Is CTI dead? What AI means for the oldest habit in ITSM

If you worked a service desk in the 2000s, you remember the dropdown ritual. Category, Type, Item. Three fields standing between you and logging a ticket.

Where it came from

Here is a detail most practitioners get wrong: CTI was never ITIL. ITIL v2 and v3 simply required that incidents be categorised, usually through a multi level scheme of category and subcategory, so that trends could be reported and tickets could reach the right resolver group. The rigid three tier Category, Type, Item structure was a tool convention. BMC Remedy popularised CTI, the Peregrine ServiceCenter lineage that became HP Service Manager ran Category, Subcategory and Product Type, and CA Nimsoft Service Desk went further with a four tier Class, Category, Type, Item. Vendors turned a principle into a data model, and the industry mistook the data model for the framework.

The reasons were sound at the time. Managers needed to know what was breaking and how often. Tickets needed to land with the right team. A structured taxonomy was the only way a database could answer either question, because nobody could query free text at scale.

Where it went wrong

Category trees bloated. Organisations built taxonomies with hundreds of items, and agents responded rationally: they picked the first plausible option, or Other, and moved on. Miscategorisation became endemic, the reporting built on top of it became fiction, and every few years someone launched a recategorisation project that solved nothing.

The automation era

Then categorisation got a second job: driving assignment rules, SLAs and workflows. The taxonomy became load bearing. From around 2018, machine learning entered the picture. ServiceNow shipped Agent Intelligence in its Kingston release, later renamed Predictive Intelligence, auto filling the very fields designed for humans. Which quietly raised the real question: if a machine reads the text and picks the category, what is the category actually for?

Where AI leaves it

Modern language models read a ticket, understand intent, sentiment and urgency, and can route it directly. No middle layer required. Routing no longer needs a category tree.

But categorisation is not dead. It has changed jobs.

It is no longer an input burden on agents. It becomes an output artefact, applied by AI, consistently, after the fact. And that still matters: trend analysis and problem management, SLA and compliance mapping, cross system reporting where taxonomies act as a shared data contract, and audit trails, because the model decided is not an answer a regulator accepts.

Our take

The future is fewer, flatter, outcome focused categories that nobody types in. Humans describe the problem in their own words. AI classifies silently in the background. The taxonomy shrinks from hundreds of items to a few dozen that mean something in a boardroom report.

ITIL never asked for three dropdowns. It asked us to understand demand. For thirty years, dropdowns were the best proxy we had. Now we have something better.

Hikon helps organisations across the UK, Europe and Israel modernise their ITSM platforms on ServiceNow and Halo. If your category tree has become a liability, talk to us.