HCLTech: AI agents handled 40% more IT incidents
Two bottlenecks in support
In a case study published on November 2, HCLTech describes a global electronics distributor whose manual triage and reliance on specialists delayed the diagnosis of incidents. The vendor does not identify the client or say what period was evaluated.
One agent classifies and routes incidents according to the team's escalation rules. The other suggests solutions based on previous records, the known-error database and knowledge articles. Both were integrated into the support system the company already used, according to HCLTech.
Numbers released by the vendor
HCLTech says the team was able to handle 40% more incidents without adding headcount. The vendor also reports 35% higher productivity among support engineers and a 28% reduction in mean time to resolution.
These numbers describe a single client and come from the company that sold the solution. The case study does not report the period measured, the baseline, the sample size or how each indicator was calculated. The client is not identified either. So the data doesn't allow estimating the result for another operation or attributing the whole change to the agents.
What's worth testing in the operation
The point to watch is in the flow: shortening the interval between opening and routing the ticket, and making validated knowledge available to whoever solves the problem. The source does not report the rate of incorrect recommendations, the volume of reopened tickets or whether there was human review; these points need to be checked before adopting the model.
A company that wants to test automation can start with a recurring, low-impact category. Before the pilot, it should record volume, severity, time to first response, time to resolution, compliance with agreed service levels, reopenings and escalations. During the test, it's also worth tracking the quality of the recommendations and when the agent needs to hand the case over to a person.
Compare the indicators with the baseline. If the queue drops without errors or rework going up, the team will have grounds to decide whether to expand the test. If resolution time improves while reopenings or inadequate answers grow, the result is not an operational gain.
Sources
HCLTech, 2 Nov 2026. https://www.hcltech.com/case-study/agentic-ai-it-incident-management-electronics-distributor