements.

Instead of asking, “Where can we use AI?” start by asking, “What’s the simplest way to solve this?”

Use lightweight automation where possible. For example, automate ticket categorization with logic first. When exceptions pile up, machine learning might be the right next step. The point isn’t to avoid AI. It’s to use it where it earns its place.

Start with process maturity. Automate what you can through rules and workflows first. If you can use an if-then statement to solve the problem, then you don’t need AI. That’s a cheaper, more consistent, and easier –to manage solution.

And yes, ask your vendors for ideas. Just make sure you’re asking the right questions:

“What use cases can AI solve that we couldn’t automate otherwise?”

That’s the key one. Other questions include:

• How are they applying AI internally?
• What successes have they had with other customer deployments?
• Where are people seeing the fastest ROI?

Then validate those claims. Analyst research, such as Gartner’s AI Use-Case Assessment for IT Service Desk, ranks AI applications in ITSM by feasibility and impact. That’s the kind of data you want guiding your roadmap, not a product brochure.

Once your processes and learning culture are in place, start looking at where AI is already delivering impact in ITSM. From what we’re seeing with customers and across the market, the most promising areas right now include:

• Portal or Teams bots for conversational access to services.
• LLM search to deliver context-aware results across KBAs and tickets.
• Incident correlation and anomaly detection for pattern recognition (if you’ve got great sensors and integrations in place).
• Knowledge synthesis to summarize resolution data, not just auto-generate articles.

The goal isn’t to reject AI, of course. Instead, we just need to use it deliberately and intelligently.

The Bottom Line

We should be excited about the AI future, but getting value today does not come from being generically “AI first.” It comes from understanding the technology and what kinds of problems you can solve with it that were unsolvable without it.

But think about processes and automation first, and only apply AI where AI makes sense. Start simple. Fix what’s broken. Share what works. Then apply AI where it truly extends your capability.

AI won’t fix what’s broken. But if you build a solid foundation and use it intentionally, it will revolutionize your operations.