Skip to main content

Tackling the public sector puzzle with AI  

How can we use AI to lower the administrative burden on public sector workers, and meet increasing demand for public services?  

Whether someone is a nurse, a social worker, or a knowledge worker, being part of a public sector organisation can be uniquely rewarding. This is due to the significant value they deliver to people across the UK and their deep connection with local communities.  

Yet it’s also uniquely tough.

Resources are precious everywhere, but especially within the NHS and local government. The independent think-tank Institute for Government reports that local authority spending in England fell by 17.5% between 2009 and 2020, largely because of reductions in central government grants. The government’s Department for Levelling Up, Housing & Communities also published a local government funding reform policy paper in March 2024, concluding that local authorities have seen significant reductions in their spending power coincide with increasing demand for their services and inflationary pressures exceeding those in the wider economy.  

Building on this, a new report called Harnessing the Power of AI for the Public Sector, featuring findings and analysis from Goldsmiths, University of London, reveals the scale of the administrative burden on public sector workers, and how it compounds the effects of existing resource constraints. 

Three doctors in conversation walk down a hospital corridor
Hopsital staff meet in the corridor and chat while looking at a digital tablet

The admin burden takes a toll 

Researchers found that simply managing information and data is taking each public sector worker more than eight hours every week, and that this is having a major impact on staff performance and morale. A shocking 45% of public sector respondents say they are “drowning in unnecessary administrative tasks” and 45% also say this high admin workload is negatively affecting their mental health and wellbeing. More than half (55%) say the sheer amount of admin work is having a negative impact on their ability to “get on with the day job”, and 54% feel the admin load is reducing their job satisfaction and motivation.

Half of respondents also say high admin workloads are compromising the quality of service they provide (48%) and limiting the time they can spend with the public or patients (49%).   

How AI can help

The good news is that Goldsmiths’ analysis1 indicates AI could save each public sector worker more than four hours a week on administrative tasks. With an estimated 5.93 million public sector employees in the UK (as of December 2023), this equates to an overall saving of 23 million hours every week. That’s a lot of time that could be clawed back, enabling public sector employees to focus on what matters most, while also enhancing service delivery for citizens.  

These latest findings follow research on AI’s potential from Public First, commissioned by Microsoft, showing that, if AI is rolled out effectively across public services, it could save the UK’s public sector over £17 billion by 2035. That sum could fund the salaries for all vacancies currently in the NHS or to re-invest in driving better public health outcomes, a key contributor to productivity.2 

For an example of AI in action, we can look to Barnsley Council in South Yorkshire, which is using Microsoft Copilot for Microsoft 365 to reduce admin load, summarise documents, automatically record meetings, and produce notes and actions. This frees up Barnsley’s office workers and social workers to be more present in meetings or focused on the family they are there to support. Thanks to the power of generative AI and its seamless integration into the productivity apps staff use every day, from Microsoft Word and Teams to Excel and Outlook, Copilot is already achieving more than a 50% adoption rate and saving workers a great deal of time. 

As Wendy Popplewell, Executive Director Core Services at Barnsley Council, explains:  

“Our employees spend a lot of time reading emails, reports, spreadsheets and compiling various documents for things like funding bids. That first draft where you spend time getting started and think: ‘What do I need to write here?’ Copilot just creates that for you, which is a total game changer.  

We also have people within social care who spend a huge amount of time writing up case notes, meeting minutes and actions – taking them away from face-to-face interaction with our residents. If Copilot saves them a few hours every week, the impact on community and employee wellbeing is massive. People don’t become a social worker to do admin, they want to be spending time with families to help solve their problems.  

Of course, we want people to look at the job they do and figure out where there are potential efficiencies, but we’ve got to give them the space to think in the first place.  

This is what this Copilot is creating – time to think.”

a man and a woman looking at a computer screen
Office workers interact around a computer screen

Augmenting human expertise 

As well as easing the general admin burden, narrower AI use cases show how this technology can boost human performance in extremely high-skill areas, such as breast cancer screening, a disease that remains hard to detect. Oncologists can look at around 5,000 breast scans per year on average, and view 100 in a single session, and an element of fatigue and potential distractions in the workplace are part of any human review process.  

NHS Grampian, which provides social and healthcare services to more than half a million people in the north-east of Scotland, carried out the first official prospective evaluation of Kheiron Medical Technologies’ “Mia” AI tool. Mia was piloted alongside NHS clinicians and analysed the mammograms of over 10,889 women. 

While most of these patients were cancer-free, Mia successfully flagged all of those with symptoms, as well as an extra 11 the doctors did not initially identify because the tumours were so small. This means Mia helped doctors find an additional 12% more cancers compared to routine practice.  

If deployed across the entire NHS, a 12% uplift in the detection of breast cancer could lead to better outcomes for thousands of women across the UK. The AI-augmented workflow also showed a decrease in women recalled unnecessarily for further assessment and modelled a significant reduction in workload. Learn more about how AI can reduce workload by up to 30%.

While it’s still early days in the public sector’s learning and experimentation with AI, especially generative AI, the promise and progress so far are hugely encouraging. Public and private collaboration, especially around responsible deployment, and a rigorous approach to measuring the ROI and social impact of projects will be crucially important for sustainable success.  

Find out more

Harnessing the Power of AI for the Public Sector presents comprehensive research with seven key recommendations for UK government


1 Research commissioned by Microsoft in partnership with Dr Chris Brauer, Goldsmiths, University of London, in May 2024.

2 Unlocking the UK’s AI Potential: Harnessing AI for Economic Growth, Microsoft, 2024.

About the author

Hugh MilwardHugh Milward leads Corporate, External and Legal for Microsoft in the UK, with a seat on the UK leadership team. His focus includes work to help organisations overcome legal and regulatory hurdles to their technology adoption and transformation, managing some of the complex geo-political issues relating to tech, and working to ensure no one is left behind from the onward march of technology. Hugh’s background is in politics, corporate affairs and reputation management, working for some of the world’s most high-profile brands including Starbucks and McDonald’s. He is passionate about the interrelationship of society and technology. Hugh is a Board Director of the New West End Company, chairing its Public Affairs committee, and a member of the Board of Directors of British-American Business. Hugh sits on the SE Council of the CBI and on the Advisory Board of the Institute of Coding.