Prime Minister Keir Starmer and Peter Kyle, the secretary of state for science, innovation and technology, intend to conduct a sweeping reform of the civil service, with AI set to a play a central role. They hope the technology can not only make civil servants but also cut headcount.
Starmer says he is “determined to seize” the “golden opportunity of artificial intelligence” to deliver cost savings. In an interview with Sky News, Peter Kyle said it’s likely that digitalisation and AI would reduce headcount in the civil service, echoing the function of the public-sector AI task force proposed by the previous Conservative government.
Kyle complained that employees in the Driver and Vehicle Licensing Agency are spending their time “opening envelopes” while those in HMRC passed their days answering thousands of phone calls.
A workforce displaced by AI doesn’t just vanish – it creates ripple effects across industries
Whether paperless offices or automated call agents constitute artificial intelligence is open to debate, but the talking points are an unusually candid reversal from the normal messaging around artificial intelligence – that AI will be used to augment human abilities rather than replace them.
While the Public and Commercial Services Union, which represents workers in the UK government and public bodies, acknowledges technology has a part to play in improving public services, it also states that AI “cannot be used as a blunt instrument to cut jobs”.
Industry figures, though, do agree with Starmer’s conclusion that AI represents an opportunity to improve public services. Amanda Brock, the CEO of OpenUK, an open source technology advocate, says: “Our governments have no choice but to bring in AI, but they need to do this in an informed a manner.”
The key to a successful future, Brock says, will be “to adapt and work with these changes and not to fight them”. “AI will be really important to the public sector and we simply can’t let it lag behind enterprise,” she adds. “So the government is right in looking to use it.”
Bill Conner, a former GCHQ adviser and CEO of Jitterbit, an automation company, says enterprise AI can introduce “untold possibilities” for the government. But, he warns a “rip and replace” approach to implementation threatens to “hinder internal adoption and introduce unnecessary risks”.
“AI isn’t a one-size-fits-all solution,” he adds. “Its success hingers on the humans who operate and manage it. Even with AI automating tasks, human expertise remains crucial for decision-making.”
How would AI replacement work?
The use of AI in these reforms raises questions about the replacement of human labour with machines – whether in the public or private sector. How should the process be managed, what are the main goals and how can its effectiveness be measured?
“If the goal is pure replacement, then the primary outcome will be cost reduction,” says Dr Alexandra Dobra-Kiel, who is innovation and stategy director at consultancy, Behave. “But even cost reduction is questionable when considering broader economic and social costs. A workforce displaced by AI doesn’t just vanish, it creates ripple effects across industries, from mental health crises to shifts in tax revenues and public services.”
Organisations that seek to replace human employees with artificial intelligence will not be able to find a like-for-like replacement. Currently, although generative AI can be useful in writing code, there’s still a need for well-trained software engineers to collate, document and audit what it writes. Even the most advanced agentic AI will likely require some human oversight, too.
If AI reduces costs but degrades services, the trade-off may not be justified
So the real question should not be whether AI can replace humans, but how AI can work “in tandem with humans to achieve excellence rather than just efficiency”, according to Dobra-Kiel. Using both in conjunction could enhance creativity and strategic thinking, she says.
Fabian Braesemann is a social data scientist and the departmental research lecturer in AI and work at the Oxford Internet Institute (OII). Braesemann co-authored a paper at the OII, Winners and Losers of Generative AI in the Freelance Job Market, which investigated the complex ways in which tools such as ChatGPT are reforming labour markets.
Braesemann, who is positive about the public sector introducing artificial intelligence, notes that job roles consist of many different tasks, each of which require different skills.
Some jobs, such as the typists in vast office typing pools, have been automated out of existence and few would suggest bringing them back. But for many others, it’s unlikely that entire roles will be directly replaced, says Braesemann, but elements within those roles might be.
Look at the admittedly extreme example of a painter and decorator, he says. While most of their work is not replaceable, some cognitive tasks, such as writing and chasing invoices, could potentially be automated. The same is true of any job, according to Braesemann. Attempts at automation must follow a “task-based approach” rather than a wholesale like-for-like replacement of any single role.
Can humans compete with AI?
Sometimes, introducing AI can complicate workflows or introduce new challenges. For instance, with the inclinations of generative AI to “hallucinate”, employees have to spend greater time scrutinising its output, should errors find their way into important documents.
Dobra-Kiel advises that organisations work to evaluate AI against human benchmarks. These could cover accuracy and error rates – the number of errors found in AI-generated decisions versus human decisions – to ensure AI systems are not introducing biases or systemic failures into organisations. Or they could benchmark cost savings against service quality. “If AI reduces costs but degrades services, the trade-off may not be justified,” she says.
Human expertise remains crucial for decision-making
It will also be crucial to measure employee engagement before and after AI integration, she adds, to track whether employees are facing undue stress or alienation, as well as ethical considerations such as audits of demographic biases.
Organisations should scrutinise the efficiency and productivity gains promised by this technology. The metric here could be the time taken to complete tasks before and after AI integration. “AI’s primary advantage is speed,” Dobra-Kiel says. “So organisations should measure how much time is saved without sacrificing quality.”
However, the civil service may be at a disadvantage to the private sector when it comes to implementing AI. Some institutions, such as the civil service, were designed differently to the digital-native businesses are now racing to adopt AI, Braesemann notes. While Starmer may be keen to reap the benefits of AI in the public sector, this may create some organisational challenges when establishing more agile ways of working.
Prime Minister Keir Starmer and Peter Kyle, the secretary of state for science, innovation and technology, intend to conduct a sweeping reform of the civil service, with AI set to a play a central role. They hope the technology can not only make civil servants but also cut headcount.
Starmer says he is "determined to seize" the "golden opportunity of artificial intelligence" to deliver cost savings. In an interview with Sky News, Peter Kyle said it’s likely that digitalisation and AI would reduce headcount in the civil service, echoing the function of the public-sector AI task force proposed by the previous Conservative government.
Kyle complained that employees in the Driver and Vehicle Licensing Agency are spending their time “opening envelopes” while those in HMRC passed their days answering thousands of phone calls.
