Planning for AI in IT service management starts with decisions a feature list can’t answer. What should your team automate? Where should people remain involved? What information will an automated workflow depend on? And how will you know whether the change delivered a useful result?
Before adding another AI initiative to your roadmap, make those decisions explicit.
Gartner® Top Trends in AI for ITSM for 2027 identifies seven trends shaping service management and recommended actions for heads of I&O: Agentic ITSM, ITSM Framework Depreciation, AI Agent Tracking, Trusted Service Data, Outcome-Based Measurements, Zero-Touch Service Desk, and Distributed ITSM Tooling.
Rather than summarizing all seven trends, this post draws on four to explore practical planning questions about where AI and automation may create meaningful value. The questions below are SolarWinds editorial planning prompts, not a summary of the Gartner recommendations.
1. What would you need to see before investing?
An AI investment discussion becomes more useful when it moves from labels to a specific scenario.
Choose a workflow your team understands. Describe the starting conditions, the result you want, and the limits on what the system may do. Consider the evidence you’d need before moving forward.
For example:
- What information would the system need to complete the work?
- How does it handle missing or contradictory information?
- Which actions require approval?
- Can your team understand and review what happened?
- What would count as a successful result?
In its Agentic ITSM section, Gartner states:
“Avoid overcommitting to agentic ITSM based on vendor roadmaps or marketing. Prioritize solutions that are demonstrable, supportable, and aligned with current I&O maturity.”
Bring those questions to your next evaluation discussion. A defined scenario gives the team something concrete to examine, rather than a general conversation about what AI might eventually do.
Not every workflow needs AI
A useful automation plan distinguishes work with a defined path from work that requires interpretation.
Your team can set rules to route a request to a designated team, request approval, and assign fulfillment tasks. Service desk automation can handle routine routing, approvals, and notifications without making every step an AI initiative.
For an application-access request, start by separating the work:
- What can follow an established rule, such as routing by request category?
- What requires a person’s approval?
- What needs further investigation or interpretation?
- What should stop until someone resolves an exception?
That breakdown gives you a clearer evaluation question: what does AI need to contribute beyond the workflow you can already define? Your team can examine the proposed capability against that specific need rather than treating AI as the default answer.
2. Which service request would you examine first?
“Automate the service desk” is an ambition. An individual service request is a place to begin a planning conversation.
Consider a request your team handles regularly, such as access to an application or a standard equipment request. Map the process as it works today: what the requester provides, who approves it, where work changes hands, and how the team handles exceptions.
The purpose of this exercise isn’t to assume every step needs AI. It’s to identify the work clearly enough to discuss the options.
In the Zero-Touch Service Desk section, Gartner states:
“Start with automation of service request workflows rather than attempting to automate incident management first. This will deliver quick wins without affecting user confidence in automation initiatives.”
For your own planning, choose one request and document the exceptions alongside the standard process. Those exceptions are useful discussion points for deciding where automation ends, and human involvement begins. Once the standard path is clear, a service catalog can make the request easier to complete and easier to fulfill. For an application-access request, it can collect the information employees need to provide, define eligibility and approval requirements, and set expectations for what happens next.
Workflow automation can then handle predictable routing, approvals, notifications, and fulfillment steps. Keep people involved when information is missing, an approval is denied, or the request falls outside the standard path.
Before putting the process into production, test the paths that are easy to overlook. Include approval denials, reassigned approvals, missing information, and alternative fulfillment paths. Teams can use Service Desk’s process-workflow test mode to review draft service catalog workflows before approval. Use that review to confirm that the next action, responsible person, and expected outcome are clear at every stage.
3. Which service data would the workflow rely on?
Examine the information your candidate workflow would use.
For an application-access request, that might include the application owner, the requester’s role, the required approval, and the permissions being requested. For an equipment request, it might include eligibility, available inventory, and who owns fulfillment.
Ask the people responsible for the process:
- Who maintains each piece of information?
- How would the team know it’s current?
- How does the process handle unavailable records?
- Which uncertainties would prevent the request from proceeding?
In the Trusted Service Data section, Gartner states:
“AI and advanced automation initiatives depend on accurate service context to function safely and effectively.”
Identify the information your chosen workflow needs and make responsibility for it visible. That gives your team a specific set of questions to resolve before expanding the initiative. The workflow also depends on service context. A CMDB can connect an application to the infrastructure and services it depends on, helping teams investigate incidents and assess the potential impact of a change. The goal is not to document everything before taking action. It is to make the information most important to the workflow visible, current, and owned.
4. What result would matter beyond the service desk?
Before introducing a new workflow, write down why your organization should care about the result.
For an access request, your team might examine how long an employee waits before they can begin the work that requires that application. For an equipment request, you might explore whether a new employee has what they need on their first day.
The useful question is whether your team can connect the proposed change to an organizational priority and explain how it’ll assess progress.
In the Outcome-Based Measurements section, Gartner states:
“Anchor all ITSM reporting dashboards or scorecards to value stories and outcomes that executives care about.”
Bring the research into your 2027 planning
You don’t need to settle every AI question in one planning session. Start by identifying the decisions your team needs to make: what to evaluate, which workflow to examine, what data it depends on, and how you’ll assess the result.
Access Gartner® Top Trends in AI for ITSM for 2027 as you consider those decisions and your team’s priorities for the year ahead.
