For the past two years, customer experience leaders have been told the same story.
AI would remove friction.
Automation would eliminate human error.
Personalization would finally scale.
By 2025, CX was positioned as the year everything “leveled up.” Predictive AI would anticipate needs. Zero-click journeys would eliminate effort. Real-time dashboards would tell us everything. And humans, in many cases, would quietly fade into the background.
At Glance, we wanted to understand whether that vision actually translated into better experiences for customers. So instead of relying on assumptions, benchmarks or vendor promises, we went straight to the source.
We surveyed more than 600 U.S. consumers to understand how AI-driven customer experience is really performing today. What we heard was clear, consistent and impossible to ignore.
The AI story CX leaders were sold
The narrative of 2024 and 2025 was compelling. AI would remove friction, eliminate mistakes and scale personalization across every interaction. Bots would handle volume. Agents would step in only when needed. CX would become faster, cheaper and more efficient.
The market reinforced that optimism. The AI customer service market is projected to reach nearly $48 billion by 2030, and AI-powered interactions were predicted to make up the vast majority of customer interactions by 2025. On paper, the business case felt undeniable.
Many organizations moved fast, trusting automation to do the heavy lifting.
But our survey shows that speed and scale didn’t translate into better experiences for customers.
What customers actually experienced
When we asked customers how AI-powered CX is working today, the results were sobering.
Thirty-four percent of consumers said AI customer support makes things harder. Another 32% said it made no noticeable difference at all. That means nearly two-thirds of customers are seeing little to no improvement from AI-driven support experiences.
The impact on loyalty was even more concerning. Nearly 90% of customers told us they’re unlikely to remain loyal to a company that removes human support. Forty-seven percent said they would not stay with the company at all if human support disappeared. Only 14% said they would remain confidently loyal.
Speed, one of the biggest promises of AI, consistently fell short of expectations. Seventy-five percent of customers said they’ve received a fast response that still left them frustrated. Quick replies didn’t compensate for unresolved issues, repeated steps or being pushed through self-service paths that couldn’t finish the job.
Context loss amplified the frustration. When customers switched between chat, phone and email, only 7% said they rarely or never had to repeat themselves. Forty-four percent said they almost always have to start over, and another 49% said it happens sometimes. For most customers, omnichannel still feels fragmented, not seamless.
Self-service told a similar story. While nearly all customers attempt self-service at least some of the time, only about a third of issues are actually resolved that way. Customers told us they abandon self-service when it loops, can’t handle real complexity or blocks access to a human entirely.
Across open-ended responses, the same themes kept appearing. Customers felt rushed. Trapped. Dismissed. Automation didn’t feel like help. It felt like a barrier.
The trust gap automation created
What surprised us most wasn’t just frustration. It was how deeply automation impacted trust.
When asked what builds confidence, 85% of customers said a human who makes a mistake and fixes it builds more trust than an AI system that delivers a quick, correct answer. Accuracy and speed alone weren’t enough. Customers valued accountability, empathy and reassurance more than perfection.
This exposes a critical flaw in how many AI systems were deployed. Too often, they were designed to deflect rather than resolve. Escalation was treated as a cost to avoid instead of a moment to support. Customers noticed immediately.
Even when metrics improved on paper, trust quietly eroded underneath.
What customers say they want next
Just as important as what went wrong is what customers clearly told us they want going forward.
When we asked what matters most in a support interaction, speed ranked last. Sixty-eight percent of customers said getting a complete resolution is the top priority. Eighteen percent said feeling understood or listened to matters most. Only 15% chose speed of response.
Customers overwhelmingly want humans in the loop. Seventy-five percent said they prefer human-first support. Only 7% prefer AI-first, and 18% said the channel doesn’t matter as long as it actually works.
Customers aren’t rejecting AI altogether. They’re asking for it to play a different role. They want AI to prepare, detect and guide, while humans handle nuance, judgment and reassurance. They want clear escalation paths, seamless handoffs and full context carried across every interaction so they never have to repeat themselves again.
Personalization also needs a reset. Thirty-seven percent of customers said they’re not comfortable with companies using their personal data to personalize support. The message is clear. Customers want personalization rooted in their current intent and journey, not experiences that feel invasive or overly monitored.
The CX reset ahead
The takeaway from Glance’s survey is not that AI failed. It’s that AI was asked to do the wrong job.
AI is exceptionally good at detection, prediction and preparation. But customers don’t want it to replace human connection. They want it to make human help easier to reach, better informed and more effective when it matters most.
The next era of customer experience won’t be won by the companies with the most automation. It will belong to the organizations that prioritize resolution over deflection, trust over speed and long-term relationships over short-term efficiency metrics.
Customers have already told us what they want.
The bounce-back starts with listening.
