On 18 October, the Woden Town parkrun in Canberra joined the national family of parkruns – free, weekly, timed 5km events that take place across Australia every Saturday morning. We were confident we had chosen a great location, but as with any new public event, there was one major unknown: How many people were going to show up?
Our first event drew 102 participants. A good start. But the next week, attendance surged to 731. We had congestion before people even crossed the line, we ran out of finish tokens, and volunteers had to write numbers by hand to ensure every result was captured. Operationally, it “worked” – but only just. Unintentionally, we had set a new ACT attendance record.
After that, planning became guesswork. Would turnout stay above 700? Drop to 500? Slide to 250? We had to make decisions about course design and volunteer staffing without any precedent. We needed something better than intuition.
I turned to a widely available generative AI tool – not as a gimmick, but as operational support. Instead of simply asking it to predict numbers, I provided context: attendance patterns at established ACT parkruns, how catchment areas affect participation, the impact of weather, the influence of less busy long weekends and competing local running events.
The tool responded with clarifying questions and structured modelling. It behaved less like a conversational novelty and more like a junior analyst working alongside me.
By week 4, we could begin to see a trend emerging. The AI recognised that the second week had been inflated by novelty and curiosity – a rare surge any new event can receive when everyone wants to “try the new one”. But rather than assuming interest would collapse, the model suggested that Woden was likely to stabilise toward the upper end of the ACT attendance range. That insight mattered: it told us not to expect a quiet fade, but to prepare for sustained high volume.
As we moved forward, we began to test forecast against reality. The model made predictions that frequently landed close to actual turnout and, more importantly, its reasoning was visible.
When a large running festival occurred in Canberra, the model anticipated a dip in attendance and we saw exactly that pattern play out. When rain was forecast, it adjusted expected turnout downward – drawing on patterns it had detected at other events affected by wet weather.
That process developed into a rhythm: each week, I provided actual turnout, weather outcomes and contextual variables; the AI refined its estimates and updated its assumptions. Over time, this back-and-forth helped us anticipate demand with much greater confidence. It didn’t just give us numbers – it helped us understand why those numbers might move one way or another.
This improved our ability to plan, particularly how many volunteers to roster and how to manage peak flow at the finish. Instead of under- or over-preparing, we were able to size our operations intelligently, based on structured reasoning rather than hope.
This was not about outsourcing judgement. The decisions remained fundamentally human. They were rooted in our local knowledge and situational awareness. But the AI supported this judgement by offering a disciplined framework for interpreting trends.
Throughout this process I found myself impressed by how grounded the tool was. Artificial intelligence is often framed at extremes – either as miracle or menace. My experience here was neither. It wasn’t glamorous, it wasn’t dramatic, it simply added clarity, consistency and rigour to our thinking. It reduced uncertainty. It helped translate scattered data points into actionable insight.
If you work in event operations, community programming, public engagement or any area where participation forecasting matters, I would encourage you to think of generative AI as a thinking partner rather than a decision-maker. Used thoughtfully, it can provide analytical support that strengthens your planning without supplanting your judgement.
At no point did the AI run the Woden Town parkrun. But it helped us prepare for it – and that made us better at what we were already trying to do. That is the role technology should play: enabling people, not replacing them.
Andrew Dempster is a former senior adviser to the Albanese Government and an avid recreational runner.
Do you know more? Contact James Riley via Email.