AI agents are likely to be adopted in healthcare as a way to provide patient guidance in any modality and carry out tasks like transportation, appointment scheduling and follow-ups.

Those were a few takeaways from two healthcare leaders during an industry panel at Google Cloud Next. Richard Clarke, Chief Data and Analytics Officer at Highmark Health, and Sameer Sethi, Chief AI Officer at Hackensack Meridian Health, outlined what their organizations have done so far with generative AI and the groundwork in place for agentic AI.

"We're very excited for AI agents directly interacting with our members and patients with guidance that is always on in whatever modality they wish," said Clarke.

Sethi was also upbeat about the promise of agentic AI. "Imagine a patient calling to schedule an orthopedic appointment and also needing a wheelchair, a ride, a pharmacy visit—seven different actions handled by one AI agent," said Sethi.

AI in healthcare is a hot topic given that the industry is facing pressure on multiple fronts. Aashima Gupta, Global Director of Healthcare Strategy Solutions at Google Cloud, said the industry is using generative AI to alleviate administrative burdens such as paperwork and searching medical records.

Gupta said that AI in healthcare is also transitioning from simple chatbots to single purpose agents and ultimately multi-agent systems across multiple departments. The overarching goals are to reduce the burnout in the field and improve patient care.

"GenAI has really evolved from a buzzword to a business essential," said Gupta. "We're seeing a paradigm shift in how we interact with healthcare. Agents represent a strategic opportunity to reimagine care and journeys, underscoring that conversation is becoming the new interface. Our customers are able to reimagine patient care and how it could be personalized for everyone."

With that backdrop, here's a look at some of the takeaways from Highmark Health and Hackensack Meridian Health.

Highmark Health

Adoption across the enterprise. Clarke said Highmark Health already has more than 14,000 of its 40,000 employees regularly using internal genAI tools built on Vertex AI and Gemini. Highmark Health inked a 6-year strategic partnership with Google Cloud in 2020 and the two companies have worked together on multiple projects.

Ambient intelligence. Clarke said ambient listening by AI during patient visits is an underrated breakthrough because clinicians can focus on the person, not the notetaking. "Everything in the ambient listening space has been a true gift to bring joy back to practice for many of our clinicians," said Clarke. This point about ambient listening has surfaced before with healthcare leaders.

Multi-modality matters. Clarke said the ability of models to deliver in multiple ways is critical. "We were stuck on a particular use case and when Gemini 2.0 came out, it kind of made us get over the hump," said Clarke.

The vision for agentic AI. Clarke said the promise of agentic AI rhymes with his ambient intelligence points. The big idea is that AI agents can interact with members and patients to provide guidance in multiple formats on demand.

Cost concerns so far. "There was some concern that our cloud costs would be challenged, but we just haven’t seen that," said Clarke.

Guardrails. Clarke said Highmark Health puts AI projects into shadow mode followed by assisted audit before going into production with automation. "We need logging, monitoring, and governance," he said.

Hackensack Meridian Health

Sethi said his organization's genAI strategy aims at delivering personalized experiences, addressing efficiency, reducing burnout and offering disease prediction and precision treatments.

Hackensack Meridian Health started on BigQuery and Looker with Google Cloud before moving to Vertex AI and Gemini.

He added that there are a few use cases for AI agents that are appealing. "We want to make sure we were focusing on the patient and our workforce," said Sethi, who noted the following use cases:

  • A nurse agent. "The nurse can sift through large amounts of data and instead of going through binders or PDFs, the agent provides that insight directly," said Sethi.
  • Patient agent. This agent would be able to string together multiple tasks that are required during discharge processes or care that usually involve multiple people.

Now that use case for agents isn't ready just yet, but Sethi noted that the innovation is happening at a fast pace. Sethi said that these agents would require guardrails and testing. "We built a whole test suite," he said. "Every time we receive feedback; we add that to our test library and test against it in future releases."

He added that implementing AI requires a heavy dose of process optimization and change management. "The biggest barrier in technology enabling is actually humans," said Sethi, adding that human-in-the-loop is critical to determining how processes can be automated. "Figuring out what should be agentic is the hardest part," he said. "It’s easy to create an agent but identifying the elements where a human has to come in—that's what takes time."

More on healthcare transformation

 

 

 

Business Research Themes

Data to Decisions, Future of Work, Innovation and Product-led Growth, Next-Generation Customer Experience

Roles

Chief Information Officer

Tags

#AI #artificialintelligence #machinelearning #ML #generativeAI, @googlecloud