PublicSector
Is there a role for AI in public sector finance teams?
As artificial intelligence continues to dominate headlines, jostling for control of this hugely disruptive technology is firmly underway – and the public sector is by no means immune. Rachel Willcox reports
Words: Rachel Wilcox Illustration: Andrew Nye
Sir Keir Starmer unveiled plans in March 2025 to use AI to digitise the civil service, a move that he said would deliver up to £45bn in cost savings and productivity benefits and deliver the UK government’s Plan for Change.
The Labour administration’s commitment to all things AI was reinforced by its most recent spending review, which saw the government commit to allocating some £2bn to AI and related computing tech over the next four years, much of it aimed at implementing the 50 recommendations of the AI Opportunities Action Plan.
But talk is cheap and, despite a flurry of announcements, research published in September shows that the public sector is at risk of falling behind in the AI race.
One in three public sector respondents said their organisation is planning to invest in AI, according to a survey commissioned by Access PaySuite. However, just 10% have significant strategies in place compared with 45% in the private sector.
It brings into question if the finance officer, the AI vendor or the algorithm is ultimately responsible.
Automating everyday tasks
For public sector finance teams, AI offers opportunities to automate routine administrative tasks such as payroll and financial reporting, while offering tools to spot fraud or duplicate payments in real time and detect trends and risks in large datasets.
There is also scope for AI to be used for policy analysis, decision support and research. In financial operations, AI is already automating tasks such as invoice processing and demand forecasting, and sharpening cash flow forecasts. It can even cut down on paperwork by turning long policies, contracts and audit notes into clear, usable summaries.
Christian Bathews, government lead at training company mthree, says: “The benefit is that finance teams can close the books faster, improve forecasting accuracy and maintain tighter control. Ultimately, this frees up professionals to focus on analysis, advice and oversight, where their judgement adds real value.”
Dr Andrew Rogoyski is director of innovation at the Surrey Institute for People-Centred AI at the University of Surrey. He says: “There are some interesting uses in terms of predictive policy impact, emerging social and economic issues, evidence-based policy-making – being able to pull policy ideas not just from the UK but around the world.”
However, many public sector finance teams struggle with how to integrate the tools that are starting to emerge. “Organisations are piloting things and getting it wrong,” Rogoyski says. “Very often, it bumps up against the organisational appetite to change. Where organisations are not terribly agile, they struggle to adopt AI because it’s disruptive.”
Add to the mix the intense scrutiny on public sector IT projects against a backdrop of tight funds and it’s no wonder that caution is rife. “The risks are very real if AI is rolled out without proper preparation,” Bathews says. “Poor-quality data will lead to poor outcomes, and black box systems make it hard for public sector teams to explain why a decision was taken.”
Image: Alamy
AI at Islington Council
Paul Clarke uses Microsoft Copilot to streamline administrative processes in his role as acting corporate director of resources at Islington Council. “If it’s recording a meeting, it can give you a summary and the actions,” he says. “If I’m asked a very specific question about something I know is in my inbox or Teams messages, it will find it and be productive in the moment.
“With forecasting, the technology has pretty much got to a place where it is as accurate as accountants, so they can focus on decision-making rather than business-as-usual stuff. As an organisation, we’re on the precipice of looking at that.
“Our current system involves a lot of manual effort to get a payment out the door. In the future, we will be able to entirely automate that process so staff can focus on supplier trends and insight. We’re replacing our finance system with one built around cloud technology that incorporates AI. The vision of the future is to get to a position where financial forecasts drop out and you can overlay what you’re doing as a business.”
The capabilities of AI will produce something that makes sense, but it’s also about understanding the context of the organisation and what you are trying to achieve, not to mention the limitations of AI, Clarke says. “It’s not the future, it’s the now,” he adds. “The challenge is to do it in a safe way.”
Evolving finance roles
The opportunity is not to replace roles but to reshape them. “Instead of manually processing every invoice or transaction, clerks will spend their time dealing with the unusual or higher-risk cases that automation can’t handle, the errors, the gaps and the potential fraud,” Bathews says. “Management accountants will shift their focus towards forecasting and scenario planning, while auditors will move toward continuous, AI-assisted auditing.”
This doesn’t have to mean fewer jobs but it does demand new skillsets. Everyone should understand the basics: what AI can and cannot do, how to question its outputs and how to use it responsibly within policy.
Paul Clarke, acting corporate director of resources at Islington Council, says: “It’s not just about being excellent at writing prompts; you need strategic business thought to understand what you’re trying to do and how to do it.
“Three or four years ago, the challenge was constrained budgets. Today, we’ve got unstable politics, divided communities and a cost-of-living crisis. Day to day, the role of finance is helping organisations navigate ever choppier water. We need humans focusing on that and computers focusing on stuff that they are better at like faster, more accurate transactions.”
However, workers will need to be carefully managed, upskilled and supported throughout the journey, according to Towers. She says: “While automation may reduce the need for traditional, transactional roles, alongside more strategic work there will also be clear opportunities for finance teams to upskill around AI implementation, assurance, monitoring, governance and risk management.”
Rogoyski’s advice for future-proofing careers is to recognise that learning is a lifelong process. He says: “Be flexible, be adaptable, develop core human skills like critical thinking, good communication skills. If you latch on to something that’s very niche and highly automatable, you’re going to be under pressure.”
Provided we don’t get too sucked into the hype and focus on how to make this technology work for us, it will be a force for good, Rogoyski believes. He adds: “You’ve got to think about how you’re supporting human beings so they stay in the loop, they stay accountable and they are steering it in a direction you want to go. Get that right and it could be very positive.”
Efficiency at the cost of scrutiny
Over-automation can also weaken the checks and balances that protect public money. “That’s why privacy, security and accountability need to be built into every system from the start,” Bathews adds. “AI should inform decisions, not replace them. Every automated or assisted decision must leave an audit trail showing which data was used, which system made the recommendation and which person approved it. That way, transparency and accountability are never in doubt.”
Sarah Towers, operations director at business change consultancy Entec Si, says AI governance frameworks, including policies on when and how AI can be used, are needed to address the accountability gap that occurs when AI systems make or recommend financial decisions. “It brings into question if the finance officer, the AI vendor or the algorithm is ultimately responsible,” she says. “Ethical oversight boards should also be established for high-risk uses, such as fraud detection.”
The potential for deskilling is another source of concern. Rogoyski urges against taking the AI dividend too early. “The potential for overreliance on increasingly intelligent systems, to the extent that you forget how to do it yourself, could be very damaging in areas like finance,” he says.
