Just about two years ago (Sept 2023), I wrote a piece called “ChatGPT as Slot Director”. I had asked my computer a simple operational question: “What is the best percentage of Participation or Lease games to have on the casino floor?”
Impressively, it spit out an answer in just seconds that was over five paragraphs long. However, the response was quite generic. My conclusion then was: “Based on the non-specific output to my question, it looks like this latest version of ChatGPT relative to the casino world is at the college freshman level with little or no real-world experience.”
However, the Chat version I used then was 3.5/4o. It was just emerging from Beta and was in its infancy. Please note that “months” in AI software development are like “years” in the real world. So, has anything changed in the last 20 digital years?
This time I used the latest (paid) ChatGPT-4o from Open AI and selected their new feature called “Deep research”. I also asked a much more complicated question:
“How can a locals’ casino in Southern California make its slot floor more profitable?”
This, of course, is a ridiculous question since it is way too broad. To Chat’s credit, before it began, it asked me for more information:
“To make sure I give you the most useful advice, could you clarify a few things?”
It asked for about a dozen things, but I responded only with the casino’s square footage, and that I wanted to increase net profitability. I also asked the program to focus on slot machine selection and layout optimization. This time my desktop took a bit longer (a seemingly forever 18 minutes!).
Along the way it consulted 28 internet sources and did 57 individual searches. Some of those references were our own CDC Gaming stories along with info from Casino Style, Eilers & Krejcik, SAS, UNLV, and others. Chat even included several of my own quotes and referenced my editorial nemesis, UNLV’s Dr. Anthony Lucas (who has never met a slot machine that he didn’t think could be made tighter).
Interestingly, Chat did not include anything from the many excellent books on the subject, since they are restricted by copyright protection.
So how did it do? It is definitely no longer a freshman. I’d say the 6,000-word essay it produced was MBA-level stuff. From my teaching background, I’d give this paper an A+ if it came from an undergrad. Likewise, a newer Slot Supervisor could study this material and make a pretty good impression in a job interview to advance their career. If you read the full output (linked below), you will note it also cited reference sources, as any good “white paper” would.
To save you some time, I’ll cite just a few of the report’s suggestions:
- Leverage analytics of actual play data to identify top-performing games and remove underperformers. Modern casinos use advanced forecasting that considers trends, seasonality, denomination, and even time-of-day usage to understand what drives a machine’s success. Rather than relying on gut feeling or historical popularity alone, data analysis can reveal which games resonate with the local market.
- Consider offering lower minimum bets and more incremental bet options on machines to appeal to players who budget their play. A casual local patron might prefer a game where they can bet just $0.50 or $1 at a time and play longer, rather than being forced into a $3 minimum bet on a penny video slot.
- Instead of long, endless rows, today’s floors use clusters, pods, and curving pathways to break up sightlines and invite customers to wander. This encourages a meandering flow where players naturally stroll through different sections of the floor, increasing the chances they’ll spot a machine that catches their interest.
- The success of Station Casinos underscores that in a locals’ market, reputation and player trust in slot fairness is crucial. By not pushing hold percentages to the max and constantly updating the gaming mix (they rotate new slots from all major manufacturers, often showcasing “Coming Soon” banks to create anticipation), they keep locals engaged and on property.
- Giving players enough personal space and comfort at the machines has proven to increase time-on-device and revenue. Pre-2020, many casinos packed machines tightly – it was common to see slim stools crammed together with barely a couple of feet between players, all in the name of maximizing units on the floor.
Those are just five of the dozens and dozens of suggestions which span 15 pages. (If you have nothing better to do for the next 25 minutes, you can read the entire output here). It would be rare to find anyone who would disagree with the concepts Chat cited. But, as any decent slot director knows, the devil is in the details.
Before the year 2000, those “details” came only from years of experience and a lot of costly trial and error. Then came powerful “information warehouses” and illuminating “data visualizations”. They improved things greatly.
But, the real Artificial Intelligence revolution has been gaining steam just recently. In 2017, GPT or Generative Pre-training Transformer technology began to emerge from Open AI and others. This was like having the “Lord of the Internet” at your beck and call and using a supercomputer to make sense of it all. We can now query slot machine performance against player demographics, engagement patterns, time-on-device, preferences, devotion and promotional effectiveness. Admittedly, we could always do that; but we didn’t because it would take weeks or months. Now our answers can come in minutes.
And it’s getting faster. This last decade was also known as the age of machine learning. This allowed these AI programs to do their own version of “trial and error”, constantly improving themselves (seemingly at lightspeed).
Make no mistake, ChatGPT-4o is not perfect. Its biggest shortcoming is that the bedrock of its research comes from the internet. That’s the same source that brought us the “Tide Pod Challenge” and “fake moon landings”. I could write a book about all the false and misleading garbage online, but Walter Scheirer from the University of Notre Dame beat me to it.
However, the improvement in the credibility and capability of AI and GPT technology in just a few years has been remarkable. Current AI-based analytic software in our casino industry is also improving at astonishing rates.
Competition ensures this trend will continue. Just check the GPT developments coming from all corners: Apple Intelligence, Chatsonic, Claude, CoPilot, DeepSeek R1, Ernie, Gemini (the former Bard), Jarvis (no, it is not just Ironman who uses this admin tool), Jasper, Llama-4, Perplexity.ai and dozens of others.
Some gaming regulators are now using these tools to craft new tech guidelines. CEOs are handing off administrative tasks so that they can focus more on their core mission. IT coders are becoming faster and more efficient. So, unless you are stressing about “Skynet,” it is an exciting and potentially rewarding time for those who welcome the future.
So, is ChatGPT now a better slot director? The answer is clearly “Yes,” but thankfully not to the point of yet having to replace us, despite our comparative slow and plodding nature. Along with that encouraging news, being aggressive and embracing the tools created by AI and GPT will make us better able to withstand the assault. Conversely, failing to keep up could mean that your potential replacement is coming much sooner than I anticipated in September of 2023. Check back soon.
# # #
Postscript #1: I asked if another innovative new program called Agentic AI working together with ChatGPT-4o might do even better as a slot director:
“Agentic AI would give ChatGPT the ability to take initiative, make decisions, and act independently—like a virtual assistant with real responsibility. As a slot director, it could continuously monitor floor performance, identify underperforming machines, and recommend real-time changes without waiting for human input. It could autonomously run test configurations, analyze player behavior, and adjust machine placement or themes to boost engagement and profit. Instead of just answering questions, ChatGPT would function like a tireless, data-driven slot manager—learning from results and acting proactively to optimize the casino floor every day. It becomes a partner, not just a tool.”
Postscript #2: Then I asked ChatGPT to produce a shorter version of the main article above. In less than two seconds, it spit out the following:
“In 2023, Buddy Frank tested ChatGPT-3.5 on a casino operations question and found its response too generic, likening its understanding to that of a college freshman. Two years later, using ChatGPT-4o with its advanced “Deep research” feature, he posed a far more complex question about maximizing profitability at a locals’ casino and received a highly detailed, MBA-level analysis backed by extensive research. While AI hasn’t yet replaced the slot director, Frank now sees it as a fast-evolving tool capable of generating actionable, data-driven insights that would have once taken weeks to compile.”
Clearly, GPT won’t be a hit with freelance writers who get paid by the word?
# # #
