AI-driven technologies can cut costs, help eliminate mindless employee tasks and boost profits while giving employees more meaningful jobs to do.
Yet AI adoption in the enterprise at a scalable level has been slow, with many organizations across different industries deploying little to no artificial intelligence technology. At an executive level, the reasons, according to experts, include nervousness, pride and a lack of explainable AI.
AI by the numbers
A recent KPMG 2019 Enterprise AI Adoption Study found that 17% of companies use AI and machine learning at scale. About 30% of the 30 Global 500 companies KPMG worked with in the study reported using it for selective functions.
Despite the low AI adoption in enterprises, about half of the companies interviewed expected to use AI and machine learning at scale over the next three years.
Partly, those low AI adoption rates could have to do with a lack of explainable AI, said Traci Gusher, principal of data and analytics at KPMG, a global network of audit, tax and advisory services firms.
"There's been a lot of software developed to solve for a very specific problem where it does that very, very well," she said. But how they arrive at those answers has typically been a bit of a mystery to business users.
AI-driven software may offer general explanations into how AI technology works, but likely won't dig into the complex algorithms that power it, she noted.
A lack of a detailed explanation into how a technology works -- whether it's because it's too complicated, like a neural network, because the algorithm is proprietary, or for another reason -- can make business users nervous.
Imagine getting into an autonomous car, Gusher said. You might not be a perfect driver, but are likely pretty good, and are able to drive daily to work or to run errands without an issue. It's familiar and it's comfortable.
Even if the autonomous car can drive perfectly, you might be nervous about letting it drive you around town. You don't know how it works, and you've never tried it before. You're relying on assurances of others to determine if it's safe and efficient or not.
Business executives, Gusher said, feel similarly nervous about relying on an automated system to help make decisions. It's unfamiliar and hard to understand, and, besides, things have worked fine without it. It might cut costs or raise profits, but, then again, it might make the wrong decisions and ultimately hurt the company.
A general nervousness, according to Gusher, is one of the reasons for the lag in AI adoption in the enterprise.
According to Michael Feindt, founder of Blue Yonder, an AI for supply chain vendor acquired by the software and consulting firm JDA in 2018, nervousness may only be part of the problem. The other part, he said, could be pride.
Adding automation into the supply chain, for example, is "quite a huge change in the way of thinking and acting," Feindt said.
For enterprises -- retail chains in particular -- it's a change that requires departments to become more connected, for employees to communicate more and for executives to leave at least some of their decision-making ability up to a data-driven technology.
Michael FeindtFounder, Blue Yonder
"There is a lot of resistance in some organizations," Feindt said. "People are afraid to lose power."
Usually not at the very top, he said, but in the second or third level down.
Often, executives came into their positions because they made the right decisions, Feindt explained. With AI, they are being asked to let an automated system make some of those decisions, and it becomes "a matter of pride," he said.
While it's becoming clearer that AI is becoming necessary, it's "hard to start to change," he said.
Yet, he added, "If [your company is] the last one to automate a system, [it] might die." The company may fold, unable to compete with others that have already jumped on automation.
"It's really dangerous," he said.