How AI Is Entering Every Corner of Casino Operations: eConnect Version 11 Explained

Sixteen years. That is roughly how much operational data eConnect says it has been collecting from casino clients, and Version 11 of its platform is built to let a manager get at all of it by typing a question in ordinary English. The company counts more than 325 clients worldwide. If you want a single, concrete illustration of what AI in casino operations actually looks like in 2026, this is a good one: not a robot dealer, but a chat box that knows which bar sold the most wings last quarter and which players are on the floor right now.

Let me walk you through it the way it would land on a property, department by department, and flag the parts where the marketing language and the reality diverge.

What eConnect Version 11 brings to casino AI

eConnect Version 11 is the latest release of a casino data and facial recognition platform, scheduled to be formally unveiled at the Global Gaming Expo (G2E) 2026 in Las Vegas on September 30. The headline change is scope. Earlier versions lived mainly in the surveillance room. Version 11 opens years of data from every business segment of a property, not just surveillance, to anyone with the right permissions.

Two features carry the release:

• Ace, an AI assistant that answers questions about the property’s data and explains where its answer came from.
• Semantic search, which lets staff ask in plain language instead of building a report or learning a query language.

“It’s crazy, it’s actually so exciting,” CTO and co-inventor Travis Whidden said of an early deployment at one casino, describing how Ace “is diving in, explaining to the user how it got the data.”

That last part matters more than the chat interface. An assistant that shows its working can be checked. One that just produces a number cannot.

eConnect was founded in 2009, built its business on facial recognition software and analytics for casinos, and later moved into arenas and stadiums. Version 11 is designed to sit on top of what a property already runs, in the same way the facial recognition tools work with an operator’s existing cameras and software. Whidden’s framing: “Casinos have been our world, because we’re heavily tied into their surveillance environments. We’ve written an integration in every video system that was there.”

AI-powered casino surveillance and security

Start where the company started. Casino surveillance AI, in practice, is mostly two jobs: matching faces against a list, and connecting one event to another event.

Facial recognition compares a face captured on camera against stored images, self-exclusion lists, advantage-play watchlists, or trespass records, and raises an alert when the confidence score clears a threshold. The system does not decide anything. It narrows thousands of hours of video to a handful of moments a human should look at. Everything after the alert, including whether a match is real, stays with the surveillance operator, and that is exactly where it should stay given the false-positive risk in any biometric matching.

Version 11’s contribution is the second job. Because the same platform touches point-of-sale, gaming and other operational data, the questions a security director can ask get broader than video. One example from the company: asking which employees hand out the most cash-register discounts. That is not a surveillance question in the old sense. It is a loss-prevention question that used to require pulling a POS report, cross-referencing shift records, then finding the relevant video by hand. Semantic search collapses those steps into a sentence.

What the AI is doing here is not magic. It is data integration plus real-time monitoring plus a language model that knows how to turn “show me discount outliers by employee” into the right query. Useful, unglamorous, and genuinely time-saving.

Gaming floor operations and player analytics

Now move onto the floor. The company’s own example is a gaming director asking which of the most valuable players are on the property right now, and where to find them.

Unpack that request and you can see what the AI gaming tools are stitching together: player-account data on historical worth, live carded play from slots and tables, and location signals from the floor. Traditionally the answer comes from a host team watching a player-tracking dashboard, or from a floor supervisor who recognises a regular. The AI version is faster and less dependent on who happens to be working that shift.

The same plumbing supports the more ordinary work of slot and table management, because it is all the same kind of question asked with different filters. Which banks of machines are producing below expectation for their location. How table utilisation moves hour by hour against staffing. Where the gap between theoretical and actual hold is widening.

Two honest caveats. First, predictive analytics on a casino floor forecasts patterns of player behaviour and machine performance. It does not change a game’s mathematics. RTP and house edge are set by the game itself, and no amount of machine learning on the operator’s side alters the odds a player faces on a given spin. Second, sharper player analytics cuts both ways. The same behavioural signals that identify a valuable customer can identify someone chasing losses, and operators using these tools have an obvious responsibility to act on the second signal as readily as the first. If you are gambling yourself, deposit and loss limits, session reminders and self-exclusion remain the tools that actually protect you.

Back-office and administrative AI integration

This is where the release is genuinely broader than “surveillance software with a chatbot.” The company describes the same assistant fielding questions from departments that had no relationship with surveillance data at all.

Here is how the examples map to real day-to-day work:

That AML example is worth dwelling on. Transactions clustered just beneath a reporting threshold are a classic structuring pattern, and surfacing them is standard compliance work. Automating the query does not automate the judgement. A human still has to assess intent and file anything that needs filing, and regulators will hold the operator responsible either way.

Security design shaped the release. Ace runs on hardware located at the operator’s premises, so interactions with the database stay inside the building. “A lot of properties don’t want their private data leaving their facility,” Whidden said. “That’s why we have the on-prem hardware that allows them to keep it all private.” Access is compartmentalised, so a food and beverage executive cannot pull gaming floor transactions. And the assistant’s access to data is read-only, with no mutation permitted, though Whidden noted that restriction may loosen later once proper controls are in place.

The bigger picture: AI in casino operations

Strip away the branding and Version 11 fits a pattern visible across gaming technology right now. The interesting AI in casino operations is not replacing people, it is replacing the report queue. Every property already sits on years of data spread across a casino management system, a surveillance network, POS terminals, loyalty databases and accounting. The bottleneck was never collection. It was that answering a moderately specific question required someone who knew which system held the answer and how to ask it.

Conversational access removes that bottleneck, which is why this category of tool is spreading sideways into every department rather than deepening in one. It also explains the on-premises architecture. Operators handling biometric data, player accounts and AML records are not enthusiastic about shipping any of it to an external model, and vendors have noticed.

Where should a professional stay sceptical? Three places. Output quality is capped by integration quality, so a property with messy or partial data will get confident-sounding answers built on gaps. Language models can misinterpret a question, which makes Ace’s habit of explaining its sources a practical safeguard rather than a nice touch. And broad internal access to sensitive data raises governance questions that permissions alone do not settle, particularly around biometric information, which faces tightening rules in several jurisdictions.

None of that makes the direction wrong. It makes the next few years about operational discipline rather than capability. The technology to ask a casino a question and get a sourced answer now exists and is shipping. Whether the answer is trustworthy will depend on how carefully each property wires it up.

Frequently asked questions

What is eConnect Version 11?

It is the latest release of eConnect’s casino data platform, scheduled to be unveiled at G2E 2026 on September 30. It adds an AI assistant called Ace and semantic search, giving authorised staff plain-language access to years of data from across a property rather than surveillance data alone.

How does AI work in casinos?

Mostly through pattern recognition on existing data: matching faces against watchlists, flagging transaction or discount anomalies, forecasting machine and table performance, and now translating plain-English questions into database queries. Decisions stay with staff.

What casino operations use AI?

Surveillance and loss prevention, gaming floor and player analytics, food and beverage performance, staffing and scheduling, and compliance reporting including AML monitoring.

How does casino AI surveillance work?

Software analyses feeds from a property’s existing cameras and compares faces against stored lists, alerting an operator when a likely match appears. A human reviews and verifies every alert before anyone acts on it.