%201.png)
Ofcom's review of AI in telecoms, published on 25 September 2026, highlights that just 8% of UK internet users have used an AI service in connection with their telecoms provider, and only 56% of that group trusted the result the last time they tried. For telecoms customers, AI adoption is just starting, but trust has not caught up with it.
Ofcom asked why operators keep their AI tools on a narrow brief such as answering routine queries, escalating anything sensitive to a person, and stopping short of autonomous decisions. Firms gave Ofcom several reasons for that caution, and one of them is not really about AI at all. It is that customer information sits across fragmented systems that were never built to answer a live question.
Three of the Five Trust Factors Are Data Problems
Ofcom's qualitative research with consumers and small businesses identified five things people weigh when they decide whether to trust AI in the telecoms customer journey. These are how their data is being used, whose profit motive is driving the interaction, how much control they are keeping, whether the information they receive is accurate, and how good the underlying technology actually is. Three of those five, data use, accuracy, and quality of response, are decided before a single word reaches the customer. They are decided by what the AI is allowed to see.
Ofcom's own worked example makes the point well. A customer disputes a bill that looks wrong, the chatbot tells them to ignore it, and the higher amount is taken from their account regardless, because the chatbot was answering from a knowledge base rather than the billing system that held the real answer. That is not a language model failing to reason. It is a data problem wearing a chat interface, and it is exactly the kind of incident that erodes the 56% trust figure further.
What Ofcom's Near-Term Scenarios Actually Require
Ofcom sets out four AI-driven developments that it expects to reshape the telecoms customer journey over the next few years. These are:
- More accessible, tailored communication
- Hyper-personalised offers and pricing
- One-stop AI intermediaries that can complete a switch or a sign-up on a customer's behalf
- Agentic AI acting with growing autonomy
If you examine what each of these developments needs to work as expected and build trust, then you see the same need for trusted data as the foundation on which these use cases run.
Tailored communication needs a chatbot that remembers a customer's stated needs across sessions with a 360° view of a customer’s knowledge graph so it answers from live account and network status data, not a static script. Hyper-personalisation needs usage, billing history, and service records combined accurately enough that a customer can still tell whether the price they are offered represents fair value. One-stop intermediaries and agentic AI need account state and consent records that a CSP can trust enough to let an agent act and can produce afterwards to answer Ofcom's own open questions: who is accountable when the agent gets it wrong, what counts as meaningful consent, and how a customer in vulnerable circumstances is protected when the front door is an agent rather than a person.
Every one of those is an evidence question before it is an AI question. A CSP cannot demonstrate transparency, accountability, or informed consent to a regulator, or to a customer, using data it cannot itself trace back to a system of record.
Consolidating BSS Data Is Not Automatic
None of this is a quick fix. Billing, order management, product catalog, and network status data typically live in systems that may have been provided by different vendors at different points in a CSP's history, and getting them to agree on a single version of the truth takes real data harmonisation, not a prompt engineered around the gaps. CSPs that treat this as a wiring exercise, connecting an LLM to whatever API happens to be open, tend to rediscover Ofcom's billing example the hard way.
How Qvantel Flex AI Experience Grounds Agents in Trusted BSS Data
Qvantel Flex AI Experience, part of the wider Qvantel Flex Suite, starts from the BSS layer rather than bolting AI on top of it. It has a data fabric layer which, importantly, contains a telco-specific data management platform that was developed specifically to understand BSS data. It has prebuilt agents for customer self-care, sales management, operations, and product catalog support which can be further extended and connected through open interfaces and MCP-based services directly into live Flex BSS data including billing, order management, product catalog, and contract state, rather than a copy of it that drifts out of sync. On top of this is the AI Foundry layer which contains the AI studio and agent co-creation service that lets CSPs develop their own agents or build them with Qvantel’s AI team.
The key here is that the agents use grounded and trusted BSS data and the platform carries enterprise-grade security and data sovereignty controls, because an agent acting on a contract or a bill carries the same consequences as a person doing it.
Every one of Ofcom's near-term scenarios turns on the same question: can a CSP show, after the fact, what an AI agent knew and why it acted. That answer starts with the data platform underneath the agent, not the model sitting on top of it. Get that right, and the new use cases such as hyper-personalisation, agentic switching, and tailored communication are built on a trusted data foundation that enables AI-driven services that customers will use and trust.
Bernhard Kraft
Head of Product Management, Qvantel
