AT&T's Andi Chatbot Traps Customers in AI Support 'Death Loop'

What happened

AT&T is facing mounting criticism over its AI-powered customer service assistant, Andi, which users report traps them in a repetitive support loop with no clear route to a human agent. According to reporting from Wccftech, customers describe the chatbot repeating scripted responses and declining requests to escalate the conversation, leaving people stuck in what has been dubbed a "death loop."

The complaints centre on Andi's apparent design priority: keeping interactions contained within the automated system rather than routing unresolved issues to live representatives. Multiple affected customers say the bot cycles through the same suggestions regardless of how the query is rephrased, with no obvious command or menu option to bypass it.

Why it matters

This is a textbook case of automation optimised for cost containment rather than resolution — and it illustrates a core service-design failure mode: removing the customer's sense of control. Behavioral economics is clear that perceived agency matters as much as outcome; a customer who cannot find an exit path experiences disproportionate frustration even if the underlying issue is minor. That frustration compounds quickly into public complaint and reputational damage, which is exactly what is unfolding here.

For CX leaders, the episode is a warning about deploying conversational AI without a visible, reliable escalation mechanism. Efficiency gains from automation are quickly erased if customers feel imprisoned by the system meant to help them.

The Renascence take

The backlash isn't really about the AI's competence — it's about the absence of a dignified exit. Most operators focus on making bots smarter; few focus on making the handoff to a human effortless and visible.

The real design flaw here isn't Andi's answers — it's the missing "off-ramp." Any automated journey needs a clear, always-available path to a human, because the moment customers feel cornered, trust collapses faster than any efficiency metric can justify. Brands deploying AI-first service should treat escalation visibility as a non-negotiable design requirement, not an edge case handled later. If a customer can't find the door, they'll assume there isn't one — and they'll say so publicly.