The conversation about AI in healthcare is moving beyond humblebrag mentions of how companies are using it toward genuine discovery, sharing, and learning. At Included Health, bottom-up experimentation sparked excitement and adoption across the company, but regular use of our enterprise AI platform plateaued at around 70%. So we shifted toward a stronger department-led approach, tasking leaders to rethink workflows and assess ROI at the team level.

Bottom-up experimentation builds momentum, while top-down direction sets clear expectations and priorities. We've learned the two are complementary. Here are five takeaways from the discussions I was part of at the inaugural Second Opinion Summit:

• AI adoption starts at the top. Leaders need to model the behavior and make the learning visible.
• Measure impact, not activity. Spend, adoption, and token usage aren't measures of success on their own.
• Technology is no longer the moat. Distribution is. Established companies can close the gap through disciplined AI adoption.
• Culture matters as much as capability. Leaders need to address the understandable fear that AI will replace jobs and position AI fluency as professional development.
• Agency and judgment will matter more than specific skills. As AI handles more technical and repeatable work, sound judgment, problem-solving, and creativity become more important in hiring.

As I wrote in a recent article for Second Opinion reflecting on the event:

Many companies in our space are figuring this out in real time, as are we. Don't be afraid to invest when the gains are clear, but measure the outcomes that matter and share what you learn with your peers. We're all learning together, and there's immense value in transparency.