Major League Baseball umpires missed 34,294 ball-strike calls during the 2018 baseball season. After studying a total of 4 million pitches through 11 seasons between 2008 and 2018, researchers determined that umpires incorrectly call 14 pitches per game—an approximate error rate of 5% in every Major League Baseball game!
It’s stats like these that have the Atlantic League, an alternative professional baseball league, testing a sort of robot umpire that calls balls and strikes. A human umpire will signal the decision made by TrackMan technology. This implementation of the Internet of Things (IoT) automation is part aimed at ensuring the right ball or strike call is made, and part at speeding up the pace of the game to keep up with the changing demands of fans.
It’s a huge jump from what was considered cutting edge when data-driven “Moneyball” tactics started to proliferate across sports more than a decade ago. Today, most teams are leveraging technology to supplement and, in some cases, replace the gut-driven decisions that guided the game for decades. The analytics that were so bleeding edge when employed by the Oakland Athletics some 15 years ago are now available to (and employed by) dads coaching town Little League teams.
That’s a familiar scenario in the enterprise, where having the access to historical data no longer provides the clear business advantage it once did. Businesses are aware of the advantages that come from looking at data to identify patterns and tendencies to reliably project what will happen. AI in business can further refine these types of analyses through enhanced reasoning and systems that self-correct based on new data and changes to the environment being studied. But understanding how and why to deploy artificial intelligence and machine learning can be challenging.
As it strives simultaneously for accuracy for the sake of the game and to evolve the business of the game, baseball’s early applications of artificial intelligence offer many lessons for businesses on making the case for, and getting started with the technology.
Analytics help humans make decisions based on historical data. AI is the simulation of human intelligence processes by machines. Systems use machine learning technology to analyze current conditions, learn from experience and either help the human make the decision or “make” the decision on its own.
Baseball provides a good analogy for considering the differences between analytics and AI. The drive to take some of the guesswork out of baseball starts at its core: pitching. And the aim is both to win games and to minimize pitching changes that lengthen those games, to keep fans engaged.
Analytics tools like TrackMan and Rapsodo gather and analyze 27 or more different data points per pitch, including pitch velocity, spin rate (of the ball), speed, the pitcher’s arm angle, the pitch release point and the distance the ball travels. But those data points only provide an historical reference point.
Using AI, pitching coaches can run simulations of all 162 games, blending the historical information with current and projected future conditions, with the ability to automatically update the analysis to predict how different scenarios play out. This allows teams to model what would happen if certain pitches were thrown to certain hitters, or what would happen when pitching changes were made, enhancing the accuracy of predicting the result. In other words, instead of running the models based on what happened in the past, teams can now look to the future, running some 10,000 simulations between one pitcher and one batter and viewing predictive results based on variables such as field type (grass vs. artificial), time of day, place of the batter in the batting lineup, and whether runners are on base.
“We’re simplifying the data we have to the point of actionability,” said Jim Brower, a former major league pitcher who now serves as a pitching coach for the Seattle Mariners. “Even though the game is starting to look more like checkers. You think you’re just seeing this guy throwing as hard as he can, [but] there’s so much more going on in the background.”
As in baseball, business technology that leverages artificial intelligence provides a distinct advantage over the analysis of historical data alone. Analyzing real-time conditions can optimize business processes.
Consider Zamboni, the company that is known for re-surfacing ice rinks in between periods at hockey games. The company is leveraging the internet of things (IoT) by equipping its machines with devices that capture and unlock the data produced by the Zamboni to improve operational efficiency through predictive maintenance. This can decrease unplanned downtime and optimize asset availability and efficiency for its customers.
Companies are using logistics data and applying it to minimizing shipment distances in light of, say, changing fuel prices or a shortage of drivers to ensure on-time delivery within budget constraints.
For the CFO, the potential to leverage AI in business to spot outliers and provide real-time adjustments is immense. Transaction processing still takes up almost half of the finance department’s time. But with AI, finance can automate processes like analyzing corporate transactions for various compliance requirements and to mitigate anomalies, for example, or to automate closing the books.
AI can even be applied to finding the right people for critical positions. For instance, if a recruiting manager is filtering graduates from nearby colleges and universities, an autonomous agent can help identify ideal candidates, such as multidisciplinary students who combine a degree in the sciences with strong communication skills.
For some, AI drums up images of a future in which machines dominate the workplace and render people secondary. That fear is a big obstacle to implementing and scaling these projects. AI will displace jobs, but studies show that it will actually create more—as many as 58 million new jobs in the next few years. As Tesla’s Elon Musk noted, “Excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated.”
Once again, baseball also provides guidance. For instance, the pitching coach must strike a balance between respecting the player’s experience and imparting the value of getting instant, instructive data-driven feedback. To accomplish this, the coach must turn the information from the machine into something of real value and relevance to the actual situation. It helps to start by using the data to validate or disprove, something.
In baseball, change management involves developing systems that are relatable. For instance, some pitching coaches are marrying AI data with data the pitcher is already familiar with, like batting averages, according to Brower. It’s also important to create visualizations based on parameters determined by AI, and that make sense for the end user—in the pitcher’s case, for instance, using red and green zones for pitch location.
“You have to tailor your information to what is already known,” Brower said.
That analogy is applicable to the manner in which technology providers are weaving artificial intelligence into solutions, for example leveraging integrations with intuitive UIs and bots that can connect to conversational interfaces, including text messaging and voice. A bot designed to help ease filing a business expense can make the process seamless from a smartphone: from an interface like Slack, the phone’s camera can scan a receipt, recognize the date, amount, vendor and purpose of the expense, and categorize it. If the bot recognizes a meal, it can file it as lunch or dinner based on the timestamp.
Just as the cloud presented its challenges when it was first being deployed, companies are starting with small pilot artificial intelligence projects. Although artificial intelligence and machine learning are relatively new and unknown opportunities, ROI is the ultimate arbiter of risk, and that will be key to its growing adoption.