Last week, I spoke on a panel at Microsoft’s GovTech Futures conference in Wellington. The event brought together IT professionals and public servants. I was probably invited because someone knew I was a tech geek working in policy. Or maybe they just liked my accent.

Anyway, it was fun. Not least because the first question caught me off guard. The moderator asked me if AI could have helped this year’s coalition negotiations. What a great question!

Rather flippantly, I replied that I am not certain all our political leaders even knew how to use a PC, let alone ChatGPT.

And indeed, the most high-tech item traditionally involved in coalition negotiations used to be a bottle of whiskey. Except with two teetotallers among the three leaders, that won’t work this time. Maybe that’s why agreement is taking so long.

Seriously, though, there is no reason that artificial intelligence (AI) should not play a bigger role in policy development – and not just during coalition talks.

For many years, and well before ChatGPT turned up on the radar late last year, I have been playing with various AI packages. Image generation, text-to-speech and voice recognition, large language models: all this techy stuff fascinates me.

For the past 12 months, I have felt like a kid in a candy shop. Hardly a day went by without a new AI tool being released somewhere – and many of them are free, at least to try.

Most people have hardly used AI yet, or at least not properly. Many of those who have will only have done the usual silly stuff. And yes, I have also made it write sonnets in Shakespeare’s style, had it draw a picture of Donald Duck as the Mona Lisa and compose a new Beatles song. (The latter was awful.)

But AI can do so much more - if you take the time to explore the possibilities.

There are two barriers to entering the world of AI. First, you need to know what tools are out there and what they can do. Second, you need to understand how the tools work and how to use them effectively.

As well as time, overcoming both barriers requires tech expertise. Most politicians are unlikely to have much of either.

I remember MPs discussing how to regulate Uber at a select committee hearing a few years back. One of them seriously asked how to recognise an Uber when you hail one. Ubers, of course, come in all shapes, colours and sizes (as long as they are a grey Toyota Prius), but they are unmarked – and you don’t hail them.

With politicians being notoriously time-poor and, often, not tech-savvy, there will be a premium for the first politicians embracing AI. They will have a superpower that will make them stand out from their colleagues – in their own parties and across the aisle.

For example, AI can deliver information more quickly than their parliamentary research assistants. They just should not try this with ChatGPT, which has been known to provide “alternative facts”.

There is a raft of specialised AI tools dedicated to answering research questions. One of them, Perplexity.ai, does not only answer complicated questions, it also provides links to websites to back up its claims. Better still, Perplexity.ai suggests intelligent follow-up questions that invite users to deepen their research.

For the more academically inclined politician, there is Elicit.org. Ask it any question, and it will give you not just a list of the most relevant academic papers with short summaries but also generate a summary of the top papers.

For example, when I asked it about the effects of mandatory water meters, it gave me detailed results from real-life case studies in Oklahoma, California and the Canary Islands.

Still, even with such specialised AI tools, don’t write off good-old ChatGPT just yet. Especially in its GPT-4 mode, it mimics real political creativity quite effectively.

I put ChatGPT to the test for one of my latest columns and asked it for suggestions on how to condense the 70-plus ministerial portfolios in New Zealand into more coherent and wider ministries. ChatGPT passed that test with flying colours.

I also entered the three potential coalition partners’ policies on tax cuts and foreign buyers, asking ChatGPT to think about potential compromises. It did that well, too.

I even asked it what Act and NZ First should do if they felt they were not being heard by National. ChatGPT advised the two small parties to team up and work jointly on areas in which there is agreement between them. And indeed, that seems to be what is happening in the real world right now.

If politicians discovered AI as a way of brainstorming ideas, they might be surprised at how creative their policies could be. And they could then let other tools evaluate their ideas based on an analysis of the relevant literature.

To avoid misunderstanding, I probably would not want to be governed by AI just yet. Even though being governed by ill-informed humans often is not much better.

So perhaps, just perhaps, AI could make our politics more interesting and creative.

At least I think so, and my small army of AI tools agrees with me.