Gartner says organisations scaling AI in infrastructure and IT operations should focus on a small number of strategic investments, not try to do everything at once. Its forecasts point to two practical tests: whether agentic tools can shift routine work towards more proactive operations, and whether companies can keep their costs visible as usage grows.
Watch Desk analysis
What happened
In a report published on 26 October, IT Pro describes Gartner’s recommendations and forecasts for infrastructure and IT operations teams. Gartner surveyed 272 leaders in the field; 58% said scaling AI proofs of concept into wider adoption would be a priority in the next 12 months.
Gartner expects half of organisations to be using AI agents with minimal human involvement in IT operations by 2030, although it says virtually none do so now. It also warns that, through 2027, organisations using uncapped, token-based pricing for agents in IT operations could exceed their initial cost expectations by more than 50%. Its proposed response is a “no telemetry, no deployment” policy: make usage visible before deploying.
Key findings
- Agentic operations
Gartner points to incident triage, change validation and remediation as possible uses, with people moving from hands-on oversight to monitoring the work. - Token costs need watching
Gartner forecasts cost overruns above 50% for some uncapped pricing models through 2027, and recommends monitoring usage before deployment. - Hybrid infrastructure is growing
Gartner links AI inference across cloud, data centre and edge to privacy, cost, performance and data-sovereignty needs. - Operations teams may converge
Gartner predicts a quarter of enterprises will have merged network and security operations centres by 2030, up from one in 20 now. - Infrastructure teams have a bigger remit
Gartner says organisations without strong infrastructure and operations support for AI will be more than twice as likely to fail in their AI initiatives over the next four years.
Why it matters
This is useful advice for organisations moving beyond pilots: agent capability is only part of the decision. Pricing terms, telemetry and where inference runs can shape whether a deployment remains affordable and manageable. The report also puts infrastructure teams closer to the centre of AI planning, rather than treating them as the people who arrive after the exciting announcement to ask where it will run.
The figures are Gartner forecasts and survey findings, not guarantees about what every organisation will experience. But the cost warning gives buyers a concrete question to ask before signing up: can they see and control the usage their agents generate?
Our read
The strongest advice here is pleasantly unglamorous: know what your agents are doing, and what they are costing, before giving them more room to do it. Organisations should scrutinise token-pricing terms and require usage monitoring as part of any deployment plan. “Autonomous” is a fine word; an uncapped invoice is still an invoice.
What to watch
- Whether organisations make usage telemetry a condition of deploying operational AI agents.
- How agent pricing models change as companies move from trials to wider adoption.
- Whether Gartner’s forecasts on agent use and operations-centre mergers track with real deployments. IT Pro’s report on Gartner’s forecasts and recommendations
Discussion spark: Should organisations require usage telemetry and predictable pricing before deploying AI agents in IT operations, even if that slows adoption?
Sources and evidence
- Source update (26 October 2026, 10:44 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.