When AI supports the service desk, something unexpected can happen to the key metrics. Mean time to resolution (MTTR) climbs. Ticket volume falls. And both trends look like bad news, until you understand what’s actually changed.
“Those numbers are a sign that AI is working,” says Jason (Jay) Perry, product marketing manager, ITSM at ServiceNow. “AI handles the routine tickets. Humans are left with genuinely harder work that takes longer to resolve. If you’re still measuring with the old metrics, you’re going to misread the whole picture.”
MTTR, ticket volume, and service level agreement (SLA) attainment were built to measure the speed and volume of human effort. They still matter operationally. But with AI in play, they no longer describe how the system works. Resolution time rises because what’s left requires more judgment. Volume falls because AI can autonomously resolve routine requests before they reach the queue. The metrics haven’t changed. The work has.
Organizations anticipating this shift can move to three measures that reflect the new reality, typically starting with a handful of repetitive workflows before expanding more broadly.
- Automation coverage. This new metric can track the share of requests resolved end-to-end by AI without human intervention. ServiceNow’s own internal deployment handles 90% of IT requests this way, allowing them to expand support capacity and free skilled staff for higher-value work.
- Agent accuracy. This can measure whether AI performs at the level of a strong human agent. Organizations can compare AI’s reopen rate, escalation rate, and customer satisfaction scores (CSAT) against the organization’s top human L1 agent. “If AI matches your best agent, it’s a good candidate to scale,” Perry says.
- Prevented incidents. They are the hardest to measure and the easiest to dismiss. They’re the outages that never materialized; P1s that stayed anomalies because the platform caught an early signal and acted before any user felt it. The obvious objection is that you can’t prove what didn’t happen. However, you can measure what the platform acted on or track suppression rates on known failure patterns. That shift in direction is enough to change a budget conversation — if you’re measuring prevention rate at all. Most IT teams aren’t. “IT is measured on resolution speed, which shapes what gets reported, funded, and celebrated,” says Perry. “Changing that is as much a management and culture decision as it is a tooling one.”
CIOs can use these new measures to have a cost-to-serve discussion with their CFO counterparts, delineating how much it costs an IT organization to run its operations using AI and the additional value they are gaining from the technology.
IT organizations winning with AI lead every executive review with these new metrics — automation coverage, agent accuracy, and prevented incidents — to signal a new operational mindset. MTTR and ticket volume are still tracked for the human work but are moved to the appendix, no longer in charge. “You have to make the new numbers the headline,” Perry says. “Everything else follows from that.”
MTTR and ticket volume still have purpose. But with AI powering the service desk, they no longer define what good looks like.
For more information, click here.