In my first two articles, I discussed experimenting with AI thoughtfully and understanding a process before deciding where it fits. Now, I’m interested in what we might discover by doing this. Often, practical AI applications are straightforward. They might help someone find information, flag a possible duplicate, or direct a request to the right person.

A task might take only a few minutes but happen a dozen of times a day. Making it easier could have a real impact.

How AI can help council staff find information

Councils came to mind when I reflected on the Public Sector Data & AI Summit. Imagine a caseworker trying to help a resident, but different teams hold the information they need across separate records. They might know the answer exists somewhere, but still have to search for it. Then they need to check that it relates to the right person and work out whether it is up to date.

AI functionality could help by finding the right information in those records and bringing it together for review. The caseworker would still need to check what the tool found and decide what to do next. The useful part would be getting to the relevant material faster, especially when they have to search through notes.

This is only one example, but councils are exploring related problems. MHCLG’s Local AI team, set up this year to help councils use AI responsibly, has noticed that frontline staff spend a lot of time managing information across different systems and re-entering data. The Local Government Association has also published a case study of North Yorkshire Council’s work. It explores using AI to help children’s social workers find information across their case management records. The study talks about the heavy admin workload that led to this project. It was only a proof of concept, so not every council will get the same results, but it’s an area worth exploring.

Other practical AI applications

Finding information is only one of the more obvious possibilities. A tool could flag records about the same person to speed up checking them. It might suggest which team is best placed to answer a resident’s query, with staff able to correct it when needed.

In each case, I’m interested in whether it saves people from repeating searches or sorting through information all day. What could people do with that time instead? If a caseworker can find the right record sooner, they may have more time to listen to the resident and work through what they need. That’s a benefit we should test, not just assume. However, it’s a better way to judge an application than whether the technology looks impressive.

This idea was also behind the NHS example I mentioned before: if we reduce the admin, clinicians could spend more time with patients. It’s a possibility worth exploring, but we still need accurate records and professional judgement.

Exploring Practical AI applications at Datagraphic

At Datagraphic, we work with document communications, data, and the processes behind them. That makes these practical uses interesting to us. I’d rather see if a tool actually helps someone find what they need or handle a request well, instead of choosing something just because it looks good in a demo.

Across these three AI blogs, I’ve argued for experimenting with care and being clear about the problem we want to solve. This is where I think that approach can lead: to changes that may look small from the outside but make a working day easier, or help someone get an answer without being passed from one place to another. That would be useful progress.

Missed any of the previous AI blogs in this series?

Read the first blog.

Read the second blog.