More healthcare leaders invest in agentic AI, but gaps emerge

More healthcare leaders are investing into agentic AI, but they aren’t all moving at the same pace.

The vast majority (85%) of healthcare leaders say they will increase their investments in agentic AI over the next two to three years, according to a Deloitte report released last week. Over half of those leaders (61%) say they are building agentic AI initiatives, the survey found.

But the survey also shows some are adopting agentic AI tools more aggressively, and the survey suggests that those who move are better positioned to see more efficiency and savings.

Dr. Jay Bhatt, managing director at Deloitte, tells Chief Healthcare Executive® that the report found three groups when it comes to agentic AI. The “early adopters” are building and implementing, “starters” are securing budgets, while “watchers” are waiting for more evidence.

“Our research points to early adopters expecting a greater than 20% savings, far more often than watchers,” Bhatt says.

By contrast, those moving more slowly, the watchers, are projecting savings of 13%.

Deloitte surveyed 100 healthcare leaders, with 50 coming from health systems and 50 coming from health plans.

Beyond tech pilots

Health systems are using agentic AI to help patients with checking in and can also answer questions about billing and lab results. But organizations are also looking at agentic AI in other ways, including more personalized outreach to patients.

“Leaders are increasingly treating it as an operating model change, and not a tech pilot. So you know, it's not just generating text. It can plan, sequence, execute multi-step tasks,” Bhatt says.

In addition to easing some headaches for patients in navigating their health systems, Bhatt says agentic AI is also reducing some work from staff.

“Leaders are still under pressure on cost, workforce strain and access, and consumer experience and agentic AI are showing up as a practical way to redesign the work, not just patch it,” he says.

There are still some obstacles to adopting AI, but the survey suggests some of those hurdles are getting a bit smaller. Four in ten (40%) say technical talent limitations are no longer a major issue, and more say they are seeing less resistance to change.

“Adoption friction is easing, particularly for those organizations that are thinking about this as part and parcel to their strategy and operations and having an intentional approach to … addressing adoption by the workforce,” Bhatt says. “So as adoption friction eases, then scaling becomes easier.”

Health systems are also more likely to see success in their transformation efforts if they are developing AI solutions with clinicians.

“When you co-design, engender trust and have governance and approaches to test, iterate and then scale and do that together with various professionals, both administrators, clinicians and other functions in an organization, you get better results and will be in a better position for being successful on adoption and scaling,” Bhatt says.

Bhatt also suggests the message to doctors matters. By explaining to doctors that incorporating AI reduces their cognitive load on administrative tasks and offers more time for work that matters, organizations are likely to build more support.

“In my clinical experience, you are able to address care gaps, do care coordination, and do a complete workflow using agents in a way that not only is a better experience, but it brings more time for me in front of my patient and family and to explain and work through decisions together that could be complex,” Bhatt says.

Bhatt also points to the use of AI tools to identify trends and intervene earlier.

“Risk isn't static. Risk evolves and changes day to day, week to week, month to month,. Risk is dynamic,” he says.

“And so what agents can help us do, if it's part of an end-to-end orchestration, you can see things as they're emerging before they become high risk and can act on it,” Bhatt continues. “That’s, I think, some of the power of the opportunity and the technology, and that's why you see better outcomes, lower cost, and the ROI in some of these places beyond, some of the revenue cycle, back office piece.”

‘Start with quick wins’

For those organizations that have been waiting to adopt more agentic AI solutions, Bhatt suggests picking one area, such as following up with patients after they have been discharged from the hospital.

He also sees the potential for health systems and health plans to work together to use agentic AI tools to improve care coordination and help reduce gaps in care.

“Pick one end-to-end workflow, invest there, see the impact, and then that could help free up resources to continue growing,” he says.

“The smaller, cautious organizations should not try to boil the ocean,” he adds. “Start with quick wins that don't require deep re-architecture. And then you could build these reusable components through strategic initiatives. So the sequencing should match the complexity and the AI maturity that organizations have, and I think that can help.”

Healthcare organizations need strong governance for agentic AI. Health systems need to ensure they are using AI tools responsibly and ethically, with guardrails for data quality, privacy, transparency and accountability. And he stressed the need for feedback on governance of AI from staff across the organization.

Bhatt says he’s encouraged by the data showing more thoughtful uses of AI>

“I see this as an opportunity to improve care, improve operations, improve outcomes and for organizations to deliver on their mission for the health of communities and the health of their workforce,” he says.