A poker-playing robot may help find the answers to some of the most intractable challenges in the business world, such as optimizing e-commerce and auction applications.

Programmers have historically tried to teach computers to play chess, setting up the iconic 1996 match between IBM’s Deep Blue computer and human champion Gary Kasparov. But poker provides a better test of artificial intelligence (AI), says Tuomas Sandholm, a professor at Carnegie Mellon University: While chess players can see all the game pieces, poker players face many hidden details, like what cards the opponent has been dealt.

Sandholm specializes in the correlation between strategic behavior and computational complexity, and runs his own company, CombineNet, which develops algorithms to optimize procurement for companies like H.J. Heinz.

He put his latest technology to the test in July when he brought his poker-playing robot, called GS2, to compete in a Boston tournament sponsored by the American Association for Artificial Intelligence. What’s the greatest strength of GS2’s technology? About 26 million possibilities exist for poker hands in the second round of a Texas Hold ’Em game—too many for a computer to analyze: GS2 uses an algorithm that reduces the number of possibilities, to consider just 2,465 strategically similar hands.

GS2 has not beaten a top human player yet, but Sandholm continues to refine.

“The research problem is how to come up automatically with better and better strategies,” Sandholm says. “The computational and combinatorial complexity of solving the game makes this enormously challenging.”