Nine years of running AI agents across hundreds of venues has taught me one thing: the technology is never the hard part.
Every week, I sit across from a stadium executive holding a list. Twenty items, sometimes thirty, are on their AI wish list. While the list is always ambitious, it’s almost never organized around what the venue actually needs.
I’ve been building AI agents for sports and entertainment venues since 2016. We have roughly 2,500 of these agents running across more than 300 venues, handling tens of millions of fan interactions a year. So, when I say the list isn’t the problem, it’s because I’ve seen it all before.
The issue is the framing.
Start with problems, not projects
My advice to executives always starts with encouraging them to set the wish list aside and write down their ten biggest operational cost drivers and ten biggest missed revenue opportunities per event. Then ask whether AI has a legitimate application for any of them. That’s a strategy you can execute.
Instead of asking “Where can we plug in AI?” owners should ask, “If I could scale my best employee, what would I have them do?” The answers become the objectives and key results (OKR) that define what each AI agent should accomplish. We apply the OKR methodology to every agent we deploy. Each one is assigned a clear goal, such as driving ticket revenue, optimizing parking yield, or increasing food and beverage conversion.
Let’s look at driving ticket revenue as a goal. We analyzed more than half a million venue conversations found that 19 percent already contain explicit purchase language. Guests who use words like “buy,” “book,” and “reserve” represent a revenue opportunity. Nearly 28 percent of conversations begin with event discovery, and 18 percent of those convert to ticket intent in the same thread. Your AI needs to capture that transition and turn it into revenue.
Hold it to a payroll standard
The frame I use internally is simple: would you put this agent on your payroll? What job does it hold? What does it produce on a Tuesday in the third quarter when nobody is watching? How do you know if it’s performing?
An agent is only as useful as what it can actually do. One that answers questions is a glorified FAQ page. Agents who can complete a ticket purchase, check parking inventory, and escalate guests to premium experiences are making a direct impact on the bottom line.
We already know that fans are bundling purchases in a single conversation. Our data shows that when guests ask about tickets, parking and upgrade requests frequently surface in the same thread. Therefore, when a guest asks about tickets, that’s the moment to surface parking availability, not after checkout. The technology enabling these agent-to-agent integrations is in place at several tech-savvy venues. These are the types of metrics you’d use to evaluate any hire, and should be the same for your agents.
Beware the demo
There has never been a better time to build a convincing AI demo. A polished prototype can be assembled in days. While it will look finished and impress executives, it is not a finished product.
The difference between a demo and a production system is the last mile, that 20 percent where data engineering, privacy guardrails, model optimization, live integrations, and systems to protect the organization from failure kick in. When you evaluate vendors, look at their production track record under live event loads and how they handle the last 20 percent. Ask whether they have the required domain expertise and whether they have ever been a venue operator or an employee.
Make it a C-suite conversation
When you assign OKRs to AI agents, ownership moves up the org chart. The ticketing agent reports to revenue leadership. The food and beverage agent reports to concessions. AI stops being an IT side project managed by a junior contact and becomes a strategic initiative measured, adjusted, and forecast at the executive level. That’s the shift that separates venues running AI experiments from venues running AI as a business unit.
AI in a venue context is the highest-leverage employee you can put on the floor, as long as you treat it like one.
