Ask a robot to schedule your dentist appointment. It fails. You call the office. A human picks up — and the first thing they ask is your name, your birthday, and the reason for your call. The exact three things you already typed into the robot. That small, familiar humiliation is now the defining experience of American customer service. And the people on the other end of the line are done pretending otherwise.
UJET put a number on the frustration. The company surveyed 250 U.S.-based frontline contact center agents — real people, aged eighteen and up, at mid-market and enterprise companies with 500 to 5,000 employees, working the phones in financial services, retail, healthcare, travel, and technology. These are the industries where the phone still rings constantly, where a bad interaction costs a customer, and where AI was supposed to be the great rescue. The report is called The Agent and AI Disconnect: A Blueprint for Futuristic CX. The title is polite. The findings are not.
Here is the headline that should keep a lot of vendors awake: 100% of agents use AI every single day, and 0% consider it critical to their success. Not "few." Not "a minority." Zero. When you deploy a technology to every worker in a building and not one of them would fight to keep it, you haven't built a tool. You've built furniture.
AI is everywhere. The impact is missing.
Universal adoption is the part the press releases love. It's true, too. Three-quarters of agents — 75% — now lean on AI for more than three-quarters of every customer interaction they handle. Nineteen percent report AI touching more than 90% of their workload. Sixteen percent have hit what UJET calls "total integration": AI in 100% of their conversations, all day, every day.
And yet. The same agents drowning in AI describe it the way you'd describe a co-worker who means well but can't be trusted with anything important. Fifty-four percent say the tools help with some issues but lack the context or depth to actually do the job. Another 21% shrug — neutral, unmoved, unchanged. Put those together and a majority of the workforce has already hit what the report calls a utility ceiling. The software is present. It is simply not powerful.
"93% of agents believe they could do their jobs without AI, yet 100% also say it saves them time. What gives?"
That's the riddle at the center of the whole document, and UJET has a sharp theory about it. Agents and supervisors care about two different things. Agents are laser-focused on one question: can I solve this person's problem? They know — from thousands of calls — that empathy, judgment and a calm voice don't require a machine. Supervisors care about speed, quality and efficiency at scale. Both are right. The tension between them is the disconnect, and right now it's being resolved on the backs of the people wearing the headsets.
The dependency that never showed up.
Every serious technology creates dependence. Take away a surgeon's imaging, a trader's terminal, a writer's keyboard, and the work grinds to a halt. That's how you know a tool matters. By that test, contact center AI fails spectacularly. Ninety-three percent of agents are confident they could perform their jobs without major issues if their AI tools vanished tomorrow. Only 8% believe they'd genuinely struggle. And seventy-eight percent — more than three in four — flatly state that the AI in place today, from agent assist to automated coaching, is not transformative.
This is the quiet scandal of the enterprise AI boom. Companies are writing enormous checks for tools their own frontline would happily unplug. UJET's own AI Maturity Benchmark found the reason: the money is going to the shallow end. The top investments are simple self-service chatbots (47%), performance and call analytics (39%), and inquiry routing (38%). By UJET's maturity scale, 55% of those investments rank as low maturity — automation that answers FAQs and shuffles calls, but never actually collaborates with the human trying to help.
Where the AI money goes (top investments)
Where AI quietly earns its keep.
To be fair — and the report is fair — AI is not useless. It's just modest. Sixty-nine percent of agents credit it with killing the drudgery of after-call work: the note-taking, the data verification, the copy-paste tedium that eats the minutes between calls. That efficiency shows up on the clock. Fifty-four percent say routine interactions run up to 25% faster; another 37% report an even bigger 26-to-50% boost. And here's the one number everyone agrees on: 100% of agents say AI saves them time.
The question is what they do with the time back. Two-thirds (66%) simply handle more requests. One in five (20%) move up to higher-value work — upselling, retention, cracking the genuinely hard problems. Fourteen percent use it to breathe. And the payoff is real and human: agents tie those savings to promotions (78%), more money (24%), and better performance reviews (18%). Freed from manual entry, they say their soft skills sharpened — empathy and active listening, multitasking, troubleshooting, creativity under pressure. Give a person back their attention and they'll spend it on the part of the job a machine can't do.
Customers and agents finally agree.
Now the ugly part. When AI is done badly — and it usually is — it doesn't just fail neutrally. It actively poisons the relationship. Sixty-five percent of agents say customers express frustration on most or every call about having to repeat information they already handed to a chatbot before reaching a human. The continuity breaks, and when it breaks, the temperature rises. Fourteen percent of agents say customers arrive more emotionally charged after a failed AI self-service attempt. Thirteen percent say that heat lands on them as extra stress and burnout.
Then there's trust — the thing that, once gone, is hardest to rebuild. Ninety-three percent of agents feel compelled to double-check or verify what their AI tells them before they'd dare repeat it to a customer. Fifteen percent call the real-time AI recommendations outright unreliable or inaccurate. Think about the absurdity: a tool built to speed you up that you cannot use without slowing down to fact-check it.
"We have reached a point where bad CX is more damaging than no automation at all."
The root cause isn't the AI itself. It's the plumbing. Rather than build native, integrated systems, organizations bolt shiny AI features onto aging, legacy, on-premise infrastructure and siloed data. The result is a cluttered desktop and an agent playing air-traffic controller across a dozen windows. Eighty-one percent of agents run more than four tools at once. Nineteen percent are wrangling more than seven. When companies chase an AI strategy without the data architecture underneath it, the report says plainly, they outsource the frustration to their customers and the cleanup to their agents.
The desktop sprawl — a day in tabs
Four is the floor. Seven-plus is a fifth of the workforce. Every extra tab is a place for context to fall through.
Fear, and something more complicated than fear.
All of this bleeds into the one question every agent actually loses sleep over: will the robot take my job? Seventy-one percent say they're likely or somewhat likely to start looking for new work because they're worried AI will replace them. That's a retention crisis hiding inside a productivity story. But the deeper you read, the more human it gets. When agents talk to each other — not to a survey, to a peer — the mood is layered. Fifty-four percent are cautiously optimistic that AI will change their roles rather than erase them. Thirteen percent are outright excited, seeing new opportunities and easier days. Twenty-nine percent hold both hope and dread at once.
The reason for the split is almost poetic: agents are currently the testing ground for immature technology. Companies commit serious capital; the immediate ROI doesn't arrive; and the workforce is left in a kind of professional limbo, doing the emotional labor of a transition nobody has finished designing. As the report puts it, companies tethered to legacy mindsets will drive their agents toward the exit — while those that treat AI as a genuine co-pilot are positioning agents to thrive.
The closing line of the survey reframes the whole debate, and it's worth sitting with: the future of CX will not be a choice between humans or machines; it will be defined by how effectively an organization integrates and seamlessly escalates customer interactions between the two. The winners won't be the companies with the most AI. They'll be the ones that know exactly when to hand the call to a person — and do it without making the customer start over.