AI summit unites Academy, innovators, leaders for practical look ahead 

What role do you see AI playing in administrative simplification tools over the next five years?

I think in the next couple of years, we'll start to see a new layer of assistive technology, on top of the EHR either provided by EHR companies themselves or by third parties or some combination thereof. And I think we'll see them deepen their focus on trying to assist the physician in doing their work.  There is likely to be a lot more focus on administrative simplification initially because there's a lot less risk there. 

And there's an epidemic of administrative burden that we need to address. We talked about that in one of our sessions at the summit, when we talked with docs who are already using AI in their practices. And we saw that ambient listening is probably the poster child right now for assistive technology that's reducing administrative burden and improving burnout in physicians. I think we'll see that grow. 

How does AI fit for you in your experience of informatics and your vision for how systems should work?

We’re definitely still trying to solve the interoperability problem. I think AI can actually help assist us in doing that work.

I also think it’s important for people to understand that until we had these very sophisticated AI tools, we were focused on automating things. And in automation, it works the best when you know all the parts and there's not a lot of different paths. When two or three things can happen, you can kind of automate that. Once it's more complex than that, it's very difficult to manage. That’s why we have humans in the process to say, Oh, this is what we need to do. With a tool such as ambient listening, you're adding some intelligence to the automation and now the technology is able to manage a different level of complexity. I think that's the big thing that's really, truly different: You actually have another type of intelligence in there to help you simplify all of the variability that we have in primary care. 

How would you characterize the learning curve for AI tools right now, especially compared with EHR?

A lot of the learning curve for the physician is how to actually use the tool itself, and these are pretty easy because they’re using some type of natural language, as though you’re talking to an intern and saying what you want the intern to do.

The steep part of the learning curve is understanding what the system does and what it’s capable of doing, including the capability to hallucinate. That's why as part of this effort, we created some CME around that, and we're working with some other entities to create some additional CME around AI for family physicians to better evaluate and understand these tools. 

What did you take away from the summit session about challenges with AI in health care? 

I think that session was really focused on making sure we understood some high-level challenges and could start talking about how to mitigate those. I think I’d bucket them in two different areas: external and internal.

The internal challenges are things that are very specific about the AI technology itself—for instance, known bias issues in the data. We need to make sure that we can get more training data that are better representative of the populations that family docs and primary care docs see. Also, the hallucinations that can be inherent to generative AI. The incentives are aligned for companies to work to solve these things. 

On the external side are the same things that drove where we ended up with the EHR: We still have fee-for-service payment and we still have a very profit-driven health care system that’s very fragmented and doesn’t have the interoperability we need. Those things not only impact AI but also affect other parts of primary care, and they are going to be harder to solve because the financial incentives are not yet aligned. I've been working on this for 20 years, and we’ve made some progress, for sure, but we haven’t solved them. One thing that is different is the lack of interest within the federal government to regulate AI heavily as EHRs were during meaningful use. This means that the private sector will need to lead efforts to establish standards and policies for responsible and safe AI. 

So those are the big ones that we have to keep an eye on, because some type of AI triage solution might instead disintermediate primary care; focusing on getting patients to specialists for procedures would be very rewarding for a health care system, but we know that's not what's going to deliver the highest-quality care for patients. 

What are the big overall takeaways regarding AI in health care?

One of the big things is, this is going to have to be a collaborative effort. There’s not just one entity driving this to fruition. It will take a national, collaborative effort involving multiple organizations. Which I’m really excited about.  

One of the big things that I saw, and that I loved to see, was the back and forth, just the engagement between the AI community—the technology people and the entrepreneurs and the EHR vendors—and the frontline docs, talking very specifically about how to move forward. 

What does the roadmap look like for using AI in health care? 

One, what are the policies, the safety and guardrail things that need to be put in place? Two, how do we talk about primary care and its value and its opportunity in AI? Primary care has the biggest breadth of data. We see the most different types of issues as well as comorbidities. We have all that complexity. So how do we drive that to make AI even better? How can we decrease the administrative burden that is there and deliver high-quality care, leveraging these tools but not being disrupted by these tools? 

We'll be publishing a document from the summit sometime this summer that lays out a roadmap for the AAFP and others, based on a strategic plan for AI and what we’ve learned so far from this partnership and from this event. 

The other big thing is our ongoing work to identify collaborators and coalitions that are aligned with the needs of primary care, to really drive things forward. I think the private sector is probably going to have to do a lot of the heavy lift over the next year or two. I don't think we can depend upon the federal government to help us significantly in the immediate term. 

The AAFP is well known in the primary care space and also in the policy space, and we’ve done a lot of work there to drive EHR and interoperability. The space we’re less known is the AI space, which doesn’t necessarily come from health care. That’s Rock Health’s sweet spot. So I think this partnership has been extremely helpful for us, and I think it's also been helpful for Rock Health. 

And while we do this work, we aren’t stopping our efforts to improve EHR and interoperability and information blocking, the policy and advocacy work on those and related things. That work continues. We just think this is a great opportunity to leverage what we know from all of that as we use new technology to improve the primary care experience for both physicians and patients.