AI projects for the UK welfare system were canceled because officials said they had too many “false starts.”
Ministers have canceled or stopped at least six artificial intelligence (AI) projects aimed at improving the welfare system, according to the Guardian. This shows the difficulties Keir Starmer faces in trying to make the government more efficient.
Tests of AI technology meant to help staff training, improve jobcentre services, speed up disability benefit payments, and update communication systems are not moving forward, as revealed by freedom of information (FoI) requests.
Officials have admitted that making AI systems “scalable, reliable, and thoroughly tested” is a big challenge and there have been many “frustrations and false starts.”
While not every trial was expected to become a regular tool, two of the projects now canceled were mentioned in the Department for Work and Pensions (DWP) annual report as successful tests of AI concepts.
A-cubed was meant to help staff guide jobseekers into finding work. Aigent was supposed to speed up personal independence payments, which many people with disabilities depend on.
This month, the prime minister said, “AI is the way… to change our public services” and told all cabinet ministers to focus on using AI more and making it a top priority for their departments.
Imogen Parker, from the Ada Lovelace Institute, a research group focused on AI and data, said, “Failed projects don’t always mean there’s a problem, as they can be used to improve things. But these failures raise important questions about how the government is using AI in public services. Are they learning the right lessons, and does AI really work as well as they say?”
So far, no information about AI in the welfare system has been shared on the government’s algorithm transparency register, even though it has been required for almost a year.
Officials say the time spent on testing the pilot software isn’t wasted because the technology might be used in a system later on, and careful testing is important before launching any system. However, this shows how complicated it is for Labour’s plan to use AI to improve public services and boost the economy.
This week, Peter Kyle, the Secretary of State for Science, Innovation, and Technology, announced a plan for a “modern digital government” and said his department will use AI to help speed up progress on their plans, improve people’s lives, and support economic growth.
In December, Laura Gilbert, the director of the Whitehall AI incubator, wrote that there have been many challenges, delays, and false starts, but they keep trying until they find the right solution. She mentioned that out of 57 ideas tested, 11 are now being developed further. She also said they have been working with US AI companies like OpenAI, Anthropic, Google, and Microsoft.
In a private meeting in August, officials from the Department for Work and Pensions (DWP) told tech companies that the main challenges in moving AI systems from early tests to full use are making sure the products are scalable, reliable, and thoroughly tested. According to meeting notes released through a Freedom of Information request, about 9 AI tests have been completed, and one is now live, while another is about to go live.
Parker commented, “It’s good that the public sector is not being too strict or rigid with AI, especially in welfare, where using AI wrongly could increase inequalities or cause harm.” However, he also pointed out that the lack of transparency is a major issue. He said the government needs to be more open, evaluate results, and learn from them instead of keeping things hidden until journalists investigate.
The DWP didn’t explain why some AI projects were dropped but said that factors like technology readiness, business needs, and cost are considered. They test how much value the technology provides to both officials and the public.
A government spokesperson said that “proof of concept” projects are short-term and meant to explore and test new technologies. Not all projects are expected to become long-term, but the lessons learned from them will help in the future. This approach is part of their AI strategy, which aims to use AI to improve public services and save taxpayers money.
Published: 27th January 2025
Also Read:
The Question That Changed My Life: 7 Authors Reflect
UK permanent job openings decline at fastest rate in 4 years
Coffee boosts health, but only if consumed in the morning
