A company-built AI agent for ServiceNow is being used to create catalogue items, analyse incident trends and review software licences. Its developers say the pilot suggests that tightly scoped workflows can deliver more practical value than a broad, do-everything IT agent.
Watch Desk analysis
What happened
The Digital Worker, built by Bytemethod.ai for its own IT team, connects to ServiceNow through APIs and uses a custom agent harness combining language models with deterministic code. The team says it can generate a catalogue item, including its workflow and scripts, in about 20 seconds from a requirements document. People review and approve write operations before they run.
The team also used it to analyse incident and request data, find recurring issues and draft knowledge articles, and to examine whether users were making use of premium software licences. Richard Mendis and Brian King, Bytemethod.ai executives, describe the work in a VentureBeat guest article.
Why it matters
The account puts useful detail behind the much-used promise that AI agents will do more than answer questions. Here, the proposed payoff is specific: less time building forms, trawling dashboards and reconciling licence use. The authors say trend analysis that took hours can now fit into a 10-to-15-minute conversation.
They estimate that automating future catalogue development could cut costs by 80% and free a quarter of their team’s capacity. Those are the team’s projections, not independently established results; the article describes a pilot with a four-person ITSM team, not a controlled comparison across a broad set of organisations.
Our read
The most persuasive lesson is also the least glamorous: start with a repetitive job that is easy to measure and reverse. The agent’s human approval step and use of existing permissions matter as much as its speed. If you are considering an enterprise agent, the workflow and the guardrails deserve at least as much attention as the model badge.
What to watch
- Whether the catalogue-building savings are measured across more than the pilot team.
- How often people need to correct the generated forms, workflows and scripts.
- Whether the incident analysis and licence review lead to documented, repeatable savings.
Discussion spark: For an enterprise AI agent, should the first success measure be time saved, or how reliably it completes work without a human having to repair the result?
Sources and evidence
- We built an AI agent for ServiceNow. The real pain points were narrower than our roadmap assumed – VentureBeat (4 October 2026, 15:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.