Agent Alert: Healthcare’s Future is Calling (Are you listening?)

Last week, I wrote about some code snippets found on ChatGPT that pointed to OpenAI integrating shopping capabilities directly into ChatGPT. Turns out, it wasn’t a rumor: they announced that this is indeed happening. I discussed how this could impact healthcare and people searching for and booking an appointment with a provider. Whether this kind of “shopping” via conversational platforms like ChatGPT comes true for healthcare in the next few months or few years remains to be seen, but the most important element of its ability to work in healthcare will be considering whether data is accurate enough to enable this capability.

Imagine this: You need to see an orthopedic specialist. Instead of hours and hours of Googling and sifting through blue links, your AI agent handles it for you. Your agent knows your insurance, your medical history and your provider preferences (e.g., you prefer providers who are within a five-mile radius of your home). It finds the right doctor, books the appointment and even adds it to your calendar.

I’ve written about this scenario before.

Sound futuristic? Maybe.

But it’s closer than we think.

Consumers will quickly start to see what this experience is like outside of healthcare. They’ll immediately see it in retail (as evidenced by the ChatGPT integration with Shopify and the Google release of Project Mariner). Once they start to adapt to leveraging agents to work on their behalf and to work with chatbots (rather than sift through blue links) to find information, the time will come when they will expect the same for their healthcare. With agents, the learning curve will be steep – users are going to move from asking questions in a search box to delegating via conversational chat, and that takes a bit of time to get used to. But I predict it will happen.

I’m a competitive person. And frankly, I’m annoyed that the healthcare industry is never leading the way in anything but, maybe, research. Instead, it is always following in the footsteps of other industries. At the same time, the industry is always complaining “it’s too hard”... or “it’s not possible.” (And when organizations try to make changes in healthcare and purportedly “fail” – the siren call is “See? Healthcare is hard, even XYZ organization couldn’t do it.”

Why is our industry so reactionary? We see these incredible advancements in retail, travel, finance and then… it’s crickets in healthcare. We’re serving people at their most vulnerable, their most stressed, and yet we can’t seem to provide a seamless digital experience.

Think about it: shopping online is a breeze (and is only going to get easier as we leverage agents to do it on our behalf). I can order a pair of shoes and have them at my door tomorrow. But does it really help anyone? Not really. It helps people consume more. And I’d argue that’s not all that virtuous.

In healthcare, shopping for the right doctor is often a nightmare. And, it’s not just inconvenient. It matters (more than a new pair of shoes). And, if an organization can help a patient find the best care in the most frictionless way possible? Now that is virtuous. (And many times, could be life-saving!)

The difference in experiences in industries outside of healthcare and how this is experienced within healthcare will only widen as technology platforms continue to deliver more user experiences that tap into how people want to find information and now how they want to consider transacting.

A help article about shopping on ChatGPT on OpenAI’s website states:

“if a user asks ChatGPT for help finding goofy costumes for their two large dogs, ChatGPT will consider general factors, such as price, customer ratings, and ease of use, as well as specific criteria provided by the user, like sizing and the desired costume vibe. If the user had previously indicated a dislike for clowns, the model might also consider that and leave out clown costumes.”

You can’t do this with a traditional search or blue link experience. This level of personalized, contextual search is exactly what people will come to expect. And they'll want it in healthcare, too.

There is also a consideration of transactions (or conversions) – this week, I watched an interview with Ryan McInerney, the CEO of Visa, announcing a set of tools that will give agents capabilities to make a payment on behalf of the user. Visa is leaning into the idea that people aren’t going to be doing the searching and converting anymore – but agents will be, instead.

As the Visa CEO said, “we’re going to have agents and they’re going to be able to go out and scour the world’s inventory and find what it is that we’re looking for…” and they’ve brokered partnerships with organizations like OpenAI, Microsoft, Perplexity, IBM, Stripe and more to build out the infrastructure to enable shopping and buying experiences.

Imagine being able to tap into an agent that can search and transact in healthcare. Imagine an agent saying, “I found two highly-rated cardiologists in your network, all with openings next week. Dr. Smith has extensive experience with your specific condition, and Dr. Jones is closest to your home. Which one would you prefer?” And then, on your preference, books the appointment?

For this to work, healthcare organizations need to really double down on data. All that data – provider specialties, availability, insurance, patient history, etc. – needs to be organized, accessible and machine-ready.

Shopping is going to evolve – quickly. And that also means shopping for a provider is going to evolve, too.

And the agents are going to rely on data and basic website experiences to get the job done.

If someone is seeking care, an agent may need to find different data points including, (but not limited to):

• Answers to questions about conditions, procedures, treatments

• Providers with certain levels of expertise

• Provider specialties

• Locations for care relevant to the searcher

• Provider availability

• Provider ratings (or, even better, expertise as indicated by CPT codes and reimbursement records, for example)

• Payment ability (credit card or bank account)

• Patient insurance

• Patient portal login credentials

I hate to say it, but guess who has a lot of this information already?

Epic.

They are (arguably) basically already a monopoly. But maybe their monopoly status could be used for good? Epic has so much data. They could be the key to making this agent-driven future a reality. If they partnered with OpenAI, Perplexity, Google, Anthropic, etc., we could maybe see some real progress. (But, honestly, I’m not holding my breath – they’re not in the patient experience business).

Since I don’t see Epic leading the charge any time soon, it’s on healthcare organizations to take the lead. Here are a few things healthcare organizations should urgently consider:

• Outline the data an agent could be seeking: What would an AI agent be looking for on a website? Think doctor specialties, availability, patient reviews and more. Refer to the data points I listed earlier if you need inspiration – and I’ve included a prompt you could ask ChatGPT or Gemini to get a comprehensive list, too, below). (And, let’s finally ditch those useless “patient comments” on individual provider pages and focus on real expertise metrics).

• Identify where this data lives today: Some of it might be in your CMS, in provider finder platforms, in finance/billing platforms (e.g., CPT codes to determine the provider’s true expertise – e.g., orthopedic surgeon billed for 10 knee replacement surgeries in one month vs another orthopedic surgeon who only billed for 1), credentialing, EMR, etc.

• Make sure this information (all of it) is discoverable on your website. Right now, the agents are (and will be) capable of searching the internet on behalf of people – patients, consumers, etc. If the agent can’t find it, it doesn’t exist. Therefore, information the agent needs to do the research and make the transaction also needs to be website-ready. It is important to ensure that your data is presented on your website in a machine-first and machine-ready way.

• Ensure your website can facilitate transactions/conversions. How will agents book appointments? How will they log in to patient portals? How will they help facilitate payment (e.g., copays, etc.)? It’s time to start thinking about this now. Determine how an agent can log into a patient portal on behalf of a patient and/or how to book an appointment either as a logged in patient or a non-logged-in patient. Tricky? Yes. Will it be expected in the near future. Probably.

I’m pretty interested in seeing how these agents will change our shopping and consumption behaviors. I’m also interested in seeing how this will change healthcare, too. This industry cannot continue to disappoint consumers with poor experiences. Healthcare is hard. It’s only going to get harder. But it’s no excuse not to work toward finding parity with what consumers are experiencing (and will continue to experience) in the world outside of the healthcare industry.

Who wants to create a consortium of organizations to start actively working on pushing the industry to make a radical change? (In fact, a consortium of healthcare organizations approaching OpenAI, Google, etc., would go a lot further than individual organizations trying to establish one-off relationships with these organizations. Just saying…)

Who will be the first mover?

PS: I used the following prompt to get a comprehensive list of data within a health system – sharing if it would be helpful:

You are a health system executive and you want to put together a report that showcases all of the different data points you have for your providers and information patients might need to find the right provider. You have a few data points already, including: a list of providers in your health system, provider specialty, where the provider is located, time slots and availability for the provider to book appointments or perform procedures. Provide a comprehensive list of other data points about a provider that a patient might need to make an educated healthcare decision (like booking an appointment or at least doing research on finding the best provider to treat his or her condition). Make the list as complete as possible and also indicate where in the health system this data would live.