It feels like the British state either moves at a snail’s pace or at the speed of lightning. Recall the breakneck tempo at which it rolled out the furlough scheme and Covid-19 vaccine during the global pandemic. Then contrast its inability to hit housing targets, or the fact that it took so long to build a high-speed railway line that it became the most expensive in history.
Well, we are told, worry no more: AI is coming to the rescue. While it may not be able to lay track, it shall, our leaders promise, automate interdepartmental back-office drudgery and see government become smarter at the point of delivery.
But that future remains a while away, if a new survey of public sector workers and the citizens they serve is anything to go by. According to a poll of 1,000 of each group commissioned by Appian, while large segments of both cohorts are optimistic about AI’s potential in government service delivery (44 per cent of citizens versus 67 per cent of public sector workers), less than a third (29 per cent) of public sector workers say that their department is actually delivering on that promise, while only 11 per cent believe that their organisation is fully able to leverage AI.
Trust among the public in the government’s ability to weave AI into services, meanwhile, is lacking, with fewer than half of respondents having faith in central (39 per cent) or local (44 per cent) government to manage the technology responsibly. Furthermore, 75 per cent of citizens seem unable to name a specific AI application they’ve encountered in the public sector.
The underlying reasons why awareness of – and trust in – government AI usage remain so low, says Peter Corpe, Appian’s industry leader for EMEA public sector, is likely down to how it’s being applied. More often than not, he explains, it’s either used as a standalone application or bolted on to an existing process. The use cases that result – quicker meeting transcriptions or memo drafts – are hardly earth-shattering. That seems to account for the disparity in optimism between senior civil servants, who love AI, and their subordinates, who appear much more cynical about what the technology can really offer the public sector at large.
This isn’t to say government’s efforts in promoting AI’s usage throughout the public sector have been lacklustre – far from it, says Corpe. Enthusiasm for AI among public sector personnel is also high, with 67 per cent saying that they remain optimistic it will begin to make a difference within the next five years. For all the momentum, there is still a noticeable gap when it comes to imagining what AI could mean for services outcomes.
“It needs to be put to proper lifting work,” says Corpe. Such systems shouldn’t be “something that sits in the wings and gives you a helping hand on the more mundane stuff”.
Corpe acknowledges this is much easier said than done. He admits to some frustration at the occasional carping he hears from the sidelines that, if only the British government had the can-do attitude of its contemporaries in Estonia, weaving AI through the public sector would be a cakewalk.
This criticism fails to take into account the much larger challenges facing the British state when it comes to updating legacy IT systems, a number of which predate the independence of its Baltic ally, as well as a pattern of outsourcing that can, over time, make it harder to maintain clear institutional memory.
These are not insurmountable challenges to innovating with AI, says Corpe. Instead of thinking about how it might be bolted on to an existing framework, he argues, departments instead should be designing delivery systems informed, from the outset, on how best to optimise the experience of the citizen using that service.
“That’s how it used to be for doctors,” says Corpe. “There was a patient in the middle, and your local doctor knew you and your family members really well, knew everything about the environment you worked in, the job you did, and so on. Everything pivoted around you as the individual. We seem to have gone the other way around nowadays, where we, as individuals, are pivoting around the needs of the health service.”
One practical benefit might lie in using AI to manage the assignment of medicines within overall prescription regimes, argues Corpe, radically simplifying a process that usually relies on a GP consulting very thick, heavy pharmacopoeias. That wouldn’t appear to be a huge win of itself (though perhaps a prudent place to start when the survey found 63 per cent of the public trusted the NHS to use AI responsibly), but enough of these minor victories over inefficiency in sequence would help to win back public trust in AI and boost internal support for further rollouts. You also need a comms strategy to cheerlead them, which Corpe believes the government has hitherto lacked. “Give us something to like!” he exclaims. “Give us something to point out and go, ‘That’s great, I buy into that, love that.’”
Basic communication between departments also needs to improve. Corpe recalls organising public sector roundtables where department managers meet for the first time and question their peers on why they’re exchanging obscure data points.
“They seem to have stopped talking about outcomes,” he says. “The departmental silos and ever greater workloads often result in missing the bigger picture, with people saying, ‘I’ve got my bit to focus on, and that’s what I’ll be measured on at the end of the day. When we’re done, it goes over to the next team – that’s just how the process works.’”
How do you break those siloes down? Central government pushing for greater collaboration between public sector bodies would help, says Corpe, especially from what he calls the “Holy Trinity” of data-rich organisations: the DVLA, HMRC and DWP.
“They are going to exist, in some way, shape or form, for generations. I would suggest greater access to their data from, say, local authorities could enable better, more targeted support in the community where it can make a real difference.”
The wherewithal to innovate with AI is present in each organisation, Corpe continues, but managers often “simply don’t have the mandate”. They need to be given one, he says. We are now well past the point where hype alone is sustaining investment and interest in AI. Sooner or later, “the NAO [National Audit Office] is going to start lifting the lid on this and ask, ‘Where is the return on the AI investment”.
This problem is reversible if attitudes change, Corpe believes, but time may be running out. Enthusiasm for AI-led decision making remains higher within the public sector than among the public themselves: for example, 67 per cent of the former believe that benefits compliance checks do not necessarily require human oversight, compared to just 40 per cent of citizens. The enthusiasm in government for leading by example may collapse under the consequences of poor implementation, potentially botching or slowing down future state responses to policy problems.
It is all too easy to imagine a future where AI is regarded as one technology among many, absorbed by an oleaginous bureaucracy too hidebound by existing protocols and procedures to really pay attention to its potential.
As various departments and systems strain under the weight of rising demands, shrinking budgets and stretched workforces, the question is whether this is a failure of ambition the public sector can afford to contemplate.
Download Appian’s “2026 UK Public Sector AI Adoption Outlook” at tinyurl.com/yxbam4kv
Related
