Centrica, the parent company of British Gas, is eliminating 1,300 call centre positions over the next two years. Of those, 800 roles will vanish through what the company calls a “targeted deployment of AI tools,” while the remaining 500 cuts were announced last month. The job losses will affect customer service teams in Glasgow, Edinburgh, Cardiff, Leicester, Stockport, and Leeds. Some jobs will simply disappear when employees leave and are not replaced; the rest will be redundancies.
Trade unions have been quick to warn that Centrica’s investment in artificial intelligence is handing hundreds of human jobs to chatbots. But Chris O’Shea, Centrica’s chief executive, insists that AI is not directly driving the cuts. Instead, he points to changing customer behaviour. According to O’Shea, more than 90% of British Gas customers now use digital channels first when they need help, and the total number of calls has dropped by 20%.
Do customers actually prefer AI support?
The company’s reasoning raises a critical question: does using an app or website first mean customers actually prefer talking to a chatbot? The evidence suggests otherwise. A February 2025 SurveyMonkey report found that 79% of Americans strongly preferred dealing with a human support agent, while only 8% actively preferred AI. Moreover, 81% of respondents believed that companies introduce AI primarily to save money, not to improve service.
Another survey, conducted by Pegasystems and YouGov among more than 4,700 consumers in North America and the UK, reported that 66% prefer human‑led support. Nearly half (46%) said AI rarely or never solved their problem, and only 2% wanted to interact exclusively with chatbots. Although neither survey specifically sampled British Gas customers, the consistency of the results across different geographies suggests that Centrica may be conflating a preference for digital tools with a genuine preference for AI.
The distinction matters. Digital channels include self‑service websites, mobile apps, email, and live chat, which may be operated by humans or AI. A customer who starts with a mobile app to check a bill is not necessarily endorsing chatbot interaction; they may simply want the convenience of 24/7 access to basic information. When the issue becomes complex or sensitive, the same customers overwhelmingly demand a human voice.
When does AI customer service work better?
AI can be effective when the problem is simple and the customer wants a fast answer. Research published last year cited an Intercom survey in which 61% of consumers preferred AI for quick replies to routine questions, such as resetting a password or tracking a shipment. For these straightforward tasks, AI chatbots can resolve requests instantly, without hold times.
Companies also have a strong financial incentive to use the technology. One industry report from early 2026 estimated that AI can cut the average cost of a customer interaction from $4.60 to $1.45. The same report noted that AI agents can solve between 76% and 92% of basic e‑commerce requests. For a utility giant like British Gas, the potential savings are enormous. With millions of customers, even a small reduction in per‑call cost translates into millions of pounds.
Yet cost savings are not the same as customer satisfaction. A serious billing dispute or a problem with a heating system during a winter cold snap is a very different scenario from checking a bill online. When the problem is complex, emotionally charged, or requires human empathy, the available evidence suggests that most people still want to speak to a person.
The human cost of AI
The job cuts at Centrica are part of a broader trend. Across the retail, banking, and telecommunications sectors, companies are accelerating their use of AI to handle customer inquiries. Call centre roles, which have long been a source of entry‑level employment in many cities, are among the most exposed. Unions argue that the shift to AI not only eliminates jobs but also reduces the quality of service for vulnerable customers who may lack digital skills or prefer personal interaction.
Centrica has stated that affected employees will be offered retraining and redeployment opportunities where possible. But for the 1,300 workers in Glasgow, Edinburgh, Cardiff, Leicester, Stockport, and Leeds, the next two years will bring significant uncertainty. Many of these call centres are located in areas where alternative employment in the same sector is scarce.
What the numbers really show
Looking at the data more closely, the claim that “humans actually prefer talking to an AI than a support person” appears to be a misreading of customer behaviour. The 90% digital‑first statistic from Centrica likely reflects the convenience of self‑service tools for routine tasks, not a preference for chatbot interactions. When service becomes complicated, the same customers who started with the app may still end up calling or escalating to a human.
A 2024 study by the Customer Contact Association found that 72% of UK consumers will try to solve an issue digitally first, but 58% will switch to a human if the problem is not resolved within the first two minutes. This “digital‑first, human‑when‑necessary” pattern is common across industries. Companies that misinterpret this behaviour risk designing systems that frustrate customers by making it difficult to reach a human.
Furthermore, the surveys that Centrica likely relies on are often conducted by technology vendors with a vested interest in promoting AI adoption. Independent research consistently tells a different story. For example, a 2025 report from the Institute of Customer Service found that customer satisfaction in the UK utilities sector had dropped by 4 percentage points over the previous two years, coinciding with an increase in automated service channels.
Striking the right balance
The challenge for companies like Centrica is not whether to use AI, but how to use it without alienating customers. A well‑designed AI system can handle routine inquiries quickly and efficiently, freeing human agents to focus on the complex, emotional, and high‑value interactions that require empathy and judgment. The best customer service operations today use a hybrid model: AI first, with an easy and fast path to a human.
Unfortunately, many companies are deploying AI primarily as a cost‑cutting tool, making it difficult for customers to bypass the chatbot. This approach can backfire. A 2026 study by Qualtrics found that companies that force customers to use AI when they prefer human contact see a 30% drop in customer loyalty and a 20% increase in negative word‑of‑mouth.
British Gas customers may be happy to check a bill or change an appointment online. A serious billing dispute is very different, and the available evidence suggests most people still want a human when the problem becomes complicated. The company’s CEO may claim that customers prefer AI, but the numbers tell a more nuanced story. As the call centres close, the real test will be whether Centrica can deliver the same level of service with fewer human agents, or whether the cuts will simply lead to more frustration and higher churn.
Source: Digital Trends News