Have you been thinking about AI in the context of your scrutiny work at all? Conversations are certainly starting to happen - so here are a couple of things that it might be useful to reflect on.

I see that as councils are beginning to form policies, so some scrutiny committees are starting to receive reports on AI. And I wonder if the first in-depth inquiries are just around the corner - perhaps they are happening already?

And, from a scrutiny perspective, there are so many dimensions of this to look at. You have the ethics, the environmental costs, data ownership, data security, unauthorised use and workforce issues. To name just six.

Looking just at how scrutiny works, however, there are a couple of things that I’ve been thinking about.

First, and this came up during a conversation at the Centre for Governance and Scrutiny conference this year, is the idea of AI as a ‘junior scrutiny assistant’.

The AI assistant can do many things such as produce summaries of reports, transcribe meetings, analyse data or undertake research. No doubt they have other abilities and new ones will emerge.

But it’s a junior assistant because its work needs to be carefully managed and checked.

I spoke to Phil Rumens at West Berkshire Council about this. Phil is involved in a pilot to use AI for meeting transcriptions right across the council. He told me that, just as we are training AI to be more human, we shouldn’t be surprised when it makes mistakes.

And bias will also be an issue. Not least because the information that AI draws on will most likely also contain biases.

“Ask it to show its working”, suggests Phil. “Get it to cite its sources and say which articles of the constitution it is referring to, for example”.

Yes, supervising your junior scrutiny assistant requires skill and the right training.

A second idea comes from something that Catherine Howe said during her keynote at the CfGS conference. She talked about AI ‘saving time for the human work’ and what that human work might involve. This really struck a chord because it raises what is, perhaps, the most interesting question raised by AI:

What exactly is the ‘human work’ of scrutiny? In other words, what is it that only humans can do, that adds value and makes scrutiny meaningful?

I mean, we can talk about the skilled questioning, the trusted relationships and the connections with residents for starters.

Yes, your ‘junior scrutiny assistant’ can suggest some questions but it’s your scrutiny human that decides what to ask, adds details that they have picked up from other conversations, frames things in the right way and carefully follows up.

So, we shouldn’t lose sight of the human skills in our conversations about AI - perhaps that should be the first thing we talk about.

​

​

By the way, I've also posted this on LinkedIn. You can read / add any comments here.