In the context of IT Service Management (ITSM), there is a growing interest in automation, predictive analytics, and AI-supported IT support for increasing efficiency and improving user satisfaction. At the same time, the discussion is picking up around the potential downside: over-focusing on AI can complicate things, and there will be situations where it is too early to deal with AI. In that sense, there is a pragmatic variant of this approach that is sometimes called âAI Last.â
The Challenge: When AI Outpaces Operational Readiness
Many IT organizations are under pressure to adopt AI-driven tools at a rapid pace. While the promise of faster ticket turnarounds and lower operating expenses is enticing, the reality suggests that some processes arenât mature or the data required just isnât up to snuff.
Typical challenges include:
- Workflows that are inconsistent or not well-documented
- Data stored in multiple systems and disconnected sources
- A lack of governance for automation tasks
- Too much trust placed in machines rather than developers or operators
When these factors are present, AI engines can exacerbate rather than solve any issues. For example, applying automation to a process that is broken, such as incident management, speeds results, although they still are incorrect.
Context: What âAI Lastâ Really Means
AI Last doesnât mean no AI. It means we should develop the foundation for operations first before driving automation, which leads to better performance runs.
This includes:
- Well-defined ITSM processes (incident, problem, change management)
- Good quality data
- Service catalog and workflows defined
- Monitoring & feedback loop
Once we have the foundation in place, we can start to use AI capabilities.
How Organizations Should Respond
1. Prioritize Process Maturity
In order to implement AI, IT teams first have to understand organization processes and levels of maturity, like ITIL, and have processes mature enough that AI systems can deliver consistent results.
2. Invest in Data Quality and Integration
Effective AI models are dependent on accurate integrated data sources. Organizations need to:
- Aggregate data sources
- Remove duplicate and inconsistent data
- Correctly classify the data
Without this data foundation, AI wonât deliver the results organizations are looking for.
3. Adopt Incremental Automation
Instead of big bang implementation of an AI system, little automation to start with serves the purpose for an organization:
- Automation of repetitive incident requests
- Building a rule-based of workflows for common incidents
- Starting with machine learning for detecting patterns of incidents
This approach reduces the automation and operational risk to the organization over time and will incrementally build confidence in the technology.
4. Maintain Human Oversight
Even in implementing the best AI systems, IT firms should never forget human intervention. For multiple reasons:
- To make sure that decision-making is still pretty much aligned with the ultimate outcome of the firm
- Handling exceptions efficiently
- Guiding how firms deal with ethical and compliance standards
AI should always be seen to be augmenting, rather than replacing, human judgment in ITSM operations.
Strategic Significance
This trend toward AI Last is part of a larger movement in enterprise IT priorities. Organizations are realizing that technology is not a magical panacea that can fix underlying operational issues. For meaningful change to happen, people, process, and technology must be in harmony.
By deferring AI deployments until the fundamentals are in place, businesses can:
- Drive more predictable outcomes
- Diminish deployment risks
- Maximize cost efficiencies
In the longer term, it also increases overall scalability, because well-performing systems will more easily integrate and take advantage of more advanced capabilities over time.
Conclusion: A More Measured Path to AI Adoption
AI offers considerable promise as an IT service management game changer, but the path to prevailing outcomes is more nuanced, reliant on both timing and maturity. The AI Last approach is a methodological structure for ensuring a strategic balance that values maturity over velocity.
The message for enterprises stuck in the ITSM spin is ringing clearly: better AI isn’t a matter of picking up the pace as much as it is picking the right places. Focus on the fundamentals first, then go all-in on AI technologies, and you’ll be sure to have both the right steps and the right strides for whatever lies ahead.