September 22, 2017
A paper published in Scientific American together with a recent MIT Sloan Management Review study on artificial intelligence argues that expectations have gotten far ahead of actual applications. As our recent membership meeting demonstrated, there is a feeling that robotic processes will soon displace significant portions of a company's current workforce. However, the Sloan study, which partnered with The Boston Consulting Group, found only one in twenty companies have extensively incorporated AI into offerings or processes. "Getting business value from AI may be theoretically possible but pragmatically difficult," write the authors of "Reshaping Business with Artificial Intelligence." A key barrier is the difficulty of developing or acquiring the requisite talent in the context of other demands for resources inside companies. Another is integrating the capabilities of humans and machines. If an AI application completely removes a person, there may not be an integration problem. But if someone is still necessary for part of the task, then livelihoods are threatened. Interestingly, the papers stress that for machine learning to be enabled, relevant historical data must already be in place and that no amount of algorithmic sophistication can overcome a lack of data. Yet, it is human nature to record successes and block out failures, and it is the combination of negative and positive data that draws lines between what works and what does not. For where AI is working today, the Sloan study shows immediate returns being realized in the area of business process outsourcing, with tasks in the areas of system administration, IT administration, business operations and verification that once were moved to low labor-cost countries now going to the cloud. But as the CEO of Infosys, Vishal Sikka, cautions, "If you look at the history of AI since its origin in 1956, it has been a story of peaks and valleys, and right now we are in a particularly exuberant time where everything looks like there is one magnificent peak in front of us." The thesis of the Scientific American article, "Artificial Intelligence: The Gap Between Promise and Practice," is that "AI likely holds much promise for organizations—but we must first get far more precise about what the promise is in order to understand how AI can be deployed to fulfill it."