Before You Put AI in Government, Maybe Put Down the Spreadsheet

Not long ago, I found myself at a workshop with one of the UK’s big political parties. The topic? Whether rolling out more artificial intelligence in government would make public services better… or worse.

The discussion was prompted by a recent YouGov poll, which showed Brits are… well, let’s say “politely sceptical” about the whole thing. (Especially those who’ve been around long enough to remember when the fax machine was considered cutting-edge.)

Around the table, there was no shortage of big opinions. We hopped between the UK’s position in the “global AI arms race,” the ethics of machine decision-making, how it might reshape the workforce, and much more. It was smart, passionate, and thought-provoking — but mostly very high-level. The kind of debate that sounds fantastic in a panel discussion or on a podcast, but if you’re actually running a government department or a small business, you might be wondering, “OK… but what do I do on Monday morning?”

The conversation often drifted into “big picture” territory: international competition, sweeping economic trends, and how we might measure up against countries that have been investing in AI for decades. All important topics, sure — but a bit like discussing the finer points of Formula 1 engineering when your own car is still up on blocks in the driveway.

I sat back, listening carefully, and waited until the end to share my thoughts. Partly because I wanted to hear everyone else first… and partly because speaking last has the sneaky advantage of sticking in people’s minds.

When my turn came, I spoke from my experience working with both government departments and private companies of all shapes and sizes. And here’s the reality: most organisations aren’t ready for AI to swoop in and magically solve everything. They’re not stuck in the paper-and-ink dark ages, but they are still running major processes in Excel. And don’t get me wrong — I love a good spreadsheet. It’s the Swiss Army knife of office life. But it’s also a bit like running your national transport system on a fleet of bicycles: versatile, yes, but it’s not going to get you into the AI fast lane.

The truth is, before AI can deliver real value, you have to get the basics right: organise your data, streamline your processes, and get your workflows humming. And often, once you do that, people feel like they’ve already implemented AI — because the leap in productivity and efficiency is so dramatic.

It’s only once you’ve built that foundation that AI becomes the game-changer everyone talks about. That’s when you move from simple automation to predictions, smarter decision-making, and maybe even a dash of autonomy. But if you try to leap straight to AI without the groundwork, you’re effectively asking a toddler to sprint a marathon. Before you can run, you really do have to walk.

By the end of the session, the MP chairing the event summed up the discussion, and I couldn’t help but notice a few of my points made the cut. Was it because I spoke last and benefitted from the memory effect? Or was it because I made the most sense in the room? Well… I’ll leave you to decide. But I know which answer I’m going with.