What will it take for AI to transform the patient experience?

When it comes to driving digital health forward, data deserves a seat at the table

Recently, we hosted one of our Modus Monthly Salon dinners in Manhattan with digital leaders in healthcare. It was a nourishing and delightful evening on all levels and included hearty discussion around our topic du jour: How will AI affect patient experience?

On its face, it’s a somewhat basic, broad, and even tired question. But what made it interesting was the people discussing it, who are on the front lines of implementing generative AI in their organizations. Notably, our two “tableside” chat participants — an executive heading the decision science and AI transformation efforts at a large pharma company, and the head of patient experience for a large health system. Both are multi-billion-dollar enterprises that serve millions of lives. 

Both organizations have deployed and are deploying LLM-driven experiences that are focused on reducing known friction and optimizing existing patterns. On the provider side, that means moving beyond basic filtered search toward context-aware physician finders—systems that can interpret patient preferences, journeys and other complexities; traverse multiple datasets; and return more relevant, personalized, and useful recommendations.

On the drug manufacturer side, this tech is being used to allocate co-pay reimbursement funds in ways that are highly optimized to a patient’s insurance as part of a modernized patient support program. Many organizations are beginning to draft and enhance patient communications using generative AI, which, according to a study in Nature, are proving to be perceived as more empathetic than communications without the same assistance. (Perhaps because the bots are not burnt out and being pushed to see 15 patients per hour, not because HCPs are not empathetic.)

Image source: Nature

Looking to the horizon, the possibilities are even more transformative. Examples include real time 3D visualization of test results, scans, biological systems, disease pathways, and care plans in conversation with a health care provider, and a personalized “agentic” world in which AI agents advocate for, educate, support, and manage care plans, ask questions, and synthesize details on behalf of patients. This may not only improve experience but also outcomes—we know that just being married or having a partner leads to ~12% better survival rates for cancer patients. Now imagine what a highly engaged AI health partner could do.

But here’s the rub. As everyone who is working in this field knows, step one is getting your data house—lake, pond, warehouse, or glacier—in order. And this is where the entire table got riled up at the dinner. Somehow, after over two decades of mandated digitization of patient data, the datascape remains largely fragmented, incompatible, not portable, and (this is where my blood boils) the patient is not in control and often not even aware of what is happening with their data.

I like to use the analogy of money. Aside from a few bills or coins in your pocket, money is data—numbers on a bank’s server. Imagine if you were unable to view your account balance easily? Or you were unable to make a withdrawal or transfer your money from one institution to another? That would not be ok. It would create massive friction in the economy. It is data that is highly regulated and worth, well, all the money. Shouldn’t we have at least the same level of control over the data that describes our bodies?

The table circled around a few potential accelerants:

• Central authority: We need national coordination and legislation to set standards and drive interoperability and portability. We have pieces of this like HL7 and the big moment in 2021 when CMS finally made a rule on the back of the America Cares Act (2016) that guarantees patient access to their own data (the how and when and what notwithstanding). But we need to go further, faster. So, perhaps, the tech goliaths, including Amazon, Meta, Google, Microsoft, and Apple (or EHR fam of Epic, Cerner, etc.), could get together and establish protocols and accelerate the implementation.
• Blockchain: Despite being in the trough of some hype/adoption curve for this class of tech, distributed ledger tech is a very secure, portable, clear way to store and move data around, and could give patients the control and transparency they deserve.
• Gen AI: The tech that could be made incredibly useful if we can improve the patient dataverse could itself be harnessed to help understand, organize, and make use of the volume of non- or semi-structured health data out there. 

I am an optimist at heart. I believe we’re on the cusp of a convergence: technical maturity, public awareness, institutional urgency, and economic incentive. It reminds me of the mobile revolution—it was not just about the smartphone, but the confluence of lithium batteries, cloud computing, GPS, accelerometers, and 3-5G networks.

Like all technology watershed moments from fission to the printing press, the tech itself is not good or bad, generative or harmful. It is the choices we make that matter when the deployment is in our hands. Let’s choose well.