There is a lot of confusion around AI and machine learning tools, and how they work or can work to increase efficiency. For all of the panelists, most of the work—about 80%—is focused on structuring data and the program and only 20% was actual machine learning. At its core, 70% to 80% is blocking and tackling around the structuring of the data [needed] to do cool stuff of the predictive analytics, which is about 20%. That isn't seeing right in front of you, but seeing around corners and seeing trends before the happen.”

Zipori concurred that most of the process was structuring data, and he added the results of AI are really meant to support investment decisions, rather than make them. Our challenge also starts with structuring the data, and there are parts of that that are not machine learning and AI,” he said. “We put a lot of effort tin predictive models. We want to know what will happen to an asset so we work on predictive models. There are problems that can be solves autonomously and others that need human involvement. We are not trying to take in autonomous decisions. We are focused on using predictive analytics to support decisions.”

This concept was a clear takeaway from the conference—the idea that AI should be used to explain human decisions. To teach a machine to make the right decisions, you must feed it data. “You need training data sets,” said Sarkis. “There is a roar for training data, and we have an advantage because we have data, and we use that to train and retain your algorithms. AI isn't trying to come up with something different, but training it to make the same decisions that you would make more quickly.”