After years of companies routing more calls, chats and refund requests through AI, the backlash has hardened into measurable numbers: preference for speaking with a real person has climbed from 83% to 85% among consumers surveyed in 2026, while preference for AI assistance actually dropped, from 7% down to just 5%, according to customer-experience research covered by outlets including PYMNTS and CNBC this year.
The Numbers Behind the Frustration
Frustration with AI-driven support agents rose from 54% to 59% over the same period, and the share of customers who say they would simply hang up if routed to an AI system climbed from 29% to 31%. Qualtrics’ 2026 Customer Experience Trends Report found that nearly one in five consumers who tried AI for customer service came away seeing no benefit whatsoever from the interaction, a striking figure for a technology companies have spent billions deploying specifically to improve service economics.
Refunds Became the Breaking Point
CNBC’s reporting on the trend, published under the headline “I hate customer-service chatbots,” zeroed in on refund requests as a particular flashpoint. Refunds require judgment calls — assessing whether a product was actually defective, whether a policy exception is warranted, whether a customer’s account history suggests good faith — and that is precisely the kind of ambiguous, high-stakes decision where chatbots most often fail, either issuing confidently wrong answers or looping customers through denial after denial until they give up.
What’s Actually Breaking
Industry analysts tracking the failures point to a consistent set of culprits: AI hallucinations that produce confidently incorrect information, “bot fatigue” from customers who have already had multiple bad chatbot experiences before this one, an inability to handle industry-specific nuance at the depth customers expect, and a widening satisfaction gap between AI-handled and human-handled interactions. Data misuse has also emerged as a top-line concern in its own right, with 53% of consumers now citing it as their primary worry about AI-run support — a sign that trust issues extend beyond whether the bot gives a correct answer to whether it should have access to their account data at all.
The Business Case Companies Are Still Betting On
None of this has stopped companies from expanding AI deployment in the contact center, because the economics remain compelling on paper: AI agents handle a fixed cost regardless of call volume, work continuously without staffing schedules, and can resolve straightforward, high-volume requests like order tracking or password resets almost instantly. The disconnect is that companies have in many cases pushed AI into complaint and refund workflows — the emotionally charged, judgment-heavy interactions — rather than confining it to the simple, repetitive tasks where it performs best, and that mismatch is a large part of what is driving the reported satisfaction gap.
Small Businesses Feel It Differently
Forbes coverage this year highlighted a distinct wrinkle for small businesses: many adopted off-the-shelf AI chatbot tools to compete with larger rivals’ service capacity, only to find that a poorly tuned bot can do more brand damage with a smaller, more loyal customer base than a slow human response ever would, because those customers have fewer alternatives to switch to and notice the friction more acutely.
What Happens Next
Expect a correction rather than a retreat: companies that read the 2026 data are already recalibrating which interactions get routed to AI versus a human, adding clearer “talk to a person” opt-outs earlier in the flow, and building escalation paths that trigger automatically once a chatbot fails to resolve an issue within a set number of exchanges. The winners in this next phase will likely be firms that treat AI as a triage layer rather than a replacement for judgment calls — using it to handle the repetitive 80% of tickets while routing the messy, emotional, high-stakes 20% to a human faster than before, rather than slower, which is the opposite of what many chatbot deployments currently do.
Photo: Airam Dato-on / PEXELS via Pexels
