AI & Data Exchange 2024: IBM’s Ryan Macaleer on why AI success hinges on enterprise platform

Agencies can implement AI across multiple uses cases by adopting standard governance and tools, advises the IBM AI and technology leader.

Jason Miller@jmillerWFED

February 12, 2024 11:49 pm

3 min read

Agencies have been moving fast on aspects of the White House executive order on artificial intelligence issued in the fall and the Office of Management and Budget draft memo that followed.

Many have named chief AI officers. The AI use case inventory continues to grow. But perhaps most significantly, agency leaders and teams now express excitement about how best to add AI tools and capabilities to their missions to drive change.

Ryan Macaleer, vice president of data and AI technology for the U.S federal market at IBM, said the executive order (EO) and memo were critical first steps in the government’s AI revolution.

“The EO and memo are providing the guardrails and the guidelines to really help govern this emerging AI space. We also think that the solution to the challenge of supporting and enabling safe and secure AI — resides in open innovation. A diversified point of view is what we’ve always relied on as a country, as a society and certainly as a federal marketplace to bring these innovative solutions to bear,” Macaleer said during Federal News Network’s AI and Data Exchange.

He pointed to IBM’s own work on evolving technology for the government for nearly a century and recently on AI projects at several agencies.

“We have been working with federal agencies for over a decade. For example, we’re working with the Veterans Affairs Department, where we’re leveraging AI to reduce delays associated with processing veterans’ claims,” he said. “Other agencies like the IRS have used AI to help expedite the tax return process, and the Navy Fleet Forces Command is harnessing the power of AI to improve its food supply chain and, consequently, fleet readiness.”

To integrate AI in current technology stacks for these and similar uses cases, public and private sector organizations must overcome shared challenges, Macaleer said.

Ensuring AI data quality, accessibility and governance

Macaleer said one big challenge has to do with data – data quality, accessibility and governance. He said that also encompasses security, privacy and even the cost of implementing the technology.

“We do a survey every year of executives across both government and industry, and 84% of the executives that we surveyed cited at least one of those three things as being a challenge,” he said. “The other vector is, even when enterprises decide to go down the AI path, there’s the challenge of having the ability to execute. The first thing that the same executives cited is having a lack of an enterprise perspective on AI and an enterprise standard. That presents a real challenge just to get started.”

Macaleer said a typical question is, how does my organization measure return on investment across the enterprise?

He said one approach to help launch AI successfully is to use an enterprise platform to ensure security, to create governance standards, to share tools, to integrate data and to develop repeatable processes.

“When these AI models first came along, every use case had to have its own data source and its own AI model. It was very siloed. In that sense, it took a lot of resources, and it took a lot of human power as that process can get pretty time consuming and fairly expensive,” Macaleer said. “There was no repeatability. If you’re solving one use case, you solved that one use case.”

But that’s changing rapidly with a generative AI solution stack, he said. “The idea is you’re training this model once, and this idea of a large language model is, at its core, able to solve more than one use case.”

Discover more articles and videos now on our AI & Data Exchange event page.