Resolved — but they wanted a human

The selling number for AI support is impressive and everywhere: agents that resolve 76% of tickets end to end, no human needed. It's the stat in every pitch deck. And taken alone, it's quietly misleading, because the people on the other end of those chats are reporting something the resolution rate doesn't capture.

Through 2026, the share of customers who'd rather talk to a real person rose from 83% to 85%, while frustration with AI agents climbed from 54% to 59%. A growing wave of consumers actively dislikes being routed to a bot, and many companies are quietly putting humans back. "Resolved by the AI" and "the customer left happy" turn out to be two different measurements — and most teams only track the first. Let me pull them apart.

The resolution rate hides the resentment

A 76% resolution rate sounds like 76% of customers were helped. It isn't quite that. It means the ticket got closed without a human — which counts a lot of interactions where the customer got an answer but hated the experience, or gave up and accepted a worse outcome because reaching a person felt impossible. The number measures deflection from your support team. It does not measure satisfaction.

That gap is where the backlash lives. More than half of people will abandon even a resolved AI-only interaction if escalation to a human feels blocked. So a chat can land in your "resolved" column and still cost you the customer. If the only thing you watch is resolution rate, you're optimizing a number that can go up while loyalty goes down, and you won't see it until the churn shows up somewhere else entirely.

The problem usually isn't the AI — it's the trap

Here's the nuance that matters: customers aren't rejecting AI itself. They're rejecting AI that loops, forgets what they just said, and won't let them out. The frustration is rarely the model giving a wrong answer; it's the experience built around it — no memory, no context, no visible way to reach a person. People will happily use a bot that solves their problem fast. What they hate is being trapped in one that can't, with the exit hidden.

Which means the backlash is mostly self-inflicted. Teams chasing a higher resolution rate often do it by making the human harder to reach — hiding the "talk to an agent" option, looping customers back to the bot — and that's exactly the move that turns a useful tool into a hostage situation. The metric incentivizes the behavior that wrecks the relationship.

Measure the escape hatch, not just the deflection

If you run AI support, add the measurements the resolution rate leaves out:

• How easy is it to reach a human? Time it. Count the clicks. If a frustrated customer can't get to a person in one obvious step, you're building the trap, not the tool.
• Track satisfaction on AI-resolved chats, separately. A ticket closed by the bot with an angry customer is a loss wearing a win's clothing. Survey it.
• Watch the escalation rate as a feature, not a failure. People escalating isn't the system breaking — it's the system working. A blocked escalation is the real breakage.

The goal isn't to push resolution to 100%. It's to solve what's easily solvable and get everyone else to a human fast, without a fight.

The bottom line

AI support genuinely works for a large share of tickets, and the resolution rate is a real achievement. But it's half the picture, and the missing half is the half customers actually feel.

A bot resolving the ticket and a customer leaving satisfied are different things, and only one of them shows up in the number everyone quotes. Make the human easy to reach, measure the experience and not just the deflection, and you'll get the efficiency without the backlash that's quietly building across everyone who didn't.