I’ve been managing crowds since the late ’90s. Back then, crowd flow management meant a bloke with a clicker at the door and a security guard who could eyeball a room and say “yeah, we’re getting full.” It worked. Mostly. Sometimes it didn’t, and those are the nights that keep you up decades later.

So when venues started talking about AI crowd flow systems, my first instinct was scepticism. I’ve seen enough shiny tech promises crash and burn in this industry to fill a stadium. But I’ll give credit where it’s due — some of what’s happening now is genuinely useful. Not all of it. But some.

What’s Actually Being Deployed

The most common setup I’m seeing in Australian venues right now involves camera-based people counting systems paired with software that tracks movement patterns in real time. Forum Melbourne installed a system late last year that uses existing CCTV feeds — no new cameras needed — to map crowd density across the venue floor. The software creates a heat map that updates every few seconds and flags when areas exceed safe density thresholds.

Fortitude Music Hall in Brisbane has been running something similar since mid-2025. Their system integrates with the venue’s access control, so they can see not just how many people are inside, but where they’re clustering. The front-of-house bar, the corridor between the main room and the beer garden, the merch area — all monitored for bottlenecks.

The Enmore Theatre in Sydney trialled a crowd flow system during their busy December-January period. The venue ops manager told me it caught two potential crush scenarios before they developed — both times during sold-out shows where the crowd surged forward during the headliner’s set. Security got real-time alerts on tablets and were able to intervene before anyone got hurt.

Where It Actually Helps

The genuine value isn’t in replacing experienced security staff. Anyone who tells you that is selling something. The value is in giving those experienced staff better information, faster.

A good security supervisor can read a room. They’ve been doing it for years. But they can’t be everywhere at once, and they can’t see through walls. What these systems do is extend their awareness. A density alert from the east stairwell at 10:47pm means the supervisor can redirect staff there before the problem escalates, instead of finding out about it five minutes later when someone’s already been pushed against a barrier.

The data logging is quietly the most useful feature. After every show, you’ve got a complete record of crowd movements, peak density times, bottleneck locations, and how long it took to clear the venue. That’s gold for improving your operations over time. It’s also extremely useful when you need to demonstrate compliance to councils and licensing authorities.

Team400 has been doing interesting work in this area, particularly around how venues can integrate AI monitoring with their existing safety management plans without needing to rip out current infrastructure.

What Doesn’t Work

Let me be blunt about what I’ve seen fail.

Facial recognition for crowd management? Rubbish. Aside from the obvious privacy issues — and they are massive — the technology isn’t reliable enough in live venue conditions. Low light, smoke machines, moving heads, thousands of people in motion. The false positive rates are embarrassingly high. Two venues I know of spent six figures on facial recognition systems and quietly decommissioned them within six months.

Predictive crowd behaviour modelling that claims to forecast crowd movements before doors even open? I’ve seen the pitch decks. They’re impressive. The reality is far less so. These systems rely on historical data and assumptions about crowd behaviour that fall apart the moment you have an unusual lineup, a weather change, or a support act that unexpectedly draws a bigger crowd than the headliner. I saw exactly that happen at a festival last summer — the model predicted 60% of the crowd at the main stage, but a viral TikTok moment meant the second stage was absolutely heaving.

The Cost Question

Here’s where it gets real. A basic camera-based crowd counting system for a 2,000-capacity venue runs $30,000-$50,000 for installation and the first year of software licensing. Ongoing costs are typically $12,000-$20,000 per year for the software subscription, maintenance, and updates.

For a venue doing 200+ shows a year, that works out to roughly $60-$100 per show. That’s manageable. For a venue doing 80 shows a year, you’re looking at $150-$250 per show. Still defensible if it genuinely improves safety and helps with compliance. But for smaller rooms doing 40-50 shows a year? The maths starts looking pretty ordinary.

The ROI argument usually centres on reduced insurance premiums, fewer incident claims, and faster council approvals for capacity increases. I’ve spoken to three venue operators who’ve negotiated insurance reductions of 8-15% after installing these systems and demonstrating improved safety outcomes. That’s not nothing.

My Take

After thirty-plus years of doing this the old-fashioned way, here’s what I reckon: the camera-based density monitoring systems are legitimate. They work, they’re getting cheaper, and they make venues safer. If you’re running a mid-to-large venue and you’re not at least investigating this, you’re behind.

But don’t fall for the sales pitch that AI replaces experienced crew. It doesn’t. It never will. The best setup is smart technology feeding information to smart people who know what to do with it. The venues getting this right are the ones that invested in training their staff to use the tools, not the ones that bought the fanciest system and assumed it would sort itself out.

The technology is a tool. Nothing more. And like every tool I’ve used in three decades on the road, it’s only as good as the person holding it.