Public sector organisations deploying artificial intelligence (AI) must ensure they identify solid use cases for the technology to get buy-in from their teams, delegates at a recent Tech Monitor event heard.
The Tech Monitor Public Sector Technology Symposium 2021 brought together leaders from a wide range of public sector bodies to discuss the ways technology is impacting their work and enabling them to deliver better services.
AI was put in the spotlight in a panel discussion entitled Putting AI and automation at the heart of public sector digital transformation. The discussion, and all the other sessions from the event, are now available to watch. Register here to watch the full panel on demand.
AI cannot be the end game
Delegates heard from Charlie Boundy, head of data science at the Department for Work and Pensions (DWP). The department has used AI to help improve the quality of service it offers, Boundy explains, with automated systems being used to flag up potential issues and support human staff in their decisions.
“The way we’ve done this over the last few years is that it’s not AI that is the end game,” Boundy said. “It’s very much we’ve been using data in certain specific problem areas or challenging areas that have been there for a while. So it might be how we’re developing major change programs, and making sure that we’ve got the telemetry and information on how our services are being used.”
Julie Pierce, director of openness, digital and data at the Food Standards Agency (FSA), said the agency has been using AI, among other things, to help assess food businesses and predict what their hygiene score might be based on available data. This can then help inform priorities for inspectors.
While these use cases are now well-established, Pierce said this hasn’t always been the case, but being able to see AI in action has helped staff engage with the technology’s potential. “When we started [using AI] a few years ago it was hard to see what the candidate use cases were and there wasn’t a lot that was being volunteered,” she said. “But once we started to deliver and people engaged with it, that’s created a demand that is greater than our capacity, which is which is a good place to be.”
The benefits of embedding data scientists
AI systems typically rely on a high volume of good quality data. Boundy says gathering this data has been a big challenge for DWP. “A lot of the information that government departments tend to collect is quite two-dimensional data,” he says. “It’s at a certain point in time and is a description of a claim or a tax return or an application or something like that.”
Boundy said his team needed more information about outcomes to help them develop useful AI systems, so they undertook a factfinding task to identify areas where governmental data was lacking. For this, he says it was vital to ensure data scientists were involved at every step to make sure the right type of information was logged that could be used to build and train the department’s AI.
“We’ve had data scientists embedded in delivery teams and feature teams,” he says. “They can, as they go along, say, well, can you create an event there? Can you create a point of information there so we can see what happens? And that means we can do things like A/B testing much more and understand what is the better way, or get some feedback if something is or isn’t working.”
Is net zero the next frontier for public sector AI?
The FSA’s Pierce told delegates AI can be used to fill gaps not currently met by existing technologies. She said her advice to public sector organisations embarking on AI projects was to “start small, with something that’s a pain point in the existing core business or start with something that nobody’s looked at before.”
She used the example of climate change. “Suddenly there’s this big thing on the horizon and there are not a lot of established systems and technology to deal with it,” she said. “You have to go and find those kind of use cases, engage with people and just get started.”
Boundy added that, regardless of size, it is important for departments to ensure AI deployments are “strategic”. “If you come up with something which is just on the borders of the business, that nobody’s that particularly interested in, it won’t be any great shakes,” he said. “If you’ve got something which is genuinely improving the outcome that your organisation delivers, and part of something strategic, then there’s a much greater chance of success.”
Homepage image by dusanpetkovic / iStock.