Ofcom’s Technology Tracker found that 53 percent of adult UK internet users had used a generative-AI tool or an AI customer-service chatbot in the past 12 months. Eight percent of online adults used an AI tool on a telecoms service. Six percent used one to resolve an issue. Three percent used one to compare price, product, or service. Two percent used one to make a complaint. That is about one in twelve online adults, not a majority of the market.
Network automation is old. The shift Ofcom is tracking is the last five years, when retail providers started putting generative models on customer interactions. Cristina Luna-Esteban, Ofcom’s policy director, said AI could make comparison, contract negotiation, and support “easier, quicker and more personalised,” and that firms still have to keep human support available.
People want the bot until the problem is theirs
Participants split. One said they did not want to speak to a person and only wanted confirmation that someone was dealing with it. Another called an AI-only path “my idea of heaven.” Others said a model has no empathy, that some problems need a person, and that a scripted bot walks them through questions and returns them to the start.
Ofcom’s read is that speaking to a person still matters for complex or sensitive issues, and that some customers already prefer the bot for routine status checks. Providers, the review says, are mostly limiting what the tools can do. Ofcom still expects the share of customers using AI to navigate the market to grow quickly.
Agents that switch the contract are a different product
The review looks past chatbots to agents that could pick a broadband deal and prepare the switch for approval, or watch an account and move the customer when a better tariff appears. Some participants liked the time saving. Most worried about personal data and said negotiating with a person still gets the better price.
Ofcom lists what would have to exist before that product is safe to run: verification and authorisation that actually bind the agent, plus regulatory, physical, and commercial constraints. An agent with login rights to a telecoms account is a credentialed actor. A wrong switch is a billing dispute and a complaint, not a chatbot hallucination.
Two expectations, not a new code
Ofcom says it has limited evidence that AI tools have been a systemic cause of harm to telecoms customers, and that current rules still cover the sector. It then stated two expectations.
Customers should be able to tell they are talking to a generative-AI application. A chatbot augmented by a model should say so in how it describes itself, not only in a footer.
Ofcom would be concerned if a provider’s AI became a gatekeeper that blocks the support channels the provider is already required to offer. The bot can be the front door. It cannot be the only door.
The regulator said it will keep talking to industry and consumer groups and will act if harm shows up. For a provider, the near-term work is labeling and an escalation path that a person can reach without restarting the script. For anyone building an agent that compares or switches UK telecoms, the open items are authorisation, whose data the agent holds, and whether the customer can still get a human when the deal is wrong.