The State of Customer Experience 2026 Deep Dive #1: The Human Preference Paradox

The headline from the State of Customer Experience 2026 report is striking: customer preference for speaking to a human agent is up 5% year on year. Among 18-34-year-olds — the most digitally native consumers — that figure jumps to 18%.

At first glance, that looks like bad news for AI vs. human customer service. A step backwards. Evidence, perhaps, that the push toward automation has gone too far.

But spend more time with the data and a different story emerges. When customers who prefer human agents were asked whether they’d switch to automated self—service if it could fully resolve their issue, 69% said yes. Among that younger cohort, the figure rises to 93%.

So customers want AI that works. The problem is that, too often, it doesn’t.

This is the human preference paradox: rising demand for human agents isn’t a rejection of AI. It’s a verdict on how AI is all too often deployed.

Why Customers Are Shifting Back to Human Agents

To understand what’s driving the shift, you need to look past the headline number and ask a more precise question: what are customers actually reacting to?

Customers are clear on what they value most in a service interaction. Speed tops the list — getting their issue resolved quickly is the number one priority. Second is the ability to resolve the issue without needing to speak to a human agent at all. But third is easy access to a human agent when they need one.

Read that again. Customers want self-service and human access. Not one or the other.

What they don’t want is an automated service that keeps them stuck in a loop, fails to complete their request, and then makes it difficult to reach a human when the interaction breaks down. When that happens, customers recalibrate their preferences. They default back to what they know will work: a person.

The rise in human agent preference, then, is less about a shift in values and more about a loss of confidence. Customers haven’t stopped believing in AI’s potential — in fact, 85% of consumers say it benefits CX. But they have tempered their trust that the AI they encounter will actually deliver.

The generational signal worth paying attention to

The sharpest rise in human agent preference is among 18-34-year-olds — up 18% year on year. This is the demographic that grew up with smartphones, expects digital-first experiences, and has the highest tolerance for automation. They’re also the most likely to switch to AI if it resolves their issue (93% say they would).

That combination is significant. This group is the first to embrace AI when it works. The fact that they’re increasingly choosing human agents tells you something specific: the AI they’re encountering isn’t meeting their expectations.

Last year’s deep dive explored how this generation was leading the shift toward digital and automated CX. The 2026 data adds an important caveat. Younger consumers will follow AI wherever it takes them, but only if it actually gets them where they need to go.

What “resolves my issue” really means

The phrase “fully resolve their issue” is doing a lot of work in this data, and it’s worth unpacking.

Customers aren’t asking for AI that answers questions. They’re asking for AI that takes action. There’s a meaningful difference between a virtual assistant that tells a customer their account balance and one that processes a refund, updates an address, rebooks a flight, or escalates a fraud case — all without handing off to a human.

That second category requires agentic AI: systems capable of understanding complex, multi-step requests, integrating with backend systems, and completing tasks end-to-end. It’s a significantly higher bar than most automated self-service tools are currently clearing.

This is precisely why virtual assistants that handle simple, FAQ-style queries are no longer enough. Customers have calibrated their expectations against what they experience every day — AI that searches the web, drafts documents, and makes recommendations in seconds. When the chatbot on a company’s website can’t process a simple return without transferring them to an agent, the contrast is jarring. The expectation gap widens. And preference shifts back to human.

The AI agent handoff problem

There’s a second layer to this that the data surfaces clearly. When customers were asked to rank their most important factors in a good customer experience, fourth on the list was ‘not having to repeat information when transferred to a human agent.’

This is the handoff problem. And it’s costing businesses more than they realize.

When a customer spends several minutes in an automated interaction — explaining their issue, verifying their identity, navigating menus — and then gets transferred to a human agent who has none of that context, the experience falls down. It signals that the business’s systems don’t communicate with each other, which in turn signals a lack of investment in the customer’s time.

That moment — the reset, the “can you tell me your account number again?” — is the moment a customer decides whether they’ll come back. And increasingly, they won’t. The data is unambiguous: 79% of consumers say they’d switch to a competitor after a single terrible experience.

How to deploy agentic AI customer service that earns trust

The instinct, when faced with data showing a rise in human agent preference, might be to slow down AI deployment. Pump the brakes. Give customers what they’re asking for.

That’s the wrong read.

The customers who prefer human agents aren’t asking you to deprioritize AI. They’re asking you to build AI that earns their trust. The 69% who say they’d switch to automated service if it resolved their issue represent a massive, recoverable opportunity — but only if the AI experience actually delivers.

There are three things that follow from this data.

1. AI self-service needs to complete tasks, not just answer questions. Virtual assistants need agentic capabilities — the ability to take action in backend systems, handle multi-step processes, and genuinely resolve interactions end-to-end. Anything short of that pushes customers back toward human agents unnecessarily.
2. Human agents should never start from zero. Every handoff from automated to human service needs to carry full context. Agents who can pick up where the virtual assistant left off — with the customer’s history, intent, and prior steps in front of them — turn a potential point of failure into a point of confidence.
3. The goal isn’t to minimize human contact. It’s to make every contact — human or automated — fast, effective, and frictionless. Customers who reach a human agent should get the same speed and quality of resolution as those who self-serve. That means equipping agents with AI tools — real-time knowledge, interaction summaries, coaching — so the moment they pick up an interaction, they’re already ahead of it.

The bigger picture

The human preference paradox points to something important about where customer expectations are heading.

Customers are drawing a line between service that works and service that doesn’t. Right now, enough AI interactions are falling into the second category that human agents are gaining ground — not because humans are always better suited to handling every type of interaction, but, because in the minds of consumers, they’re more reliably effective.

As AI capabilities advance and businesses deploy them with greater precision, the balance will shift again. But the businesses that get ahead of this moment — that build AI capable of genuine end-to-end resolution, and back it up with human agents empowered to deliver when automation reaches its limits, won’t have to wait for that shift to happen to them.

They’ll be the ones driving it.