AI Dealer Beats Humans 10-to-1, and It Killed a Crypto Casino

Nigel Eccles says his AI blackjack dealer proved 10 times more popular with players than its human equivalent — popular enough that he shut down his crypto casino to sell the dealer to other gaming operators instead.

Eccles co-founded FanDuel and then built BetHog, a crypto casino operating in what he describes as a super competitive industry. The team developed an AI blackjack dealer in-house, one that banters and remembers players. It became so popular that the company was better served by closing the casino entirely and selling the dealer as a B2B product to other operators.

The economics driving that decision are stark. Live dealing is a huge operational cost for casinos, yet it stays popular with users. Eccles puts the scale of the human side of the business in concrete terms.

"Live dealers are a $15 billion industry. There's probably about somewhere around 30 or 40,000 people who work as live dealers," he said. The player experience, in his assessment, "just isn't great, probably 70, 80% of the time."

His forecast is blunt: "Within three to five years, I think 80% of live dealers will be AI-powered."

Eccles also argues that automation grows the market rather than shrinking it. "I actually think we are going to dramatically grow the market because there's this sort of whole group of people that would play a live dealer, but real humans are a little bit scary. This is the game they can play," he said.

The comparison he draws is with video games. "Very typically, a lot of video games you can play against computer, or you can play against other people. Playing against computers is actually a lot more popular, right?" Eccles said. "We found that AI Live Dealer is 10 times more popular than its Live Dealer equivalent. So people just find it less scary."

Why Are Some Founders Selling the Exact Opposite?

At the same time, founders across gaming and finance are building and selling the reverse proposition: verified, live humanity. Examples include a prediction market on what a real streamer does next, and a paper trading app where retention comes from other people.

The split defines two winning camps. Customers will either want more efficiency and customization, in which case AI becomes the product, or they will want more authenticity, in which case the humanity itself becomes the product. Which use case prevails remains unclear.

What Happens When the Human Is a Cost Center?

Eccles automated dealers because players were not actually buying the human. They were buying the game. When the person is the operational expense — as with live dealers — AI can replace them with something cheaper, less intimidating and more personalized.

The caveat, per Eccles, is that the threat is not to the best. The best online dealers have demand and always will. The threat targets the not-so-good ones, the less efficient operators, because AI will deliver a more personalized experience than they can. At the very high end of most categories, people will still prefer a human experience.

What Happens When the Human Is the Peer?

Paper trading is a way to learn about finance and the stock market, but doing it alone is boring, tedious and easy to fake your way through. A social paper trading app has succeeded by building a simulated environment that still contains real social and human interaction — people learning from each other without the pitfall of unlimited losses.

A finance creator who partners with the app made the same point about the fake money. "The problem is when you do paper trading, a lot of people don't take it seriously," he said. What makes it work is treating it like the real, social activity it imitates: "You should have a strategy when it comes to paper trading even though it, like, seems like it's counterintuitive."

Here the humans are peers. The product would collapse without them.

What Happens When the Human Is the Product Itself?

A Gen Z founder identified as Rama runs a prediction market on live streams, with markets that settle in about a minute. The betting is not on final outcomes.

"Instead of trading on the final thing happening in the stream, like, will you win the match? You're trading on everything happening throughout the stream," he said. "How many grenades will he pick up in the next 30 seconds? How many kills will he get in the next 60 seconds?"

His numbers, all from the first 45 days after launch: "We've crossed $10 million in trading volume. We have 42,000 registered users. This is all in the last 45 days of launching."

The launch video that pulled in millions of views came from the human side of the platform, not the esports side. "We had one of our streamers approach 50 women in the streets of Miami, and you were able to trade on each of those interactions — would he get the number or not?" Rama said.

In his experience, demand follows the humanity. "It's more entertaining. And cooler. To trade on an IRL stream or your favorite streamer stream."

Where Should Founders Place Their Bets?

The framework cuts across both camps. Founders should ask where the human sits in their product and what happens in that moment:

• Is it a moment of delight, or a moment of friction?
• Does the human add value as a peer?
• Does the human add value as competition?
• Is the human a cost center the customer never asked for?

If the human is the cost center, like a live dealer, AI can replace them with a cheaper, less intimidating, more personalized experience. If the human is the peer or the product, like in paper trading or streamer prediction markets, humanity is the draw and is harder to replace.

Eccles's three-to-five-year window for 80% of live dealers going AI-powered sets the clock. Founders on both sides of the divide are already monetizing the answer — BetHog pivoted to B2B, and Rama's platform cleared $10 million in volume in 45 days.