Five Lessons for Making AI Adoption Work

Last week, I was lucky to be part of a curated group that attended the inaugural Second Opinion Summit. It was a great event, bringing together a diverse group of voices from across providers, payers, technology companies, investors, and patient communities.

Beyond simply being in the room, which was fun and interesting in its own right, I’ve spent the past week reflecting on the conversations around the panel I joined, “The AI Org Chart: What Happens to Healthcare Teams Now?”, alongside Sean Duffy of Omada Health, Abhinav Shashank of Innovaccer, and Guillaume de Zwirek of Artera, with Julie Ebert of Stifel moderating.

What’s refreshing is that the AI conversation is moving beyond humblebrag mentions of how an individual or business is using AI toward genuine discovery, sharing, and learning. I thought I’d share a few takeaways from the discussions I was part of.

Coming up · November 24 · Noon–1 p.m. ET AMA with Doximity CEO Jeff Tangney A live Q&A on where clinicians are finding value with AI and building a company from zero to IPO. Co-hosted by Chrissy Farr and Rebecca Mitchell of Scrub Capital, for Second Opinion premium subscribers and Scrub Capital community members. Register →

1. AI adoption starts at the top

Included Health initially rolled out Glean, an enterprise AI platform for finding information and getting work done across company tools, by encouraging bottom-up experimentation and launching an AI Champions Network to help team members identify useful AI applications. That approach sparked excitement and adoption among employees across the company, though many still wanted a clearer direction of how AI could further be used. When adoption — measured by regular individual use of Glean — plateaued at around 70%, we shifted toward a stronger department-led approach. This remains a work in progress, but the goal is to go beyond individual efficiency and to task our leaders to rethink workflows and assess ROI at the team level as we combine AI, operational change, and human expertise to deliver higher-value work.

We’ve learned the two approaches are complementary: Bottom-up surfaces experimentation and builds momentum, while top-down sets clear expectations and priorities.

Product development is one area where I’ve seen this come to life. Rather than relying on individual product managers and designers to experiment on their own, a clearer mandate from Product and Engineering leaders is helping transform the product development lifecycle. We’ve moved past high-level encouragement and are now treating AI adoption as an active change-management effort. We’ve set a clear, standardized approach to integrating AI, are testing it through targeted pilots with early adopters, and are pairing structured training with dashboards so managers can be accountable for scaling what works. We’ve learned that strong AI adoption takes more than access to tools or even a mandate, alone. It also takes leaders who model the behavior, map a practical plan, and create a culture where teams are experimenting and learning together.

At Included Health, this is happening on an ongoing basis: Our head of product showed his team how he uses AI to scale his work; our head of engineering is exploring agentic coding herself constantly to stay current; and I recently built a chief-of-staff agent that’s already making a real difference for me — surfacing must-cover agenda items, flagging inconsistencies in what’s being said versus reported, reminding me and my team of things we say are priorities. In short, leaders need to walk the walk and make the learning visible.