One of the most common questions we hear from IT teams right now is: where do I start with AI?
And that’s understandable. Every ITSM platform now offers AI features. Your leadership team is asking about it. Meanwhile, your inbox is full of vendors promising transformation. But when it comes to making a decision, the options are overwhelming and the path forward isn’t clear.
You’re not alone. According to the Service Desk Institute, 71% of organisations are still in the research or pilot stage of AI adoption for IT support and ITSM operations. Most IT teams are still working out where to begin.
So, let’s break it down.
Start with the problem, not the technology
It’s tempting to look at what AI can do and work backwards. But that’s how you end up with a chatbot nobody uses or an auto categorisation feature that creates more work than it saves.
The better approach is to look at where your team is spending time on repetitive, low value work. What are the tickets that come in repeatedly? Password resets. VPN issues. Access requests. Software installation queries. These high volume, predictable ticket types are where AI can have immediate impact.
The ITSM.tools State of AI in ITSM 2025 survey found that the top challenge for IT teams was “automating repetitive tasks to improve efficiency” cited by 45% of respondents. If that sounds familiar, you’ve already identified your starting point.
Gartner’s advice aligns with this. In a 2025 survey of infrastructure and operations leaders, they recommended that I&O leaders “start with high value, feasible pilots and flexible upgrades” rather than chasing big AI projects.
Your data matters more than you think
AI is only as good as the data it learns from. If your knowledge base is out of date, your ticket categories are inconsistent, or your asset records are incomplete, AI will inherit those problems.
But it’s not just knowledge articles. There’s a whole set of ITSM data that AI relies on to work properly:
- CMDB data: If your configuration items aren’t accurate or your CI relationships are incomplete, AI can’t map incidents to affected services or understand the impact of a change. You’ll get misrouted tickets and impact assessments that miss the mark.
- Ownership data: AI needs to know who’s responsible for what. If your service ownership, support group assignments, or technical ownership fields are out of date, automated routing and escalations will go to the wrong people.
- Escalation paths: If your escalation rules are inconsistent or haven’t been reviewed in years, AI will follow them anyway. Garbage in, garbage out.
- Service catalogue: If your catalogue doesn’t reflect what you offer, AI powered self-service will suggest options that don’t exist or miss the ones that do.
This doesn’t mean everything needs to be perfect before you begin. But it does mean being realistic about where you are. Pick the data that matters most for your first use case and focus there.
This isn’t just our view. Gartner’s 2025 Market Guide for Customer Service Knowledge Management makes the point clearly: “Knowledge management is no longer just a support function. It’s the foundation of successful AI adoption. Without structured, well governed knowledge, GenAI can confidently serve the wrong answer.”
Many AI initiatives stall not because the technology fails, but because the underlying data wasn’t ready. It’s worth spending time here before you switch anything on.
Governance isn’t optional
Here’s a statistic that should get your attention: 51% of IT professionals cite governance and compliance as their top barrier to AI adoption, according to the ITSM.tools 2025 survey. And yet, the same research found that 80% of ITSM professionals are already using free AI tools like ChatGPT at work.
That gap is a problem. Your team is likely using AI already – just not in a way that’s visible, governed, or secure.
The good news is you probably don’t need to start from scratch. If you’ve got ITSM governance structures in place – change advisory boards, risk assessment processes, data handling policies – AI governance can build on what’s already there.
A few questions worth working through:
- Who approves new AI use cases? Could this sit with your existing CAB or a similar body?
- What data is AI allowed to access and process? Your current data classification policies are a starting point.
- How do you handle AI outputs that need human review? This is similar to how you’d manage any automation that affects live services.
- What’s your policy on staff using external AI tools like ChatGPT for work tasks?
These questions are easier to answer now than after something goes wrong. And if you’ve already done the work on ITIL aligned governance, you’re not starting from zero you’re extending what you’ve built.
Pick a use case you can measure
One of the risks with AI is investing time and budget without knowing whether it’s helping. The ITSM.tools survey found that 44% of organisations said it was “too early to tell” whether AI had improved their efficiency.
The way to avoid this is to pick a use case with clear, measurable outcomes – and be honest about what you’re measuring.
A few principles:
Measure what matters to the business, not just the toolset. Faster ticket resolution means nothing if the answers are wrong and tickets keep reopening. Track outcomes, not just activity.
Pick metrics you can compare before and after. If you’re automating password resets, track how many are resolved without human intervention. If you’re using AI for routing, measure reassignment rates. If you’re suggesting knowledge articles, check whether first contact resolution improves.
Don’t ignore the human side. Customer satisfaction and analyst experience matter. If AI is making your team’s lives harder or frustrating end users, that’s a cost – even if the numbers look good on paper.
Be realistic about timelines. You won’t see the full picture in week one. Build in checkpoints – 30 days, 90 days – and be prepared to adjust.
This also helps you build a case for further investment. The same survey found that organisations allocating more than 10% of their IT budget to AI were far more likely to report positive ROI. Starting small and proving value is how you get there.
You don’t have to figure this out alone
AI in ITSM is still new territory for most organisations. McKinsey’s 2025 State of AI survey found that agent use is most commonly reported in IT and knowledge management – but most organisations scaling AI agents are only doing so in one or two functions. Everyone is learning as they go.
The teams getting the most value aren’t necessarily the ones with the biggest budgets – they’re the ones asking the right questions and learning from others in similar situations.
That’s why we run roundtable sessions: to bring IT professionals together to share what’s working, what isn’t and how to move forward. If you’re trying to make sense of AI in your service desk, it helps to be in a room with people asking the same questions you are.
