Artificial Intelligence has quickly become one of the biggest talking points in IT Service Management. Almost every ITSM vendor now has an AI story, from virtual agents and intelligent ticket routing to automated summaries, knowledge generation and predictive analytics.
There is good reason for the excitement. Used well, AI can remove repetitive work, help service desks respond faster and give IT teams insights that would be difficult to uncover manually.
But there is another side to the conversation.
Introducing AI into ITSM does not automatically create a better service desk. Poor data, weak governance, unreliable outputs, security concerns and unrealistic expectations can turn an AI initiative into another expensive technology project that fails to deliver meaningful value.
The interesting question, therefore, is not:
“How can we add AI to our ITSM tool?”
It is:
“Where can AI genuinely improve the way we deliver and manage services?”
Where AI Can Add Real Value to ITSM
AI has potential across many ITSM practices, but some use cases are already particularly relevant.
1. Smarter Incident Classification and Assignment
Anyone who has worked on a busy service desk knows how much time can disappear into categorising, prioritising and assigning tickets.
AI can analyse the description of an incident, compare it with historical records and suggest:
- Category and subcategory
- Assignment group
- Priority
- Related services or CIs
- Similar previous incidents
- Relevant knowledge articles
Even when the final decision remains with a technician, reducing these repetitive administrative tasks can make a significant difference.
More importantly, better classification means better data. Better data eventually means better reporting, problem identification and continual improvement.
2. Giving Technicians Better Information, Faster
One of the most promising uses of generative AI in ITSM is not replacing the service desk technician, but giving the technician a better assistant.
Imagine opening an incident and immediately seeing a concise summary of a long conversation, similar historical incidents, relevant knowledge articles and suggested troubleshooting steps.
Instead of searching through several systems, the technician starts with useful context.
This can reduce resolution time while allowing experienced technicians to concentrate on diagnosis and decision-making.
3. Improving Self-Service
Traditional self-service often expects users to know what they are looking for.
“Choose the correct category.”
“Search the knowledge base.”
“Complete this form.”
Users do not necessarily understand the organisation’s service catalogue or support structure. They simply know what they need.
Conversational AI can potentially change that interaction.
A user could explain their problem naturally:
“I can’t connect to the VPN since changing my password this morning.”
The AI can then identify the likely issue, locate appropriate guidance, ask relevant questions or create a properly categorised incident when human support is required.
That creates a very different self-service experience.
4. Turning Knowledge into Something People Actually Use
Most organisations do not suffer from having absolutely no knowledge.
They suffer from knowledge being fragmented, outdated, duplicated or difficult to find.
AI can help identify relevant information across knowledge articles, previous incidents and technical documentation. It can also assist with summarising information and identifying potential knowledge gaps.
There is an important distinction, however:
AI can help create and retrieve knowledge. It should not automatically make that knowledge trustworthy.
Human validation and knowledge governance still matter.
5. Finding Patterns Hidden in ITSM Data
ITSM platforms can contain years of valuable operational data.
Thousands or even millions of incidents, requests, changes, problems and configuration records may reveal patterns that are difficult for people to identify manually.
AI and machine learning can help uncover:
- Recurring incident patterns
- Emerging service degradation
- Common causes of SLA breaches
- Unusual ticket volumes
- Change-related incident trends
- Frequently failing configuration items
- Opportunities for automation
This is where AI becomes particularly interesting for continual improvement.
Instead of simply reporting what happened last month, IT teams can start asking what the data is telling them about what should change next.
The Risks We Shouldn’t Ignore
The potential is significant, but so are the risks.
AI Can Be Confidently Wrong
Generative AI can produce an answer that sounds completely convincing and is still incorrect.
In a general conversation, that may be inconvenient.
In IT operations, incorrect guidance could result in inappropriate troubleshooting, incorrect configuration changes or misleading information being given to users.
AI-generated recommendations therefore need controls appropriate to their potential impact.
There is a major difference between:
“Summarise this incident.”
and:
“Execute this remediation action on the production server.”
The level of governance should reflect that difference.
Sensitive Data Can Become an AI Governance Problem
ITSM systems contain significant amounts of organisational information.
Tickets may include:
- Personal information
- Usernames and email addresses
- Infrastructure details
- Security incidents
- Application information
- Configuration data
- Business-sensitive discussions
- Customer information
Sending that information to an AI service without understanding where it goes, how it is processed and whether it is retained creates obvious security, privacy and compliance concerns.
AI adoption therefore needs to involve security, information governance and data protection teams from the beginning.
Automation Bias Is a Human Risk
There is another risk that receives less attention.
If an AI recommendation is correct 95% of the time, people may gradually stop questioning it.
That remaining 5% then becomes much more dangerous.
A technician may accept an incorrect categorisation, troubleshooting recommendation or risk assessment simply because “the AI suggested it”.
Good AI governance must therefore preserve meaningful human judgement rather than turning people into approval buttons for automated decisions.
Your AI Is Only as Good as the Environment Around It
This may be one of the least exciting but most important points.
Imagine training or grounding an AI system using years of incident data where:
- Categories were inconsistently selected.
- Priorities were routinely overridden.
- Resolution codes were rarely completed properly.
- Duplicate tickets were everywhere.
- Knowledge articles had not been reviewed for years.
- CI relationships were incomplete.
- Service mappings were inaccurate.
AI will not magically repair the underlying management problems.
It may simply process bad information much faster.
Before investing heavily in sophisticated AI capabilities, organisations should ask whether their ITSM foundations are mature enough to support them.
The Integration Challenge
The service desk does not operate in isolation.
Useful AI may need context from multiple sources:
ITSM records, the CMDB, monitoring platforms, knowledge repositories, identity systems, collaboration tools, endpoint management and other enterprise systems.
Connecting these safely and reliably can be considerably more difficult than demonstrating an impressive chatbot.
The challenge is therefore not simply selecting an AI model.
It is designing the architecture, integrations, permissions, data flows and governance around it.
Don’t Automate a Bad Process
This lesson predates AI, but AI makes it even more important.
If an approval workflow contains five unnecessary approval stages, adding AI will not make the underlying process sensible.
If users cannot understand your service catalogue, putting a chatbot in front of it does not necessarily fix the catalogue.
If your incident categorisation structure is badly designed, automatically categorising incidents against it simply creates bad categorisation more efficiently.
Before asking:
“Can AI automate this?”
ask:
“Should this process work this way in the first place?”
Sometimes simplification delivers more value than automation.
Start with the Problem, Not the AI
A practical AI journey in ITSM should begin with a business or service management problem.
For example:
“We spend too much time manually assigning incidents.”
“Our technicians struggle to find relevant knowledge.”
“Users abandon self-service because finding the right service is difficult.”
“We have years of incident data but struggle to identify recurring patterns.”
These are problems worth investigating.
Once the problem is understood, AI can be evaluated alongside automation, process redesign, better knowledge management or other possible solutions.
This avoids the increasingly common situation where an organisation buys AI capabilities and then starts searching for problems to justify them.
Finding the Right Balance
AI is unlikely to replace IT Service Management.
It is much more likely to change how IT Service Management is performed.
The service desk technician may spend less time categorising tickets and more time solving complex problems.
Problem managers may spend less time manually searching incident records and more time investigating patterns.
Service owners may receive insights that previously required hours of analysis.
Users may interact with IT through natural conversations instead of navigating complicated forms.
That is a meaningful transformation.
But the organisations that benefit most from AI will probably not be those that automate everything first.
They will be the ones that understand where human judgement creates value, where automation makes sense, what data can be trusted, and where appropriate governance must remain in control.
At Northera IT Solutions, we believe AI should support good service management rather than distract from it. From assessing ITSM maturity and identifying practical AI opportunities to improving processes, CMDBs, knowledge management, automation and ITSM platforms, the objective should always remain the same:
Start with the problem. Understand the value. Apply the right technology. Measure whether it actually made things better.
