Itil Governance

Most large organisations run their technology departments on a framework you have probably never heard of by name.

It is called ITIL, originally an acronym for IT Infrastructure Library, and it was developed in the 1980s by the British government to standardise how big technology teams should operate. It describes how to handle incidents when something breaks, how to manage changes so new releases do not break other things, how to keep track of every asset the organisation owns, how to respond to the steady flow of small requests that come in from users. It is a sprawling, bureaucratic, slightly dusty thing, and it is also running quietly underneath the systems you use at work every day, from banking to healthcare to education. The framework is not glamorous. It does not promise innovation. What it promises is that things do not fall over at two in the morning on a Tuesday.

Where this gets interesting, and slightly alarming, is in its application to AI. Adaptive learning platforms, predictive models, automated grading systems, the whole new generation of education technology, are being treated as if they were regular IT services. Log an incident. Raise a change request. Follow the approval workflow. The trouble is that AI systems have failure modes the framework never imagined. Model drift, bias emerging over time, hallucination, confidence without correctness. An incident response process designed for a server that has stopped responding is not going to catch a model that keeps giving answers, just increasingly wrong ones. Using ITIL for AI is better than using nothing for AI, but it is not the same as using something that was actually designed for AI. When you think about the rules your workplace follows, how many of them are governing yesterday's problems?

Last week we looked at separating decisions from judgements. This week we look at the unsung framework that runs underneath most large technology departments, and at how badly it fits AI.