© 2019 SourceMedia. All rights reserved.

CIOs must take charge in implementing artificial intelligence governance

The role of many chief information officers is changing, as organizational strategies advance from data governance initiatives to artificial intelligence governance. That is placing new pressure on CIOs to take a more advisory role in what is for many is still cutting-edge technology. But this is welcome news, and brings new opportunities for those that are successful.

That was the message of Jean-François Gagné, chief executive officer at Element AI, at the Strata Data Conference in New York this week. Gagné spoke on the topic “From data governance to AI governance: The CIO’s new role.”

Data governance has grown in importance in recent years, as organizations try to get more value from the mountains of data they collect or create. As evidence, Gagné said 39 percent of organizations currently have a data governance framework in place, another 15 percent are now implementing one, and 24 percent have just started the process.

Driving these numbers are three key goals: improving efficiencies in processes (cited by 54 percent of IT leaders in a recent survey); regulatory pressure (cited by 39 percent); and customer service demands (cited by 7 percent).

CIO and AI governance.jpg

Success with artificial intelligence governance will require getting past a couple of myths, however, Gagné said. One is that big data is all too much for most organizations to master, and another is that all data is created equal. A successful governance program requires that an organization be able to distinguish what data truly has value, and focus its efforts on that.

Opportunities with artificial intelligence governance include several improvements with data management, including communication, perception, reasoning, decision making and interaction. To realize those gains, organizations should create an AI governance framework that focuses on four key pillars: processes (including a focus on accuracy, bias and completeness); security (including a focus on adversarial robustness and adaptability); privacy (including focus on IP capture and impacted users); and transparency (including a focus on explainability and intent).

For reprint and licensing requests for this article, click here.