What Percentage of Tickets Can AI Resolve?
AI can realistically resolve 80 to 90 percent of repetitive, knowledge-answerable support tickets, and a much smaller share of complex, judgment-heavy, or account-specific ones. The achievable percentage is set by your ticket mix and the completeness of your knowledge base, not by the model alone. Just as important, the safe ceiling is set by the hardest ticket you allow the AI to close, not by the average difficulty of your queue.
It depends on your ticket mix
A team whose volume is dominated by order status, password resets, and how-to questions can honestly automate the large majority. A team fielding billing disputes, fraud, and integration bugs will automate far less, because those tickets require judgment, entitlement checks, or actions that should never be automated. The right question is not "what can AI resolve," it is "what share of my specific tickets are safely automatable."
The safe ceiling, not the maximum
Klarna is the cautionary example: it automated aggressively, then rehired after quality dropped on disputes and escalations. Pushing the automation percentage past what the AI can safely handle produces wrong answers and reopens that cost more than the labor saved. The goal is the highest rate at which reopen rate and CSAT stay healthy, which is often below the theoretical maximum.
How Resolve finds the right number
Resolve targets 80 to 90 percent on repetitive workloads and lets you set the safe ceiling with a tunable confidence dial and a forbidden-action list. Anything below the confidence bar or on the forbidden list becomes a decision card for a human. The self-improving knowledge base raises the ceiling over time as approved resolutions become groundable content, so the automatable share grows without sacrificing safety.