This is a re-publish of a thing that went on LinkedIn, my newsletter, and the Digital Leaders newsletter. I’ve backdated the published date on this post to reflect this.

Summary: all this tech called ‘AI’ is genuinely exciting. But the impact of it is unlikely to be felt for several years. Don’t expect quick results, and don’t expect them to come without a hell of a lot of hard, boring work first.

It’s hard to look at LinkedIn these days without being instantly confronted by AI enthusiasts, almost foaming at the mouth as they share their vision for how the public sector can save millions, if not billions, of pounds by simply using AI.

It sounds so easy! As a chief executive I would be reading this stuff and thinking to myself, ‘why the hell aren’t my people doing this already?’.

In fact, I am hearing from digital and technology practitioners in councils all over the country saying that this is happening. That the AI hype is putting pressure on teams to start delivering on some of these promises, and to do so quickly. I find this troubling.

It’s always worth referring to my 5 statements of the bleedin’ obvious when it comes to technology in organisations:

1. If something sounds like a silver bullet, it probably isn’t one
2. You can’t build new things on shaky, or non-existent, foundations
3. There are no short cuts through taking the time to properly learn, understand and plan
4. There’s no such thing as a free lunch – investment is always necessary at some point and it’s always best to spend sooner, thoughtfully, rather than later, in a panic
5. Don’t go big early in terms of your expectations: start small, learn what works and scale up from that

How does this apply to using AI in public services? Here’s my take on the whole thing. Feel free to share it with people in your organisation, especially if you think they may have been spending a little too long at the Kool Aid tap:

• The various technologies referred to as ‘AI’ have huge potential, but nobody really understand what that looks like right now
• Almost all the actual, working use cases at the moment are neat productivity hacks, that make life mostly easier but don’t deliver substantial change or indeed benefits
• Before we can come close to understanding how these technologies can be used at scale, we need to experiment and innovate in small, controlled trials and learn from what works and what doesn’t
• Taking the use of these technologies outside of handy productivity hacks and into the genuinely transformative change arena will involve a hell of a lot of housekeeping to be done first: accessing and cleaning up data, being a big one. Ensuring other sources for the technology to learn from is of sufficient quality (such as web page content, etc) is another. Bringing enough people up to the level of confidence and capability needed to execute this work at scale, for three – and there’s a lot more.
• The environmental impact of these technologies is huge, and many organisations going ham on AI also happen to have declared climate emergencies! How is that square being circled? (Spoiler – it isn’t.)
• The choice of AI technology partner is incredibly important and significant market testing will be required before operating at scale. There’s an easy option on the market that is picking up a lot of traction right now, because it’s just there. This is not a good reason to use a certain technology provider. Organisations must be very wary of becoming addicted to a service that could see prices rocket overnight. More importantly perhaps is whether you can trust a supplier, or those that supply bits of tech to them, to always do the right thing with your data. There’s always going to be an element of risk here: but at least identify it, and manage it.
• Lastly, the quality of the outputs of these things cannot be taken on trust, and have to be checked for bias, inaccuracies and general standards. Organisations need to have an approach to ensuring checks and balances are in place, otherwise all manner of risks come into play, from the embarrassing to the potentially life-threatening.

This ended up being a lot longer than I first imagined. But I guess that just shows that this is a complex topics with a whole host of things that need to be considered.

Just remember – any messages you see claiming that AI is a technology that takes hard work away for minimal investment or effort, is at best just guesswork and at worst an outright lie.

Related to this post is a set of slides I presented to a conference in Glasgow: