The Two-Day Rule: How One Company Cut Resolution Time by Deflecting 38,000 Tickets

Here's a number I keep coming back to: a British engineering and technology company used autonomous workflows to deflect 38,000 tickets and knocked two full days off its average resolution time. Two days. Not two hours. And the lever was deflection, the most misunderstood word in the support world.

Let me clear something up, because people get this wrong constantly. Deflection is not ignoring. Deflection done badly is a wall of unhelpful articles that infuriates users until they rage-open a ticket anyway. Deflection done well is resolving the request so completely, so early, that a ticket never needs to exist, and the user feels helped, not stonewalled. The difference between those two outcomes is the entire ballgame.

So how do 38,000 deflected tickets translate into a two-day improvement in resolution time? This is the counterintuitive part. The deflected tickets were never the slow ones, they were the fast, high-volume ones. But every one of those quick tickets was sitting in the same queue as the genuinely hard problems, clogging it, adding wait time, pushing the complex cases further back. Pull 38,000 simple requests out of the line entirely and the hard tickets suddenly get attention days sooner. You didn't speed up the difficult work. You stopped burying it.

That's the insight most teams miss: deflection's biggest payoff isn't the deflected tickets. It's what happens to the tickets that remain.

But, and this matters, deflection is also where AI does the most damage when it's wrong. A deflection agent that confidently gives a bad answer doesn't just fail to help; it actively erodes trust, and a user who's been burned once will route around your AI forever. So the maturity curve looks like this: start with the requests where the correct resolution is unambiguous and well-documented. Earn trust. Watch the deflection-accuracy numbers like a hawk. Expand the scope only as fast as the accuracy holds.

Get it right and you don't just clear 38,000 tickets. You give every hard problem in your queue two days of its life back. That's the rule worth chasing.