Digital Welfare State edition 015
DWS Newsletter - edition 15
October 2026
———————————————
Things to read
I find this kind of thing so fascinating - this is how policy and practice travel the globe, why we see the same kind of systems, successes and failures pop up in seemingly quite different contexts - Somalia and Pakistan deepen cooperation on digital infrastructure.
——————————————————————————-
And here’s how a much bigger official partnership (the EU) uses digital tools to make things like social security portability easier for EU citizens, partly using the EU digital identity wallet.
——————————————————————————-
An update on DWP’s JobCentre in Your Pocket which includes an AI-powered chatbot, designed to help people find work. I had a play around with it a while back and wrote it up here.
——————————————————————————-
In other DWP news, they are building internal capabilities to use AI agents effectively. It’s going to be extremely interesting to see how AI agents change how citizens interact with digital welfare systems. What the take-up of agents will be, and what kind of tasks they are capable of (could one correctly make a claim for Universal Credit for example) is going to unfold in ways that we can’t yet predict. Tom Loosemore has written a bit about his thoughts.
This recent paper from the Administrative Fairness Lab highlights the existing use of generative AI tools by benefit claimants to help them with their claims and queries - adding agents into the mix might make things even less predictable.
This project from the Money Advice Liaison Group feels like the kind of thing which needs to be replicated for all sorts of groups with advice needs, including those applying for benefits. This survey from AdviceUK shows that lots of people are turning to tools like ChatGPT for advice on issues such as benefits, but a big majority of them also don’t think AI can replace human advisers, particularly on complex issues, and three-quarters of people are worried about getting incorrect advice.
——————————————————————————-
Interesting piece about the use of AI in workers’ rights legal advice and claims. There’s been a big upsurge in claims, thought to be in part due to people using AI to ask questions and draft complaints. The piece points out the pros and cons.
——————————————————————————-
A couple of reports on the use of AI transcription in public services. This, from the UK’s National Audit Office, says the Justice Department’s use of a tool called Minute should be held up as ‘exemplary’ and other government bodies should learn from it. A few people have pointed out that the actual evaluation of the tool was not particularly robust.
This hands-on playbook from the team at Careful Industries, developed in collaboration with Greater Manchester, includes 11 principles for good use of transcription. Perhaps read both if you’re looking for a well-rounded take!
——————————————————————————-
Well this is depressing. An AI tool used to create summaries of interviews between the Home Office and asylum seekers has been found to make a significant number of errors. In addition, people whose interviews are summarised are not told that the AI is being used. When the potential impacts are so consequential this is very troubling.
——————————————————————————-
An important report from the Public Law Project about the challenges of transparency and accountability when public bodies make automated decisions and use AI.
——————————————————————————-