Illustration: Natalie Peeples/Axios
Google Cloud on Wednesday unveiled a suite of AI tools aimed at tackling health care's administrative and clinical inefficiencies, with two major health systems as partners.
Why it matters: Google is shoring up its use of generative AI tools in health care at a time when such systems are becoming table stakes for hospitals.
Context: The tech giant debuted AI tools it's positioning to exceed the capabilities of basic chatbots in reasoning, automating, and collaborating.
Inside the room: Google previewed updates to its Vertex AI and Agent Space platforms at a press briefing ahead of its Google Cloud Next conference.
- Agent Space features a centralized "Agent Gallery" of prebuilt, ready-to-use agents, including a "Deep Research Agent" to synthesize complex data sets and cite its sources.
- An "agent-to-agent" protocol allows different agents to work together across workflows and platforms — a nod to hospitals' fragmented IT stacks.
State of play: Pittsburgh-based Highmark Health and Hackensack Meridian Health, New Jersey's largest not-for-profit health network, are using the new tools.
- Highmark rolled out Vertex-powered tools to roughly 14,000 out of its 40,000 employees. It's using agents for call centers, prior authorizations and ambient documentation, per Highmark chief data officer Richard Clarke
- Hackensack is using Google AI for nurse agents, call center automation, and document summarization.
- The system is building use cases in six focus areas, including disease prediction and precision treatment, per Hackensack chief AI officer Sameer Sethi.
What they're saying "We're not just talking about incremental improvements," said Aashima Gupta, global director of health care at Google Cloud. "We're talking about fundamental shifts, big unlocks that will lead to powerful new use-cases across the entire health care value chain."
- Both Highmark and Hackensack have shifted from thinking of AI tools as departmental initiatives to thinking of them as part of their core business strategy, Clarke and Sethi said at the briefing.
- "I was chosen to have a dotted line into the CEO because this became a CEO agenda," Sethi said. "There's board oversight over this. We have pressures from from [top to] bottom and bottom [to] top."
Between the lines: The pace of innovation may be outpacing hospital risk tolerance.
- Clarke and Sethi emphasized the importance of human-in-the-loop designs and phased rollouts that move from "shadow mode" to full automation.
Reality check: Even the most advanced agent deployments are still early-stage, and the lack of integration with legacy systems remains a critical obstacle.
- Hackensack, for instance, has yet to fully implement the patient-facing scheduling agent it envisions. "We're not there yet," Sethi said.
- "The question of how you get these new technologies integrated in a way with those legacy systems is so critical," Clarke said. "We spend a lot more time thinking about that than we did before."
What's next: "Our expectation is that every single employee will be, at a minimum, using — but many building — AI, especially given the ease of some of the tools," Clarke said.
