When it comes to adoption of artificial intelligence (AI) by the public sector, Paul McGinness believes use of the technology requires trust. Tech firms like his, he says, not only need to demonstrate that it can work, but that it gives confidence to those using it.
And once you see the proven benefits – by improving a business process or helping doctors diagnose disease quicker – then it makes the case to go bigger and bolder.
“At the moment we’re mostly seeing the public sector take their first steps with the technology,” says McGinness, founder of Edinburgh-headquartered digital consultancy, Storm ID.
“At the lower end of the adoption curve, they’ll maybe look at AI solutions that make small and pragmatic improvements to the workflow, but when you start to demonstrate that trust, then the big, more dramatic solutions that promise to reimagine the way we do things, they suddenly become a little less frightening.”
I speak to McGinness midway through the first public sector AI Challenge event, run in partnership with Futurescot during the summer months.
From a standing start, the competition, in which the company offered the chance for public sector agencies to win £20,000 in consultancy time towards AI adoption, generated more than 30 responses, many of them highly innovative.
For McGinness, pictured below, it demonstrated not only that there was a strong appetite for AI across the public sector – be it healthcare, central or local government, or some of Scotland’s most important public institutions – but also that organisations are starting to think creatively about how they can deliver services.
For him, that signifies not only that the technology is being considered as a means of doing work that humans find either repetitive or laborious, but also that leaders are starting to think about paradigm shifts in how services should be delivered in future, made possible, in part, by the dawn of large language model (LLM) technologies.
“What we’re trying to do is look at the opportunity to do things that not just improve the workflow, but perhaps radically changes it. And I think those conversations are easier to have when you’ve introduced AI, maybe with a human in the loop, to provide that assurance and trust,” McGinness says.
And as for the AI Challenge itself, he adds: “We’re delighted with the response so far. I think the interest shown by a broad section of the public sector shows there is a lot of demand out there for introducing AI across a range of areas.
“People are looking at AI as a potential solution for automating certain processes within their organisations, to improve operational efficiencies, but also to provide decision support, which is where that higher value comes in.”
And it is not just the value to clients. Storm ID has seen its own revenues grow from AI deployments by 86 per cent in the last year, as more organisations turn to the technology.
The company is already working with some of the most forward-thinking government agencies across the UK, including Historic England and the Foreign Office, as it sees the demand for AI rise as more people realise its potential.
One deployment has seen integration of an LLM into a web forms engine that can convert PDFs and policy information into web forms, which saves time and energy in recreating information in different formats for stretched government teams. “If you look at government websites, there are a lot of PDFs, so this simple use of AI short circuits a lot of effort involved,” adds McGinness.
“Another is where we used the technology to interrogate large amounts of unstructured data to answer questions on whether aid programmes met the objectives set out in their original business case, were delivered to budget and generate key findings to inform future business cases.
“Humans are still relied upon for decision making, but the AI removes a lot of the administrative burden helping staff to focus on higher value activity.”
Incremental improvements that the technology can bring to workflows for certain departments is one thing. But the evidence base for scaling up AI across government is also starting to mount.
According to a paper released by the respected Alan Turing Institute think tank this year, AI could help automate about 84 per cent of repetitive transactions across 200 government services.
According to its researchers, UK central government carries out around one billion citizen-facing transactions per year spanning almost 400 services – including passport applications and universal credit processes – and 57 departments.
For McGinness, it comes back to the trust matrix. “So, I think it’s going to take time before we delegate decision-making to machines,” he says. “But the more it’s able to perform the tasks that are routine, the higher the likelihood that we’ll see organisations move up the trust axis. And when that happens, we’ll maybe start seeing attended workflows, with humans in the loop, become autonomous ones.”
Storm ID has developed a strategy to support its delivery of AI, and the challenge process has refined some of its own thinking. Its view is that AI will become more embedded into organisations over the next decade and the ability to deploy, manage and orchestrate AI agents will be a key differentiator for organisations.
Failed systems have human impacts, as the Post Office scandal so tragically illustrated. When computers are trusted over humans, it causes something of a moral panic. So companies like Storm ID have to ensure their “approach” is more than just a technology one.
“Since going through the AI Challenge process, we’ve actually strengthened our approach around ethics and inclusion,” says McGinness. “We’ve developed discovery workshops for AI, and one specifically for LLMs, which tries to show people what might be possible within a pilot or prototype.
“In a healthcare setting, we’ve also looked at assessing the performance of models based on protected characteristics including age, ethnicity and gender, and to really scrutinise whether we are introducing any unfairness or bias. But asking the questions, and understanding potential harms, is key for all of us to try and address.”
And then you have to weigh the risks of not introducing AI on under-pressure public services. Recent Office for Budget Responsibility (OBR) forecasts paint a stark picture for policymakers to address.
In its September fiscal paper, the OBR projected government spending will rise from 45 to 60 per cent of GDP, based on the current policy landscape, driven largely by rising costs of healthcare, an ageing population and the ballooning national debt.
McGinness adds: “At some point we’re going to face the question of can we afford not to adopt these kinds of technologies into public services, in healthcare and other places.
“Because the demographics are such that we’re going to have to find better ways of delivering public services. AI is definitely part of the solution and we should embrace it.”
Up for the Challenge
Three public sector organisations went forward to the final stage of the public sector AI Challenge, run by Futurescot in partnership with Storm ID. The winner will be announced at Digital Scotland on 26 November at the Edinburgh International Conference Centre.
- Scottish Parliament – Developing an AI system to transform the way citizens engage with Scottish Parliament live webstreams.
- National Library of Scotland – Using AI to help audiences discover and connect with Scotland’s rich audio-visual heritage.
- NHS Forth Valley – Harnessing AI technology to improve both the processing of complaints data and to draw out insights and learnings for service improvement.
Partner Content in association with Storm ID