Skip to Content

These Are IBM’s 3 Tips for Using AI in Business

Opening Day Of The World Economic Forum (WEF) 2017Opening Day Of The World Economic Forum (WEF) 2017
Rometty spoke at the WEF in Davos on Tuesday.Bloomberg Bloomberg—Getty Images

Here’s something that artificial intelligence may not have predicted: As AI has gotten better, businesses, particularly because of ethics and privacy issues, are having a harder and harder problem putting it to use.

At the World Economic Forum, IBM CEO Ginni Rometty shared some of the important lessons her company has learned by working with businesses to implement AI.

1. It’s not computers vs. workers.

There’s a lot of fear that AI will elimination jobs, Rometty said. It’s possible that some professions will be wholly replaced by automation, but at most of IBM’s business partners “it will not be man or machine,” she said. The purpose of the technology is instead to augment and supplement what humans do. That needs to be clear.

Subscribe to The World’s Most Powerful Women, Fortune’s daily must-read for global businesswomen.

2. Industry domaine really matters.

Every institution has its own data to contribute to AI, Rometty said, but to truly unlock its value it must be combined with industry data. “These systems will be most effective when they’re trained with domain knowledge and in an industry context,” she said. IBM, for its part, has focused on healthcare.

3. Know your business model.

Any company—new or old—has accumulated assets in the form of data, Rometty said. Data is a competitive advantage, “so we believe strongly as a business, that the insights you get from your data belong to you and that applies also to how these systems were trained,” she said.

Joichi Ito, director of the media lab at the Massachusetts Institute of Technology, pushed IBM and Microsoft, whose CEO Satya Nadella appeared on the same WEF panel with Rometty, to go a bit further and ensure that the people who create AI technology reflect those who will actually use it. He referenced a black researcher in his lab who realized some facial recognition technology didn’t pick up darker faces. “[O]ne of the risks that we have in the lack of diversity among engineers is that it’s not intuitive to the kind of questions you should be asking, even if you have design guidelines,” he said.