Agentic AI

A branch of the HHS that funds breakthrough and high-risk research, awarded contracts to six teams to build an FDA-authorized, autonomous artificial intelligence tool to treat patients with heart failure.

The Advanced Research Projects Agency for Health will dish out $33.7 million in the first year of the four-year, $62.7 million contract. Recipients include Kaiser Permanente, Duke University and Stanford University, as well as health AI vendors Atman Health, Tempus AI and Updoc. 

Teams will build a patient-facing AI agent to serve as a “digital member of the clinical care team,” a second supervisory agent that oversees it and a deployment plan to scale the tech in health systems, according to ARPA-H.

The Advocate program, first announced in January, aims to address preventable deaths from heart disease — the leading cause of death in the U.S. —  and fill cardiac care gaps in rural areas, ARPA-H said.

Nearly half of U.S. counties do not have a practicing cardiologist, and those that don’t often have higher rates of cardiovascular disease.

ARPA-H says it wants to tackle those gaps by funding an AI tool to provide 24/7 autonomous cardiovascular care that patients can use between visits to real clinicians. The AI agent will essentially be part of the patient’s care team, fielding questions independently and only engaging human providers when necessary.

The program is split into three focus areas, with each contract awardee working toward a shared goal of improving cardiac care with AI. 

Still, ARPA-H’s programs are fueled by competition — total award amounts are contingent upon meeting aggressive research milestones, and those who don’t reach them can be kicked out. It’s a model ARPA-H got from the Defense Advanced Research Projects Agency, which led to technological breakthroughs like the internet.

Under the first focus area, three health tech companies will compete to create an AI agent that will be submitted for FDA authorization in two years.

Atman Health, a health AI company focused exclusively on cardiovascular care and chronic disease management, plans to use up to $7.7 million in Advocate funding to advance its large language models with a voice-based interface that heart failure patients can talk to.

Telehealth company Updoc will use $9 million to develop its own conversational AI assistant, while precision medicine company Tempus AI will leverage $9.5 million to extend its existing patient health app and add continuous monitoring features to detect changes in a patients’ health.

Under the second focus area, researchers at Stanford University will build a system to identify any unsafe recommendations produced by the patient-facing AI tool.

Teams from Duke University and Kaiser Permanente will use their funds under the third area to map out a deployment strategy.

North Carolina-based Duke, awarded $15.5 million, will create a platform to test the agents across five health systems on electronic health records systems from Epic and Oracle.

Kaiser Permanente will use $16.3 million to deploy the agents into its Epic EHR workflows for heart failure patients across 21 medical centers and more than 260 clinics to test its safety and efficacy.

ARPA-H has also contracted with the Johns Hopkins University Applied Physics Laboratory, which will serve as an independent evaluator of the tech.

Additionally, program leaders said they’ll work alongside the FDA throughout the program to develop a regulatory framework for this class of clinical AI.

The FDA has not yet approved an agentic AI tool for clinical care. If this program is successful, it could create a regulatory pathway for future applications of clinical AI, which the Trump administration has advocated for in recent years.

Artificial intelligence has dominated much of the conversation on healthcare technology over the past three years.

That trend continued at the HIMSS conference in Las Vegas in March 2026. The sector is hopeful AI could assist with major challenges like workforce shortages, burnout and an aging population that needs more care. 

As the technology rapidly advances, the industry is increasingly adopting AI agents that can act more autonomously and updating their governance strategies to keep up with the rapid pace of innovation, experts said at HIMSS. Those agents could be a cybersecurity risk, and health systems will need to pay close attention to how they verify non-human users of their systems. 

Plus, electronic health record vendors are doubling down on their own AI offerings — though the companies say their provider customers will still be able to integrate third-party AI products into their records systems. 

These are the five biggest takeaways from the conference:

1) The rise of AI agents

AI agents, or AI systems that can act more autonomously and handle complex, multi-step tasks, will increasingly be deployed in the healthcare sector this year as the technology advances, experts said. 

The industry faces a significant burden of back-office work that could be handled by agents independently, like onboarding new workers, managing the claims and reimbursement process, and helping patients schedule appointments and navigate their care, said Aashima Gupta, global director for healthcare solutions at Google Cloud. 

That work is costly and labor-intensive. Clinicians have long said administrative tasks distract from patient care, drag into after-work hours and increase burnout. And administrative work is expensive for the industry as a whole, accounting for up to one-quarter of healthcare expenditures in the U.S., research shows. 

Now, agents are rapidly becoming more advanced in their reasoning and decision-making capabilities, said Puneet Maheshwari, senior vice president and general manager of Optum Real, Optum’s new AI system for claims reimbursement.  

Healthcare reimbursement and claims processing often takes a long time to complete, he said. Clinicians document the care they provided, and coders check their work for errors — potentially sending those encounters back to the provider if they flag concerns. Claims clearinghouses might also return claims if they see missing information or mistakes. 

And by the time claims arrive at payers, weeks may have elapsed. It could take another 28 days for providers to get paid, he said. As AI agents advance, they could execute a multi-step workflow on their own, experts said.

“Autonomous agents will become more and more proficient and will be utilized more and more across the board,” Maheshwari said. 

2) Interoperability isn’t optional

The ease of health data exchange has been a long-term challenge for the healthcare industry, complicated by siloed information that doesn’t easily flow between providers and a reliance on dated technology like fax machines.

Interoperability is key if healthcare organizations want to take advantage of AI tools, as the technology is only as good as the data it’s trained on, experts said. 

As the sector moves to adopt agents, providers are being forced to take a hard look at their data, said Edwina Bhaskaran, chief clinical systems and informatics officer at Mayo Clinic. They need to examine data quality, where it’s stored and whether agents can easily pull information from different sources. 

“This happens quite often, you still have to have someone call over and say, ‘Can you fax me records?’” she said. “So while there’s a lot of data that we do share, there’s still quite a bit of data that we don’t.”

The federal government is also prioritizing data liquidity, Dr. Thomas Keane, assistant secretary for technology policy and the national coordinator for health information technology, said during a presentation at HIMSS. Treating patients without a full picture of their health background can lead to medication errors, care gaps and adverse events, he said.  

Plus, the wealth of data generated in the U.S. healthcare system can now be effectively analyzed with AI to improve diagnosis and treatment, pushing healthcare to focus on interoperability, said Greg Samios, CEO of information services and software firm Wolters Kluwer Health. The sector has often been overwhelmed by data that couldn’t easily be used. 

“Grassroots pressure often drives this more than anything else,” he said. “And now that there’s an ability through a lot of the AI technology to leverage this data to make a real difference for your life and my life, I feel like that pressure might force that issue a bit more.”

3) EHR vendors tout AI tools

Electronic health record vendors are rolling out suites of AI tools to their customers, adding new competition for startups and health tech firms. 

Providers are already reliant on EHRs, so it’s likely easier for them to implement products from their vendors. More than 85% of Epic’s customers use their AI products, including tools for drafting discharge summaries, documenting patient care and revenue cycle management, the market-leading EHR vendor said at HIMSS. 

Meditech, another large EHR vendor, is also expanding its AI offerings, including an ambient scribe for physicians and nurses, an agent that can help manage claims denials and appeals, and a tool that lets clinicians query patients’ records. 

EHRs are key to physicians’ workflow, so integration with the health record is often important for adopting AI products, experts said. Plus, the EHR already holds the health system’s data, so it can offer a better quality output, said Sarah Hatchett, chief information officer at Cleveland Clinic.

Competing AI products will still be able to integrate with their EHRs, leaders at Epic and Meditech said. Epic is building capabilities, but the firm also wants to “give our customers the choice” and keep options open if they prefer another product, said Seth Howard, Epic’s vice president of research and development. 

Meditech also allows health systems to integrate other AI tools with their records system, said Chief Operating Officer Helen Waters.

“If I’m asking you for an investment of millions of dollars, you should ask of me the ability to take that investment and do what you choose with it,” she said. “We’ll have a free opportunity to compete for their business, but we cannot control their business, neither should we.”

4) AI governance shifts from priority to practice

Although experts have emphasized the importance of AI governance in the past, vendors and health systems this year said they were focused on evolving, refining and innovating their AI governance structures.

As AI tools mature, healthcare organizations need to evolve their AI governance structures, according to Jane Moran, chief information and digital officer at Mass General Brigham. Static governance is not an option in the face of rapidly evolving AI tools.

“We’re finding these tools are super easy to use. You can go out and just get OpenAI or Claude or Perplexity or whatever and start to use it. And so, what we found in terms of governance was that we actually needed to do more education with our employees about safety and security and keeping our patient data private,” Moran said in an interview.

Dr. Ed Lee, chief medical officer at Nabla, an ambient AI scribe for clinical documentation, also stressed the importance of having a governance structure with the ability to adapt to new innovations.

There should be different tiers of governance for different use cases, Lee said. Administrative tools do not necessarily need the same guardrails as those that touch clinical care.

The rise of agentic AI has further amplified healthcare's need to maintain nimble governance structures.

“Especially as we’re talking and moving into things like agentic AI, where things are potentially happening in an autonomous way, there may be even further guardrails so that we don't do the wrong thing,” Lee said. “We always have the patient’s wellbeing in mind and safety and quality are of most importance.”

Throughout HIMSS, experts discussed the need to level up AI governance in healthcare and expanded on the tangible ways they are doing so — with structured frameworks, tiered guardrails and user-friendly tools.

“What I see in different organizations is a reflection of maturity and the discipline to do the mundane tasks that establish foundational elements of governance,” Skip Sorrels, field chief technical officer and chief information security officer at cybersecurity company Claroty, said in an interview. “And the ones that do that well, I put in the category of quite mature and usually having a higher level of governance, versus the opposite end of the spectrum, where it’s like, ‘Yeah, we have governance, but we haven’t done anything with governance.’” 

“And unfortunately, I do think there are still some out there that kind of have their heads in the sand,” he said.

5) Healthcare faces the challenge of AI agent identity management

Healthcare is confronting the rise of non-human identities in its AI systems, as AI agents are increasingly deployed in administrative, clinical and operational settings. As these agents gain autonomy, robust identity management programs are needed to prevent unauthorized access to sensitive systems or data. 

Russell Teague, chief strategy and security officer at Fortified Health Security, a company that offers advisory services and threat defense capabilities, said he’s still waiting to see how agentic AI plays out.

“From a security perspective for healthcare, we’re going to have to pay close, close attention to identities, both human and non-human,” Teague said.

Vendors are already trying to corner the agent identity management market. At HIMSS, the digital identity and access management company Imprivata introduced Agentic Identity Management, a new set of capabilities in its platform designed specifically to govern AI agents.  

“It’s a complex need to have someone not from your system who’s not your employee come in. And you don’t want to give them the keys to the castle. So, you have to do things like govern exactly how and when and where and why they're going into a system,” said Dr. Sean Kelly, chief medical officer at Imprivata and emergency physician at Beth Israel Deaconess Medical Center.

“And that becomes very important when you start talking about non-human ID. So now if you have agentic agents doing things in healthcare, it's a very complex and messy situation.”

Healthcare is hastily adopting agentic AI tools, Kelly said, given the tech’s promise of enabling efficiency and better care outcomes.

“The problem is that leaves you very vulnerable because it's even harder than human identity to control and to understand,” he noted.

Sponsored content

By athenahealth

A few months ago, a team of our engineers, designers, and product managers set up in a break room at practices in Louisiana and Mississippi. Between patient visits, physicians wandered in, grabbed a chair, and started talking. They weren't reviewing a polished demo or reacting to a roadmap presentation. They were explaining where their day broke down, what slowed them down, and what they wished technology would do differently. The room stayed loud most of the day, equal parts clinical complaint and technical debugging, and by the time anyone left, the products and features being described weren't sketches. They were already being built, right there, so physicians could watch features evolve in real time, challenge assumptions, and push the team in directions they never anticipated.

In that room, clinicians weren't simply evaluating our work. They were doing it with us. That's the difference between feedback and co-development, and it's how trust in technology actually gets built.

The healthcare industry spends a lot of time talking about how quickly AI is advancing. I think we're asking the wrong question. It's not how fast AI models improve. It's how fast clinicians are willing to trust them.

Trust should determine the pace of AI

Research continues to reinforce this reality. Recent surveys from organizations like the AMA, KLAS Research, and Wolters Kluwer consistently show that clinicians want transparency, governance, and meaningful involvement before AI takes on higher-stakes decisions.

We're experiencing this firsthand. Once we saw that physicians trusted ambient tools to accurately capture a conversation, we could then layer in a next step: pulling together the 10 or 15 documents that can follow a single hospital discharge into one summary, with a link straight back to the source, so a physician isn't hunting through 50 pages to find the detail that matters. From there we could add an intelligence layer, which would allow a physician to ask a question of a chart or an inbox the way they would ask a colleague.

Like stepping stones, each innovation depends on the one before it holding steady, earning enough trust, before the next can be reached. Because of this, we deliberately don’t build based on what the models are capable of; we build based on what clinicians are ready to allow AI to augment next. Trust that can’t be verified is akin to no trust at all.

Co-development isn't another word for customer feedback

"Co-development" has become one of those nice-sounding phrases every health IT company uses. Too often, though, it means inviting clinicians to react to something that's already been designed. That's valuable feedback, but it isn't co-development. Real co-development means clinicians, engineers, designers, product managers, and data scientists solving problems together before the solution exists. When those conversations happen together instead of sequentially, you go way beyond building features – you solve problems. The result is software that clinicians recognize as inherently helpful, as something they helped create.

The best feedback is the feedback that changes your roadmap

Against all reason, we find that the most valuable moment in co-development isn't when clinicians validate an idea – it’s when they reject one. Some of our favorite concepts have died in co-development sessions. It's far more valuable to learn at the iteration stage what will, or will not, pass the adoption bar. Launching an elegant feature that no one trusts or adopts is far more expensive than abandoning a promising idea before it's built.

Those conversations also reveal nuances we rarely would have discovered on our own. Our clinicians push the gas in some areas and pump the brakes in others, and it's not always in areas we might presume. They'd like us to go further on personalization, but hold back on automation. They might advocate hard for ambient documentation, but then request an override option so they can still dictate appointment notes word for word. A survey or a focus group captures just one of those directions. Co-development lets you hold both at the same time and build the tension directly into what ships – and what you hold back.

Building AI clinicians actually want

Technology succeeds in healthcare when it fits naturally into the way clinicians think, work, and care for patients. And that only happens when clinicians help shape it from the beginning. The companies that win won't necessarily be the ones with the biggest models or the fastest release cycles. They'll be the ones that consistently earn the confidence of the people who rely on their products every day. Technology may power the future of healthcare, but clinicians will determine whether that future is trusted.

Amazon Web Services rolled out a suite of agentic artificial intelligence tools in March that aim to handle a range of healthcare tasks, like helping patients schedule appointments and summarizing medical data for clinicians. 

The product, called Amazon Connect Health, includes five capabilities: verifying patients’ identities; handling appointment scheduling; creating summaries of patient medical histories; creating clinical notes based on conversations between clinicians and patients; and generating medical codes from clinical documentation. 

Amazon Connect Health should help patients more easily access care and assist with clinicians’ administrative work, according to Naji Shafi, general manager and director of healthcare AI at AWS. “Our healthcare workers are overburdened, drowning in administrative complexity, and it’s costing everyone,” he said.

The agents — AI systems that can act more autonomously and handle complex tasks — include functionality geared toward patient engagement as well as tools to help providers prepare for appointments and document care, according to AWS.

The patient-facing agents are focused on assisting patients as they schedule appointments with providers on the phone, Shafi said. For example, a patient could call and ask to see a specific doctor about knee pain, and the tool could confirm the patient’s identity, check their insurance, review their doctor’s availability and schedule the visit. 

Health systems can customize their patient verification requirements and appointment scheduling capabilities based on their preferences. For instance, a provider might want the tool to only schedule an appointment if the patient’s insurer has approved a prior authorization request for an expensive service, Shafi said. 

The scheduling tool is also trained on examples of frustrated patients and complex cases, so the agent can route these calls to a person if necessary, he added. Additionally, if a caller says they want to speak to a human, the tool sends the conversation to a human worker. 

“If I’m getting frustrated, our AI agent will detect that, automatically escalate and bring in a human, and all that context will carry over,” he said. 

Amazon Connect Health also includes functionality geared toward providers. One tool synthesizes patients’ medical histories from electronic health record data and health information exchanges, so clinicians don’t have to dig through charts to find relevant information, AWS said.

Additionally, the suite of tools includes an ambient documentation assistant, which generates clinical notes after recording conversations between providers and patients. The offering builds on Amazon’s HealthScribe tool, launched in 2023, by adding the ability to automatically populate after-visit summaries and suggest diagnosis and billing codes, according to AWS.

The provider tools also include guardrails — a key concern for experts who worry about inaccurate or misleading responses that could cause patient harm.

A human clinician needs to review documentation and codes before they’re approved, Shafi said. And any time the AI is summarizing information or taking an action, users can see the source information for the suggestion.

Strong evaluation of the models is key too, Shafi said. AWS uses clinicians on staff to analyze AI outputs, and the company has also built what it calls a “large language model as a judge” — a separate AI that can critique other models. 

“We’ve built it up so it’s highly correlated to humans,” Shafi said. “So that enables us to have a secondary system that is looking at the model outputs, and it’s trying to say, ‘Well, is this right? Is this good? Is this accurate?’”

Healthcare organizations are rapidly adopting agentic artificial intelligence, but governance frameworks for managing the technology are not evolving at the same pace, according to a new survey from digital identity security firm Imprivata.

More than 85% of leaders responsible for AI strategy at healthcare organizations say they’re confident they have visibility into AI agent activity and are able to fully control and govern an autonomous agent’s actions.

However, 72% of respondents admit AI tools are deployed without IT approval at least occasionally within their organizations, highlighting a disconnect between confidence and reality.

Healthcare organizations are embracing agentic AI while simultaneously struggling with fundamental governance and oversight challenges, according to the report. 

Agentic AI are AI systems that can act largely independently, making complex multi-step decisions and solving problems with minimal human intervention. The healthcare sector has invested heavily in agentic AI, especially for back office use cases such as prior authorization and revenue cycle management automation.

More than a quarter of leaders have already implemented agentic AI at their organizations, according to market research firm Vanson Bourne, which conducted the survey on Imprivata’s behalf. An additional 44% are piloting or conducting proof-of-concept projects and another 21% plan to implement it in the next year.

The vast majority of respondents expect agentic AI to have a “transformative” impact on clinical and operational workflows.

However, agentic AI presents newfound security risks — especially as use cases expand, giving agents access to a wider variety of internal systems, experts say. That’s because, unlike traditional software that relies on user input to act, agents can act autonomously, traversing different information systems and executing actions on behalf of users.

“If an agent has excessive permissions, operates outside its intended scope, or takes a high-risk action without appropriate oversight, the consequences can directly impact care delivery,” said Dr. Sean Kelly, chief medical and growth officer and senior vice president of customer strategy at Imprivata. 

“An agent could access or expose sensitive patient information, enter incorrect information into a medical record, alter a medication or dosage or act under a clinician’s authority in a way that the clinician never intended. Because agents can operate autonomously and at machine speed, a single error could also propagate before someone recognizes what is happening,” Kelly said.

More than half of respondents ranked security among their top concerns when adopting agentic AI. Although leaders expressed confidence in their ability to govern AI agents, instances of shadow AI, when employees use AI tools that haven’t been authorized by their companies, are still rampant.

According to a separate survey by Wolters Kluwer, 40% of medical workers and administrators said they were aware of colleagues using unauthorized AI tools, while nearly 20% reported they have used an unsanctioned tool themselves.

Imprivata’s survey shows that organizations are taking fragmented approaches to managing AI agents. Sometimes information technology departments centrally manage the tools, while security teams manage others. Some organizations report ad hoc or unapproved deployments.

And healthcare leaders are still ironing out the specifics of AI governance and how much humans need to be in the loop when an agent is making decisions, according to the research. 

That process varies widely based on where the agent is deployed — whether in an administrative, operational or clinical setting.

“Oversight must match the level of clinical risk. Agents need clearly defined identities, permissions, and boundaries, with human review for higher-risk activities and an audit trail for accountability,” Kelly said. “The more autonomy we give these systems, the more important those safeguards become to detect and contain problems before they create broader operational or patient safety risks.”

The healthcare industry has long sounded the alarm about potential patient safety risks posed by improperly implemented AI. ECRI, a nonprofit focused on healthcare safety and quality, named insufficient governance of AI in healthcare as one of its top 10 patient safety concerns.

Various industry consortia have issued guidance to alleviate these concerns. The Coalition for Health AI, a network of thousands of healthcare systems and industry groups, issued AI governance playbooks last year to help health systems responsibly and safely roll out AI tools.

Google Cloud revealed several artificial intelligence partnerships with healthcare organizations, including for projects that summarize clinical notes and automate prior authorizations. 

The partnerships come as more healthcare and life science firms are deploying AI agents, or advanced tools that can more autonomously plan and perform tasks, according to a Google Cloud survey of 605 leaders. Forty-four percent of executives said their organizations were actively using agents, with 34% reporting they use 10 or more agents.

For example, Hackensack Meridian Health built multiple AI agents using Google’s generative AI technology, including a tool that can recap patients’ medical records for doctors.

The health system’s note summarization agent has helped more than 1,200 clinicians generate more than 17,000 summaries since it went live in June, according to a press release. 

The provider began offering a general summary of patient records, but has since been able to customize results by specialty — so an oncologist may see different, more relevant information compared with a urologist, said Sameer Sethi, SVP, chief AI officer, Hackensack Meridian Health.

“We all have been in situations where we go to our physician offices, and the physician is reading the screen while trying to understand and talk to you, right?” he said during the press call. “We’re trying to reduce that time that the physician spends on screen looking at the patient chart.”

The health system is also rolling out two other agents, including an assistant for nurses in Neonatal Intensive Care Units that provides information on best practices and policies, and a tool that summarizes lab results so primary care providers can more quickly draft notes to patients.

The technology giant has also partnered with IKS Health, a company that provides revenue cycle, clinical and other operational products, on an AI platform that detects when prior authorization could be required and helps manage the process.

Its multi-agent system includes a documentation assistant that creates clinical notes on patient visits, a coding agent that aligns billing codes with documentation and a tool that aims to detect when prior authorization is necessary. That agent works with payer systems to collect administrative and clinical data to support the prior authorization request, executives said.

The process also loops in a human worker to ensure the AI is accurate when medically necessary, IKS said.

Additionally, Google announced a partnership with virtual cancer care company Color Health to launch an AI assistant that helps women determine their eligibility for breast cancer screenings and schedule care. 

The CMS wants to deploy artificial intelligence tools to Medicare beneficiaries to help navigate their care, CMS officials said at the 2026 HIMSS conference. 

The agency is already using the technology to detect fraud. But the CMS also hopes to get the technology into patients’ hands, both to assist seniors and to hopefully bring down rising healthcare spending, which continues to outpace the rest of the economy. 

“The fundamental problem right now is that other sectors of the U.S. economy have advanced and been deflationary with their use of technology,” CMS Administrator Dr. Mehmet Oz said during a panel discussion. “Healthcare has remained inflationary.”

The federal government wants to introduce AI agents, or AI systems that can act more autonomously, to Medicare beneficiaries to help them find doctors or choose Medicare Advantage plans, Oz said. That tool should be available “by the time we’re done in this administration,” he said.

The biggest challenge is that Medicare beneficiaries don’t trust AI, Oz noted.

Only 31% of Medicare enrollees aged 65 and older trust AI “a great deal” or “a fair amount” to access medical records and provide personalized health advice, according to a survey published by health policy researcher KFF.

“No one has gotten to them with the use case of why it’ll transform their life for the better,” Oz said. “It seems like a tool that we use to market to them, or help hospitals deal with issues, but not necessarily their issues.”

One goal of the CMS’ Health Tech Ecosystem — an initiative launched in 2025 that aims to ease data exchange through partnerships with the private sector and increase availability of digital health tools — is to expand access to conversational AI to help patients navigate care and discuss their health, said Amy Gleason, acting administrator of the U.S. DOGE Service and senior advisor to the CMS, during the panel.

Some companies that have signed onto the Health Tech Ecosystem pledge have begun to release dedicated AI chatbots geared toward health queries, she noted. For example, OpenAI launched its tool in January, and Microsoft announced its Copilot Health assistant in March. 

Still, misuse of AI chatbots could be a risk to patients, given the tools can provide false or misleading information, experts say. Research on OpenAI’s consumer health tool found the chatbot often under-triaged cases, failing to send patients to the emergency room for serious health issues like impending respiratory failure.

Hallucinations are a concern, but doctors can also make mistakes, Gleason said. 

“The same will happen with AI, and they’re the dumbest today that they’ll ever be,” she said. “So I think we need to embrace this and help people understand how much this can help them.”

The Trump administration has also been using AI in healthcare to help the CMS detect fraud, said Kimberly Brandt, COO and deputy administrator at the agency. 

Fraud has become a significant priority for the Trump administration, which last month halted $259 million in Medicaid funds from Minnesota over what it called unsupported and potentially fraudulent claims. The agency is also imposing a six-month moratorium on Medicare enrollment for some suppliers of durable medical equipment, as part of its efforts to crack down on fraud. 

The CMS has been using an algorithm trained on previous Medicare providers that the agency has taken enforcement actions against, and tested new providers enrolling in the program against those bad actors, Brandt said. 

“We wanted to use AI to really help us be able to find out where the problems are,” she said.