Can AI improve UK public sector productivity?

The day before world leaders gathered for UK Prime Minister Rishi Sunak’s artificial intelligence summit at Bletchley Park earlier this month, there was a less heralded but arguably equally important meeting of more than 100 Whitehall civil servants at the London headquarters of consultants PwC.

Aided by teams of AI technologists from Microsoft, the officials were invited to a “hackathon” to examine ways that AI could be used to boost productivity in the UK’s cash-strapped public sector.

“We have to find the things that make the most material differences to people,” said Mike Potter, the government’s chief digital officer, before sending his officials to begin their “hacks” using the latest AI tools.

The hope is that applying AI will help to deliver on chancellor Jeremy Hunt’s ambition, announced in June, to increase public sector productivity by 0.5 per cent a year in order to “stop the state growing ever bigger as a proportion of our output”.

Hunt claims plans set to be announced on Monday by the Home Office could save police officers up to 38mn hours a year of “unnecessary bureaucracy”. A review to be published in Wednesday’s Autumn Statement “has revealed huge opportunities to cut admin, safely harness AI and deliver early interventions to relieve pressure on public services,” he said.

The chancellor’s message resonates with Tory backbenchers keen to see him create space for tax cuts ahead of an election expected next year. But any government will need to confront the need to manage relentlessly rising demand for public services, against which current spending plans already look unrealistic. 

The Institute for Government think-tank warned last month that public services were stuck in a “doom loop” of perpetual crises because erratic, short-term decision-making by successive governments had left buildings crumbling, driven away experienced staff and made it “impossible for public service leaders to plan or implement performance-enhancing reforms”.

However, economists caution that the scope for quick productivity wins may be limited and say that while AI holds potential, the real solutions will lie not so much in breakthrough new technologies as in upgrading basic IT, boosting inadequate capital investment, improving processes and management capacity and transforming workplace relations. 

“The idea that there is loads of free productivity growth to be had is wrong,” said Torsten Bell, director of the Resolution Foundation think-tank. He noted that recent revisions to official gross domestic product data had erased much of the post-pandemic drop previously seen in public sector productivity, leaving less scope for a rapid rebound. 

In the years before Covid-19, productivity grew faster in the public sector than among private businesses, but this now looked more like the effects of cheese-paring austerity than a genuine improvement, Bell added. 

Nowhere is the productivity challenge more acute than in the NHS, where staff are already strained to breaking point, pay has not kept pace with private sector earnings for many years, and the demand for care is set to rise inexorably as the UK population ages.

Some hospitals are already using AI tools to help radiographers analyse X-rays, speed up bookings and referrals, or are deploying speech-recognition technology to take clinical notes.

But NHS leaders say that while AI may be transformative in the longer term, the immediate challenge is to spread existing best practice across a long tail of poorly performing hospitals. 

An NHS hospital ward at Ealing Hospital in London
Nowhere is the public sector productivity challenge more acute than in the NHS © Jeff Moore/PA

“You can’t possibly harness AI when staff can’t access a basic computer that doesn’t have multiple logins and takes half an hour to switch on,” said Anita Charlesworth, director of research at The Health Foundation think-tank, noting that some hospitals still worked only with paper records. 

“We have underinvested in capital, which is critical to deliver productivity,” she added. 

Sir Julian Hartley, chief executive of NHS Providers, which represents health organisations across England, said the biggest productivity gains would come from tackling staff burnout to improve retention, solving the problems in social care that prevented hospitals discharging patients, and stepping up capital investment in basic IT systems. 

“Without that, it’s really difficult to layer on AI or clever pieces of software — you’ve got to have the basics right,” he said. He added that investment in electronic patient records and wider digital transformation was being squeezed by the Treasury’s refusal to fully offset the financial impact of strikes. 

AI tools can help other parts of the public sector sustain or upgrade services frayed by the funding cuts of the past decade. 

At the hackathon officials focused on six targeted areas: raising call centre efficiency, getting best value for public contracts, streamlining ministerial “red boxes” for better decision-making, improving policing, flood defence modelling, and improving skills and training in the public sector. One group focused on how AI could instantly summarise conversations, freeing up time for call centre workers.

Local government is already deploying AI to help reduce the number of callers needing human attention.

Telford & Wrekin council said the introduction of its AI-powered Ask Tom chatbot had cut waiting times by nearly 20 per cent and increased the capacity of call-handling centres covering everything from housing to council tax and bin collection.

“We haven’t reduced headcount, but we have brought in four new services — library, registration of births, deaths and marriages, homelessness and free school meals — all with no extra cost,” said Gemma Hancox, Telford’s customer contact team manager.

But senior Whitehall officials say it will take time and investment to realise the benefits of AI. 

“We’d like realism on the need for investment in the underlying infrastructure. We can’t flick a switch,” said Hartley. 


This website uses cookies. By continuing to use this site, you accept our use of cookies.