On Sept. 21, 2026, The Guardian reported that Citizens Advice was urging UK banks, energy firms, phone companies and internet providers to stop making AI chatbots the only path to customer support and to guarantee access to a human. The charity said its research found chatbots wasted time, raised stress and delayed problem-solving for more than half of users.
The warning matters because an unresolved problem with an essential service can have consequences beyond a poor app experience. Our operational reading is that a support interface should be evaluated by whether it helps a person resolve the underlying issue, including when the automated route cannot do so. Counting a conversation as handled because the chat ended would not demonstrate that a billing dispute, interrupted service or inaccessible account was actually fixed.
What providers could test
Consider a hypothetical customer who disputes a charge and cannot supply the account detail requested by a bot. A useful acceptance test would follow that person through the entire journey: whether a human route is visible, whether it works without the same unavailable information, and whether the next person receives the earlier explanation. The goal is to test a reachable fallback, rather than assume that adding a contact link solves the problem.
A second test could examine an interruption midway through the conversation. If the user loses a connection, changes device or returns later, can they recover the case reference and understand what remains unresolved? Product teams could compare the experience with a phone or assisted route and document the differences. These are proposed checks, not results from testing any named provider.
How the mechanism works
Citizens Advice’s earlier response to the Mills Review shows that its concern is broader than a bad chatbot script. In its July 6, 2026 submission, the charity said more consumers were using large language models for financial advice, but that the advice could be inaccurate, lack nuance or include fabricated examples. It also warned that consumers may not understand the degree of control they hand to agentic AI, and that AI adoption is likely to widen the digital divide.
That is the practical mechanism behind the customer-service complaint: once a provider shifts routine support to chatbots and removes phone, face-to-face or paper routes, automation only works if customers can self-serve successfully. If they cannot, the burden moves onto the user, who may have to repeat information, navigate authentication steps or wait for an offline workaround. Citizens Advice said that burden falls hardest on people already excluded by cost, device access or confidence with digital systems.
Who should care
Banks, utilities, telecoms and broadband providers are the clearest audience, but the decision also matters for product teams designing support flows for regulated services. Citizens Advice said people without a human route to resolve essential-service problems often felt powerless or frustrated, and some gave up entirely. For a provider, that is not just a satisfaction problem; it can become a complaints, arrears or consumer-harm problem.
Teams could also separate speed from resolution in their own measurement. A short conversation may be efficient, but it may also end because the user cannot continue. Track a clearly defined resolution outcome, repeat contacts and successful escalation, with privacy controls appropriate to the service. Review the difficult cases individually before interpreting a lower contact volume as an improvement. Without that distinction, a dashboard can reward a barrier that customers experience as failure.
Decision boundary
The evidence here does not show that every chatbot deployment fails, and it does not measure every provider. But it does support a narrow operational conclusion: chatbots can be useful as triage, yet essential services should not let them become the only route to fix urgent problems or reach redress.
Check whether your customer service still offers a phone, email or face-to-face route, because Citizens Advice says chatbot-only systems can delay urgent fixes and exclude people who cannot reliably use digital channels.
