"There's No Gold Standard": Lisa Johnson and Tamara Havenhill-Jacobs on Leading AI Before the Playbook Exists

Lisa Johnson, VP of Clinical IT Solution Delivery at Providence, and Tamara Havenhill-Jacobs, CIO at Alpine Physician Partners, recently completed a shared 13-week MIT course on AI strategy and leadership, finished between flights, evenings, and weekends. At a 229 CIO Summit in Napa, the two talked about what the course actually changed about how they lead.

The Confidence Shift

Johnson came to health IT from clinical care, not from a programming or analyst background, and said that history made her tentative about claiming space in AI conversations. The course reframed that.

"I realized my leadership style, the way I lead teams, what I know, I am built for this," she said.

She later applied that confidence directly in the board presentation on whether Providence should be using AI at all. Johnson framed her answer around what she called the sacred encounter, the space between caregiver and patient in moments of greatest need, arguing that AI, unlike the EHR before it, has a chance to recede into the background rather than sit between provider and patient.

"We're actually going to make sure that when we deploy it and utilize it, it's because we're trying to actually bring that human experience back in," she explained.

Governance Without a Gold Standard

Havenhill-Jacobs came into the course already running AI work at Alpine, including an internal platform for agent development built the previous fall. She said the coursework helped sharpen how she thinks about governance, an area where accepted standards are still emerging.

"There's no gold standard for governance right now," she noted.

Her current approach starts with categorizing the AI itself: deterministic, generative, agentic, or a hybrid orchestrating across all three, each carrying a different risk profile and a different governance responsibility. Rather than standing up a separate committee for every initiative, she is working toward assigning each workflow or agent its own accountable team, one that reports back to a core governance function on performance.

That accountability extends to measuring whether a tool is actually working. Havenhill-Jacobs said the harder problem is translating qualitative gains, the kind without an existing baseline, into something consistent enough to track, and added that tools that go unexamined too long tend to quietly stick around long after they have stopped proving their worth.

Johnson described a parallel version of the governance question at Providence: consistent scaffolding, with different paths through it depending on the use case. Lower-risk, third-party tools with a strong track record move faster.

"There's also kind of a fast pass, like Disney FastPass, for things that maybe... a third party is delivering that we feel really comfortable with that has less risk," she said. "The scaffolding and the framework is consistent, but there's just different routes basically that it needs to take based on what it is and the risk involved."

Letting People Play

Both leaders pointed to permission and access, not the technology itself, as the real cultural blocker. Havenhill-Jacobs described building a sandbox environment at Alpine where employees who meet certain requirements, or who have shown strong interest, can test coworker and coding tools directly. From there, a tool with demonstrated value moves through oversight, testing, and governance before a decision is made on whether it earns a permanent place in production.

Johnson said the same principle applies to clinicians, who need to be at the table experiencing AI directly rather than hearing about it secondhand. Relating the technology to tools people already use every day, she added, does more to ease fear than any policy document could.

"Once you start relating it to real things that they're utilizing every day, then it's less scary," she said.

What They're Watching Next

Johnson pointed to early cancer detection through better pathology and imaging as the development she is most hoping to see, a topic she called personal for her. Havenhill-Jacobs said her hope is that AI stops being a source of fear in these conversations and becomes ordinary: a tool in regular use that has visibly reshaped the experience for patients, providers, and staff alike.

For Johnson, that future still comes back to a simple boundary. She described AI's role for now as a partner riding alongside the people doing the work, not replacing them.

"The human is still driving," she said plainly.

Lisa Johnson is VP of Clinical IT Solution Delivery at Providence. Tamara Havenhill-Jacobs is CIO at Alpine Physician Partners.