Revolutionizing Sales With A Poker Playing AI

 artificial-intelligence-sales-forecasting

Artificial Intelligence & Sales Forecasting

Here’s a scenario that might be familiar to you: You’re on the phone with the purchasing rep at a large vendor. They’re making buying signs, but the process is starting to drag on. Are you seriously in contention, or would your time be better spend on fresh leads? If an AI known as Libratus can successfully beat four professional poker players to the tune of $1.7 million in virtual chips, the same same algorithms can tell you if your leads are serious—or simply bluffing. In terms of B2B sales forecasting, Artificial Intelligence can act as a force multiplier—letting sales organizations achieve more closed business with the same amount of effort. Here is how it breaks down:

Intelligent game-playing AIs are nothing new. In fact, the first chess-playing computer dates back to 1956. Known as MANIAC, this machine was able to beat an amateur player in just 23 moves, albeit using a simplified version of the game. Beating an amateur is nothing like beating a professional, however, and winning at chess is nothing like winning at poker.  What is the importance of this AI milestone, and how might it change the future of business?

Winning at Poker is a Win for AI

Both chess and Go have recently been mastered by artificial intelligences—but poker is a different animal. In both of the board games mentioned above, both players can see the entire state of play. There are no hidden pieces on a chessboard.

In poker, the game is all about what players have hidden. This is a particular challenge for an AI, because solving problems based on incomplete information has, until recently, been beyond the capabilities of all but the most advanced computer programs. Just as an example, no-limit Texas Hold’em can contain variables that work out to 10160 possible moves. Furthermore, other players will intentionally hide their position by bluffing. For a computer, solving poker is the maximum test of processing power and machine learning ability.

Libratus irons out the uncertainty and complexity of poker in three ways:

  1. Algorithmic Strategy: For poker-playing computers, the optimal strategy is to simulate possible endings for a given hand, then calculate which ending represents the most optimal outcome. Finally the computer attempts to achieve this outcome via branching decision pathways. Libratus refines this strategy by pruning down the more obviously wrong decision trees, lessening the overall computational load.
  2. Punishing Mistakes: In general, AI players don’t beat humans—humans beat themselves. Human players, even grandmasters, constantly make mistakes which computers can see and exploit. Libratus has an improved ability to understand the extent to which humans’ mistakes have helped it out. It then calculates risk/benefit strategies accordingly.
  3. Enhanced Playing Ability: Libratus can bluff and deploy random moves that stymie human players. This is a constant feature of advanced AI—when Google’s AlphaGo defeated grandmasters in China, the players reported seeing strategies that had never been contemplated in thousands of years of competitive play.

Winning at Poker is Also a Win for Businesses

Before you start to worry, just know that AI is probably never going to replace human-controlled business processes—or poker players, for that matter. What it will do, most likely, is make workers more efficient and more productive. As just one example, the players facing down the Libratus AI reported that playing against the AI made them better players. Since the AI would regularly make massive bets to win small pots, it forced each human player to play their most proficient game during every hand.

Imagine working with an AI that helps your sales leaders become the best managers and sales people that they could possibly be. Request a demo of SpringML to see how our application can rank opportunities on their likelihood of closing, create detailed and accurate sales forecasts, and leverage these insights into a plan for scale.