UK regulator warns broadband firms not to hide humans behind AI
Key Points
- Ofcom warned providers against AI chatbots that block access to human agents
- Providers must offer at least three complaint routes under General Condition C4
- Ofcom wants chatbots to make clear to customers that they are AI
- Ofcom will not introduce new rules on AI in telecoms for now
Ofcom has warned UK telecoms providers against using AI chatbots to block customers from reaching human staff.
The regulator set out its position in a report on AI in telecoms published on Friday (25 September). The report found that 23% of UK adult internet users had used a customer service AI chatbot in the past year.
“We would be concerned if communications providers’ AI applications were to act as ‘gatekeepers’ that create barriers to accessing support over the channels that providers are required to make available to customers,” Ofcom said.
Under General Condition C4, providers must accept complaints through at least three routes. These are a phone number, a postal address, and an email address or dedicated web page.
Ofcom said AI tools should form only one of the channels open to vulnerable customers. The others include phone, email, post, and video and text relay.
It added that chatbots handling vulnerable customers must follow the policies providers already need to have in place for those customers.
Chatbots should say they are AI
Ofcom said it should be very clear to customers when they are dealing with a generative AI tool. That includes how a chatbot describes itself.
The regulator’s own research found that some people could not tell whether they were talking to AI, a simpler automated system or a real person.
What providers told Ofcom
One unnamed telecoms company told Ofcom it automatically passes customers to a human agent after three failed attempts to resolve a query through its chatbot.
Ofcom said providers appear to take a human-first approach to sensitive cases such as bereavement and debt management.
Vodafone replaced its TOBi chatbot with SuperTOBi in July 2024, using Microsoft generative AI to handle more complex customer queries.
EE added generative AI to its Aimee chatbot in 2024. Virgin Media O2 uses a tool called Lumi AI to support human agents during customer calls.
Virgin Media O2 said Lumi AI has helped it answer customer queries faster and reduce complaints.
No new rules for now
Participants in Ofcom’s research described chatbots as rigid, robotic and frustrating to navigate. Some said they try to bypass chatbots entirely to reach a person.
Consumer groups including Which? asked Ofcom to clarify the rules on multiple contact channels and access to human agents.
Citizens Advice and Which? also called for new rules requiring providers to disclose where they use AI.
Ofcom said it does not need to introduce new rules or guidance at this time. It cited limited evidence that AI has caused systemic harm to telecoms customers.
The regulator will hold a further roundtable on AI in the telecoms customer journey between April and June 2027.
It will also publish further research on trust in AI chatbots in spring 2027.
