Agentic AI in healthcare in 2025: Separating the near-term use cases from the hype

Agentic AI in healthcare has rapidly become the centerpiece of health tech discussions. We’ve been studying it because it’s an area of rapid growth and advancement—an outlier in this moment in healthcare, when stakeholders across the industry are struggling amidst rising expenses, workforce shortages, and patient experience frustrations. Agentic AI tools are being positioned as a game-changing approach to addressing many of the industry’s biggest challenges, given their stated potential to mimic human reasoning and action and take on straightforward tasks.

We can’t fit everything we’d like to discuss in one post, so we’re splitting it into two parts. This first part will focus on the opportunities afforded by agentic AI, and the second part will cover the challenges and risks.

What is meant by "agentic"?

We’ll get to the agentic AI opportunities soon, but first: Let’s get some clarity on definitions. You’re probably familiar with generative AI—it has been a hot topic in healthcare for about three years now. You might be wondering: What makes agentic AI different? The primary difference between agentic AI and previous forms of AI, including generative, is the level of autonomy expected of the “agents." In short, agentic AI is much more autonomous and can execute across a broader range of tasks that a human would normally complete.

It’s also worth distinguishing between a single AI agent and the “agentic” term, though they are often used interchangeably in conversation. A single agent is a program designed to perform a specific task on its own (e.g., review a visit transcript to identify billing codes). When you have multiple agents working in a networked manner to accomplish more complex tasks, then you get to the “agentic” framework.

The headline buzz versus real adoption of agentic AI in healthcare

There’s a ton of investment in agentic AI—both inside and outside of healthcare. This year, it’s looking like 10% of VC AI funding across all industries will be for agentic AI solutions. This is a technical area that’s growing quickly, and, perhaps surprisingly, that’s even more true for healthcare. The global agentic AI in healthcare market is predicted to surge from $538.1 million in 2024 to $4.96 billion in 2030 (a CAGR of 45.56%).

This might be a little surprising to you. After all, we are inclined to think of the healthcare industry as lagging other industries when it comes to adopting new technologies; stringent regulatory requirements; complex, lumbering existing systems; and general risk-averseness are inhibitors. On AI, though, we might be in the middle of a sea change. Healthcare is making more investments than any other industry, with healthcare deploying AI at a rate more than double that of the US economy overall. Already we’ve seen that 22% of healthcare organizations have implemented domain-specific AI tools—that’s a 7x increase from last year and 10x from the year before that. Implementation cuts across the entire industry: 27% of health systems, 18% of outpatient providers, and 14% of payers have adopted agentic AI solutions. (Life sciences companies are lagging, probably because they are more inclined to go in-house with model development given how much internal data they have.) To put that in context, only 9% of companies in the broader economy have implemented AI, and most of those are using tools like ChatGPT for enterprises rather than investing in purpose-built solutions. (I would guess this survey doesn’t capture how many employees of companies outside of healthcare are using AI-powered tools to facilitate their own work, even if their organization overall hasn’t yet made the investment.)

The buying cycle has sped up for health systems, too—they’ve gone from 8 months for IT purchases to 6.6 months. Payers, by contrast, have seen their buying cycles lengthen from just over 9 months to just over 11. Amidst this acceleration, it’s important to be able to think critically about potential IT purchases—particularly when they involve new and therefore relatively untested technologies (like agentic AI)—before getting locked into a contract. But that’s easier said than done. Even as buzz for agentic AI grows, it’s still hard to separate the more aspirational from the observable—at least right now. According KLAS, research identified only 17 real-world healthcare agentic AI use cases, whereas it found 3,157 cases for generative AI.

The sheer quantity of activity in this area can make it hard to separate the hype from the present impact, so let’s see how the big players are investing.

Investments by stakeholder

Agentic AI is drawing interest from all segments of healthcare, though each is pursuing slightly different goals, with the overlap being that agentic AI has high potential to streamline the administrative functions that pose a widespread challenge. Regardless of stakeholder, the main ready-for-primetime use cases tend to be administrative in nature, while the longer-term view folds in clinical potential.

Payers

On the payer side, an initial focus seems to be on accelerating prior authorization, with efforts to advance agentic automation in claims review processing (not to forget payment integrity). We are seeing investment in areas like automating sales processes and streamlining underwriting. We also see solutions coming on the scene to facilitate prior authorization and review claims. Still speculative use cases could involve recommendations for network design and automated outreach to providers to address network gaps.

Providers

For provider organizations, the current investment theme is operations, while the speculative use cases include potential clinical breakthroughs. Presently, there are back-office efficiency opportunities: revenue cycle automation, inventory management, and provider scheduling (where shifts are adjusted based on predicted patient volume among other factors). Admittedly, many of the momentum behind provider adoption initially surrounds efforts to pushback against payer adoption of AI not to mention improvements that align with the patient experience using voice agents.

Life Sciences

On the life sciences side of things, the general theme is accelerating research and production. Agentic AI could help speed up research by facilitating recruitment for clinical trials; automating the paperwork needed to be in compliance with regulatory requirements; and monitoring and reporting adverse events. Plus, it could be used to identify new compounds for medications faster than ever before. A still-speculative use case: using agentic AI in service of personalized medicine, with agents reviewing patient records to make highly individualized recommendations.

Patients

We can't forget the patient experience angle, either, even though patients themselves aren't purchasing agentic AI tools. Patient-focused agents could improve the healthcare user experience. We're seeing a lot of automated call center startups alongside well-established companies. They have strong backing in that they speed access to an appointment, are available 24-7, and can conduct real-time triage to get the patient routed to the right place.

What big tech players are doing

Let's also take a look at the investments we're seeing big tech players currently making in agentic AI. So far, Big Tech investments have largely focused on enabling provider and life sciences organizations.

As we've covered, this is a rapidly growing and shifting market, but it's noteworthy that some of the biggest tech companies—in fact, the two largest companies in the world (Nvidia and Microsoft)—are investing heavily in agentic AI for healthcare. These large companies with ample time to wait to see a payoff appear to be diving into agentic AI for clinical and research use cases more aggressively than smaller companies, which primarily focus on the administrative use cases.

• Nvidia: The company has many different partnerships in healthcare; at the start of this year, it announced multiple partnerships to accelerate research and drug discovery via agentic and generative AI. And, in late 2024, it made its first non-biotech or pharma investment in an agent-focused company: $17 million in Hippocratic AI (which just this week announced a $3.5 billion valuation after raising $126 million in a series C funding round).

• Microsoft: In addition to forging major partnerships with EHRs and other companies, Microsoft is working on its own agentic AI solutions. Developers and researchers can use its healthcare agent orchestrator to configure multidisciplinary agents to suit their needs, with the quintessential example being multiple agents working together as a virtual tumor board (e.g., agents to analyze radiology and pathology data, stage the cancer, reference clinical guidelines for treatment planning, assess clinical trial eligibility, and draft reports).

• GE: The corporate giant is working on an agentic AI for radiology solution—moving from just analyzing images to supporting radiologists in generating reports.

• Palantir: In contrast with these other big players, Palantir's stated investment in healthcare-specific agentic AI is administratively focused. In March 2025, they announced a partnership with R1, the revenue cycle management giant, to launch a research lab focused on developing agentic AI approaches to revenue cycle automation.

Big companies have more runway to make investments that don't have near-term ROI. On the other end of the spectrum, smaller companies are investing in agents, too, but more often for cases with likely near-term payouts, like improved back-office functionality (rather than high-risk clinical opportunities)—revenue cycle automation and agentic call centers for both clinical and billing questions. Lots of startups tout agentic AI call centers, which figure among the lowest-hanging fruit for agentic AI (because of the time and labor-saving elements and the extent to which the tasks are often repetitive), though there are other high-value applications in life sciences/pharmaceuticals coming on the scene.

What major EHR vendors are doing

Epic and Oracle Health (formerly Cerner) are leading the agentic AI charge among EHR vendors, and both made announcements this year about their commitment to agentic capabilities.

• Oracle Health announced a commitment to voice-driven navigation and agentic capabilities, while also piloting ambient scribe and revenue cycle applications (i.e. prior authorizations initially). Perhaps more notably, Oracle announced a new AI-powered EHR for U.S. ambulatory providers in August 2025 with plans to launch an acute-care version in 2026. Agents are expected to figure prominently in these offerings in future.

• Epic rocked the market at its UGM with announcements that it plans to launch numerous AI capabilities, including targeting some existing market players’ businesses (ambient scribe technology) but also expressing plans to expand into agentic capabilities. For example: a “digital concierge” agent to answer patient questions via myChart and a provider-facing agent to answer billing codes and claims management questions.

While Epic and Oracle are leading the pack, smaller EHRs have also ventured into agentic AI.

• Athenahealth: Has a marketplace where clients can shop for tech solutions developed by partner organizations as add-ons to the EHR—some agents are in the marketplace (e.g., RCM automation, contact center, front desk).

• NextGen: Has a SOAP note generation tool that appears to edge into agentic territory.

• Meditech: Announced a commitment to agentic AI but have been vague about the specifics.

• eClinicalWorks: Rolled out an agent-powered contact center through a partnership.

None of their announcements landed with the fanfare of the two market leaders, though, because the investments aren’t yet as far-reaching as those we’re seeing with Epic and Oracle.

Takeaways: Some improvements, lots of VC deals, but watch the hype

We are seeing vendor tools proliferate and show promising early outcomes. Still, taking solutions from exciting pilots to true scale within an organization is proving an elusive outcome, at least so far. We’re very much still in the early stages of agentic AI implementation, and the extent to which this technology moves from headline-grabber to technological mainstay remains to be seen. For example, a HIMSS Market Insights survey found that only 18% of organizations feel ready to use AI in care delivery, with IT infrastructure being a major barrier to implementation.

While the hype cycle remains large for agentic AI, there are a series of headwinds for adoption related to risk and implementation challenges. Stay tuned for more details to come in part two.