Corewell’s Joseph Says IT Leaders Can Spur Innovation by Empowering Users to Address Unsolvable Problems with Novel Solutions

When the dust settled from one of the largest health system integrations in Michigan history, Jason Joseph, Chief Digital and Information Officer at Corewell Health, found himself facing a new frontier: AI.

As Corewell brought together Spectrum Health, Beaumont Health, Lakeland Health, and health plan Priority Health under one umbrella—with 21 hospitals, 65,000 employees, and 1.4 million health plan members—Joseph’s team had already executed a formidable feat: consolidating infrastructure, cybersecurity, ERP systems, and Epic instances across the sprawling organization.

But with the emergence of generative AI tools like ChatGPT, another challenge demanded urgent attention—how to foster innovation without losing control.

“In the middle of all of that, AI became the new electricity,” said Joseph, likening the technology’s ubiquity to the transformational force that reshaped the industrial age. “We realized very early on we needed to build something that would give us both the ability to explore and the framework to do it safely.”

That “something” became the AI Center of Excellence (AICOE), a governance body rather than a development team, designed to set guardrails and offer guidance on AI-related initiatives across the enterprise.



Beyond a Centralized Command

At first, the AICOE was an informal assembly of clinicians, technologists, compliance officers, and other stakeholders exploring AI’s capabilities. Over time, it evolved into a more structured function under a newly appointed director of analytics and AI services, who reports into Corewell’s data and analytics division.

Crucially, Joseph was adamant that the AICOE not become a centralized team for executing all AI projects. “The people closest to the innovation are the ones who will figure out how to harness it,” he said. “Our job is to educate, support, and ensure they’re working within the appropriate framework.”

Corewell’s approach contrasts with that of some peers who have appointed Chief AI Officers and formed centralized execution teams. Joseph isn’t convinced that’s scalable or effective. “AI is going to be in our core applications, in third-party services, in solutions we build ourselves,” he said. “Trying to own everything AI-related through a single group is short-sighted.”

He likens it to the early days of electrification. “You probably had a chief electricity officer during the industrial revolution. But eventually, the knowledge had to diffuse to every part of the organization.”

Instead of bottlenecking progress, Corewell’s model aims to push AI expertise outwards. One example: when the system piloted an ambient documentation tool, the implementation was led not by a standalone AI team but by the clinical informatics and core systems groups supporting frontline workflow.

“If something like that came in from left field by a team that didn’t understand clinical workflow, it wouldn’t be optimized,” Joseph explained. “You want people who understand both the problem and the technology.”

Right Tool, Right Job

While some health systems are still experimenting with generative AI use cases, Joseph and his team are thinking several steps ahead, delving into agentic AI—systems capable of performing multi-step tasks on behalf of users.

Yet he cautions against lumping all AI into a single bucket. Corewell’s approach varies based on the specific challenge at hand. “Radiology, for example, is all about detection,” he said. “You’re training models on millions of images. But that’s totally different from generative AI, where you’re summarizing documents or creating content.”

Each use case demands the right type of AI—and the right governance model. “We still need technical experts, we still need good governance,” Joseph noted. “But we also need the right process to match the problem.”

That extends to data strategy. While AI’s dependence on high-quality data is well unders...