UK Government's AI Ambitions: Promises and Pitfalls in Public Sector Transformation

AI won't transform public services unless someone does the hard work on process change

Governments the world over typically come into power promising to cut out public sector waste and improve productivity in the civil service. More recently, the new trend is to cut out entire swathes of the civil service completely, as in the figurative chainsaw wielded by Argentine's President Javier Milei. Meanwhile in the US, we wait to see how tech bros Elon Musk and Vivek Ramaswamy will deliver their plans for the incoming administration's Department of Government Efficiency (DOGE). In the UK, the recently elected Labour government this week revealed its plans to harness AI, with the aim of fostering economic growth as well as transforming how the public sector operates. No doubt such ambitions beat in the hearts of government leaders around the globe, emboldened by the promises of tech leaders that a new breed of tireless AI agents will soon save human workers from mundane toil. But as my colleague Stuart Lauchlan notes in his analysis today, there's a sorry track record of governments turning to technology for a magic pill that will cure public sector woes. Those mundane and redundant tasks are somebody's job, and the political will to force through change -- along with the necessary reskilling, redeployment, retirement or layoffs as appropriate -- is rarely strong enough to overcome the "administative won't," as Stuart reminds us. The reality is that sprinkling the latest technological pixie dust across the civil service, or any other long-established organization, is never going to be enough on its own to effect lasting change, however much prime ministerial or presidential passion and prestige surrounds the initial announcement. What's needed is a concerted effort on the ground to analyze existing processes, identify where change is needed, and then follow through with a proper change management plan. So I was intrigued to see another announcement cross my desk this week, revealing that the UK Cabinet Office, the body responsible for the effective running of government in Britain, is working with business process vendor Celonis to streamline government processes. The background to this partnership is an initiative first launched by the Cabinet Office in 2018, when Teresa May was Prime Minister. The aim is to consolidate back-office computer systems for over 118 government departments and other bodies into five shared services centers, involving the replacement of around 286 legacy IT systems that sit siloed in individual departments. The new centers will ultimately support around half a million civil servants, as well as over a million military personnel, reservists and veterans. Celonis has been recruited to support the creation of improved, standardized processes, such as onboarding of new civil servants so that new joiners can get up to speed more quickly. From the vendor's press release: Celonis will enable the team to identify any anomalies, roadblocks or out-of-date processes, which can be fixed or removed. It will also highlight how cross-departmental processes can be optimised by providing recommendations on best practices, presenting clear visualisation of process data collected across government. While AI isn't mentioned in the press release, Celonis sees its platform as crucial for providing essential context when applying AI to business operations. In focusing on a narrow use case like onboarding new joiners, Celonis is following its own advice, as outlined by Rupal Karia, its VP and Leader for the UK & Ireland in conversation with diginomica's Derek Du Preez in November. His view is that AI isn't a "silver bullet" and needs to be adopted thoughtfully. He comments: There are companies that try to go quite bold and do quite meaningful things, but it's actually quite expensive. If you look at the cost of failure, it is pretty high as well. I think the thing which is important is picking a particular use case to go after. Rather than using AI to try and fix the whole world, [think about] is your problem supply chain? Is it customer value? Is it customer acquisition? What is the thing? Try and do one thing well and then look at how you try and deploy. One of my big worries about the rush to adopt AI in business and government is that much of the investment will be wasted on automating redundant or broken processes that can't be fixed simply by throwing technology at them. Adding a new layer of automation may provide the appearance of superficial improvements, but it won't address the underlying problems. The way many organizations operate hasn't changed substantially since the days when people exchanged work by passing sheets of paper around and the only way to contact someone in another office was to pick up the phone. Digital connection makes it possible to cut out much of the paper-passing and disjointed decision-making that persists from those times. Taking a fresh look at how things are done and adopting new standardized processes is the way to deliver real change, which then provides a much more streamlined foundation on which AI can layer further automation. It's good to see that the Cabinet Office has the ambition to take this approach -- but given this is an initiative that has already been seven years in the making, you can see how long it takes to actually deliver change on this scale.

Is AI really the silver bullet to transform government or another tech pothole to fall into?

Amid the grand ambitions outlined in this week's UK Government AI Opportunities Action Plan and all the talk of taking on world leaders in the field, something reassuringly British stands out about one of the proof points for AI's transformative potential: It can spot potholes quicker. That's a very British obsession - complaining about the number of potholes in the road. It's a use case that presumably someone has concluded would be one that would appeal to the widest possible UK demographic. Personally I was reminded of the supposed vote-winning 'Big Idea' of the John Major premiership in the 1990s which was a Cones Hotline - a dedicated telephone number to report abandoned or excessive traffic cones left on public highways . But potholes aside, there is big public sector services transformation envisioned in the Action Plan. As Prime Minister Keir Starmer pitched it, this is about "totally re-wiring government": AI isn't something locked away behind the walls of blue-chip companies. It's a force for change that will transform the lives of working people for the better...It will make public services more human, re-connect staff with the reasons they came to public service in the first place. The Plan includes the setting up of a Digital Center of Government which will look for AI opportunities and pilot them, before scaling solutions up across the wider public sector. The Prime Minister has also written "personally" to all members of the British Cabinet, putting them on an AI war-footing and emphasising that this has to be regarded as top priority. All good. Except...well, except we have rather been here before. Some of us are old enough to remember when e-procurement was going to transform government spending, slashing away at the bureaucratic paper chase that meant that a 50p packet of paper clips ended up costing the taxpayer £50 once all the necessary administrative boxes had been ticked. What stopped that from working? Not to put too fine a point on it, the pen pushers whose jobs depended on that paper chase still being in place. {Amusingly enough, the AI Opportunities Action Plan includes a telling reference to procurement: Government purchasing power can be a huge lever for improving public services, shaping new markets in AI, and boosting the domestic ecosystem. But doing this well is not easy - it will require real leadership and radical change, especially in procurement. Indeed it will. What chance we get it right this time by throwing AI at the problem? See also Mark Chillingworth's view on this topic. ) Or there's Cloud First. The UK did a great job with its G-Cloud initiative, despite having a lot of hurdles to overcome in terms of both buyer and seller skepticism, and later with its Cloud First declaration that, akin to the US version of the same, essentially said procurers of government IT services or solutions needed to have a damn good reason why they weren't going down the cloud route. That might be marked down today as having had marginal success, but one glance at government computers still running old versions of Windows tells its own story. Or we might think back to when the genuinely radical and transformative Government Digital Service (GDS) was set up to shape government IT strategy and put an end to the grip held by what was termed The Oligarchy of the same old, big ticket tech and services vendors. No more multi-year, multi-billion pound contracts was the rallying cry and for a while it seemed to be working. But all The Oligarchy had to do was bide its time, wait for the inevitable political musical chairs and business-as-usual could be restored, helped in no small part by the internal government IT establishment not liking being told what to do by GDS and civil servants who saw their empires being eroded in the process. So, what can we realistically expect from this upcoming re-wiring of government? AI is sufficiently sexy as a subject that the political will will be there to get at the forefront of change, even if the administrative won't might prove a lot more resistant to change. One of the big selling points of AI in the private sector is the supposed opportunity to improve efficiencies and automate mundane or redundant tasks, of which there are a hell of a lot in the public sector. There's also the ongoing fear of job losses among private sector workers and that's going to be potentially even worse in government circles. In other words, the cynical conclusion has to be that the biggest barrier to public sector AI-driven transformation will come (again) from the public sector itself. Meanwhile a couple of bits of research adds a bit more grist to the mill in terms of public sector attitudes to AI both from the point of view of services deliverers and recipients.. Firstly, research from Public First, commissioned by Google Cloud into the potential of generative AI and based on polling 415 public administration workers in the UK. This estimates that greater use of gen AI could lead to annual cost savings across the public sector of £38 billion a year by 2030. The research states: By facilitating more effective working practices and automating repetitive, bureaucratic tasks, generative AI could save enough time to allow for an extra 3.7 million GP appointments, a 16% increase in the teacher to student ratio, and freeing up the equivalent of over 160,000 police officers. All of those are politically crowd-pleasing stats. In contrast though the report warns: Conversely, if the UK cannot leverage generative AI to boost public sector productivity, which has not increased in over 25 years, the Treasury will have to find another £12 billion in revenue for the next fiscal year. Among the barriers to AI adoption cited by public administration managers were presumed legal/regulatory barriers that made them cautious about using AI tools more extensively (60% of respondents); a need for better structured data sets to access (55%); while only a third of those polled felt they had the right skills on tap to take advantage of AI. (The new Action Plan does cite policies to tackle each of these.) Meanwhile new multi-country data from Salesforce, based on 11,750 political constituents, is angled more towards how successful public sector adoption of AI could improve their engagement with public services. In the current Agentforce-centric era, there's an big agentic angle at play here, of course, with 90% of global respondents saying they'd use an AI agent to interact with the public sector. The percentages vary by country - 88% of US respondents in this case vs 84% in the UK, but 100% in India. Globally, the top three reasons that would make respondents use an AI agent to engage with the public sector were cited as: getting 24/7 access to information and services (47%) more efficient access to government resources (44%); and reducing the number of websites and steps needed to address their needs (40%). In the US, 43% of respondents talked about the importance of keeping personal data safe. Again, India is particularly enthused with 70% looking for that 24/7 access to services. While I resolutely refuse to believe in silver bullets, I really want to believe that AI might be the transformative tool to change government and how we interact with it. As a taxpayer, I'd love to be proven wrong on my fears that we'll repeat the same mistakes as other tech-driven overhauls. Politicians love to be on the bleeding edge of modernity - it was the 1960s when then-Prime Minister Harold Wilson was bigging up the UK's role at the forefront of the "white heat of technology". But innovation stalls when it crashes into process. I've just had engagement with a major UK government department that involved me filling out a 40+ pages paper form, sticking it in an envelope and posting it via snail mail, and now having to wait for eight weeks before I can expect a response from said department, a response which will also come in a letter. There are success stories, of course - UK passport renewal, for example, which has become very slick and efficient and driven by online interaction. The lesson here - finding the early use cases to point to AI success is going to be critical in the coming AI "re-wiring", both in the UK and across governments worldwide. Otherwise we're just going to have to reconcile ourselves to another decade or so of spotting potholes using our eyes.