Industry library
CX Operations
Customer operations is measured by the numbers that are easy to count and judged by the ones that are not. These pieces are about the difference.
Deflection is not resolution, and AI makes the gap easier to hide
Contact center AI is very good at deflection — closing a contact without a human touching it — and deflection rate is an easy number to put in a QBR. It is not the same as resolution, and an AI system optimized to deflect will happily deflect a customer whose actual problem wasn't solved, generating a repeat contact later that shows up as a new case rather than a failure of the first one. CX operations that measure success by deflection rate alone are optimizing for a proxy metric that inherits all the incentive to game it. The programs that hold up track resolution and repeat-contact rate specifically for AI-handled cases, not just for the channel as a whole.
Handle time was always a false contract with quality
Average handle time has been the default contact center metric for decades, and AI voice and chat agents are extremely good at reducing it — sometimes by truncating the interaction before the customer's actual issue surfaced. Handle time was always a proxy for efficiency that assumed a roughly constant quality floor; AI breaks that assumption because an automated agent can hit a great handle-time number while providing a materially worse resolution than a human would have. Full-population quality review, made possible by AI grading every interaction instead of a sampled few, is the corrective — but only if the quality rubric is built around outcome and not around compliance with a script.
The escalation path is where loyalty actually gets decided
Most customer loyalty isn't won in the easy contacts an AI agent handles well; it's won or lost in the escalation path, when something has already gone wrong and the customer is testing whether the company will actually fix it. Voice agent routing that's tuned to minimize transfers can end up trapping frustrated customers in automation longer than they should be, which reads as efficient on a dashboard and reads as stonewalling to the customer. CX operations leaders getting this right treat time-to-human-escalation for genuinely complex or emotionally charged cases as a protected metric, not something to be optimized away by the same logic that correctly reduces friction on routine requests.
Deflection is not a customer outcome
Contact centers report deflection as a win. It measures a contact that did not reach a human — not a problem that got solved. The two diverge in exactly the cases that cost the most.
Read
Average handle time is the wrong contract
AHT was a staffing input that quietly became a performance target. Once agents optimise against it, the operation buys speed with re-contacts and pays for them in a different budget line.
Read
The voice agent is a routing decision
Most voice AI programs are scoped as a channel replacement. The valuable version is narrower: an agent that decides, in the first thirty seconds, what kind of problem this is and who should own it.
Read
QA samples two percent and calls it quality
Traditional contact center quality assurance reviews a statistically meaningless sample, scores it against a rubric written for a different era, and produces a number nobody uses to make a decision.
Read
Escalation is the product
Every CX operation designs the happy path and improvises the escalation path. Customers form their durable opinion of you in the second one.
Read