Healthcare Brew covers pharmaceutical developments, health startups, the latest tech, and how it impacts hospitals and providers to keep administrators and providers informed.

Healthcare Brew

illustrated headshot of caroline catherman, healthcare brew reporter

Caroline Catherman

Healthcare Brew delivers the latest news and trends in healthcare.

healthcare-brew.com

About 44% of global healthcare and life sciences companies that use generative AI have added 

 to the mix, according to a survey released Oct. 16.

The survey, conducted by Google Cloud and global insights and strategy firm National Research Group, polled 605 leaders of global healthcare and life sciences companies from April to June to see how they use AI—in particular, AI agents—and what return on investment (ROI) they’re seeing. 

Healthcare most commonly uses AI agents for tech support (53%), followed by security operations and cybersecurity (49%), then productivity and research (46%), the poll found.Agents represent a new frontier because, as opposed to traditional large language models that take commands, they can autonomously work together to complete a task, Aashima Gupta, Google Cloud’s global director of healthcare strategy and solutions, said during a virtual roundtable on Oct. 14.

“This multi-agent framework is the evolution [of AI],” Gupta said.

About 1 in 3 surveyed organizations have deployed not just one but 10+ AI agents.

Mayank Pant, EVP of product management at care enablement company IKS Health, said during the roundtable his organization uses agents to automate the journey from an appointment to a prior authorization request.

 agent takes notes on an appointment. Then, an autonomous coding agent checks the notes for accuracy and assigns medical codes. If prior authorization is needed, a prior authorization agent steps in. The agents lastly bring a human into the loop to sign off on their work.

“What we are seeing on the ground is that this clinically connected…system is delivering better experiences for patients, for clinicians, their staff,” Pant said.

The push toward AI agents is accompanied by an increase in spending, as 74% of leaders said they’ve upped the amount they spend on generative AI and 48% said they’re reallocating money from other areas toward their AI investment.

The good news: It seems to be working. The majority (73%) of healthcare and life sciences executives said they’re seeing ROI in generative AI, compared to 74% in 2024.

Among 221 executives reporting ROI from AI, 83% estimated generative AI directly increased their company’s annual revenue by 6% or more.

When it comes to AI agents, however, the numbers get a bit more complicated.

Healthcare executives pointed to a few specific categories where they felt this newer technology was already making up for its 

Among 161 health execs whose company used AI agents, 34% said they saw ROI within patient experience and tech support. This was closely followed by software development at 33%.

Sameer Sethi, SVP and chief AI officer for New Jersey-based Hackensack Meridian Health, said during the roundtable it’s difficult to precisely link patient outcomes to AI.

Hackensack Meridian Health uses AI agents to call patients post-discharge to ensure they’re adhering to instructions. While Sethi said this was followed by a reduction in the number of readmissions within 30 days, there are other factors, like a patient’s support system, that also contribute.

“Attribution is going to be very difficult,” Sethi said. “What matters is that we’re doing the right thing for the patient and making sure that they’re healthier, and AI and other capabilities are helping with that.”

Correction 10/17/2025: This story has been updated to reflect that Mayank Pant, EVP of product management, was the speaker for IKS Health, not Sri Madabushi, EVP of health system partnerships.

Business hand handing money over to AI hand representing venture capital funding to AI startups.

Healthcare, life sciences companies are betting big on AI agents

From scribes to coders to prior auth, AI agents are starting to work in teams and deliver ROI.

Cassie McGrath

UnitedHealth Group, Cigna, and Humana all have something in common—besides being major health plans, that is.

They’re all currently facing lawsuits from patients alleging AI programs led to denied care. These suits claim (pun not intended) the insurance companies used AI-based algorithms to deny care without human oversight.

 of healthcare by leaders across the industry. AI scribes have 

 to give clinicians more face time with patients and better work-life balance, while also saving hospitals money. 

, able to carry out complex tasks autonomously, have turbocharged the tech’s potential.

 these tools are being adopted faster than safeguards, despite concerns about 

. Now, with all the emerging lawsuits, the outlook of AI in healthcare is looking even more complicated.

Fawad Butt, former chief data officer of Kaiser Permanente, UnitedHealthcare, and Optum, told Healthcare Brew there’s been a lack of oversight regarding “whether algorithms are behaving as good actors or bad actors.”

UnitedHealth spokesperson Eric Hausman told us it isn’t used for “adverse coverage decisions,” Cigna spokesperson Justine Sessions told Healthcare Brew AI isn’t being used to deny claims, and Humana didn’t respond to a request for comment.

But providers say the opposite—that plans 

 using AI for denials. While bigger health systems have more resources to appeal denials, some experts say independent rural hospitals

could be hit harder. A 2023 Stat investigation 

 UnitedHealth was using predictive algorithms to decide when some patients should be cut off from care.

Alexa McKinley Abel, director of government affairs and policy at the National Rural Health Association, said the research and advocacy organization is hearing rural hospitals are increasingly burdened by responding to prior authorization and claims denials because they have 

“If plans’ use of AI means even more prior authorization denials, then we would have concerns about the increasing paperwork and time associated with rectifying the denials for hospital staff as well as delays for rural patients that need to receive care,” she said.

The Centers for Medicare and Medicaid Services doesn’t require insurance companies to share how prior authorization decisions are made or how predictive technology is used in the process, so most of the evidence comes from providers, patients, or insurers that choose to self-report.

 from the National Association of Insurance Commissioners found 84% of 93 insurers used AI, with 50% saying they use it or are exploring its use to detect fraudulent claims, 68% for prior authorization approvals, 61% for disease management programs, and 45% for sales and marketing.

Only 9 of the 93 (12%) said they used or were considering using it to deny prior authorization requests.

Infinitus is an agentic communication platform that currently works with providers such as Medicare Advantage company Zing Health and Gramercy Surgery Center. The voice agent software can do things like conduct health risk assessments and handle specialty pharmacy prior authorization approvals for plans.

“The direction of travel for the industry is to introduce things that we fundamentally couldn’t do today because of a lack of human capital,” Infinitus CEO and founder Ankit Jain said.

On the other hand, healthcare providers and patients are using AI, too, for tasks like appealing denials. Startups 

 are an example of this, creating tools that collect patient and coverage information, then format and send a letter to the plan.

But insurers say some providers are taking AI too far.

Elevance CEO Gail Boudreaux alleged in a July earnings call “isolated pockets” of hospitals are using AI to “increase documented acuity,” driving up unit costs for the insurer. And Centene CEO Sarah London said in her company’s July earnings call that the insurer uses AI in response to that trend.

“As [hospitals] integrate AI into the revenue cycle, we’re integrating AI into payment integrity to make sure that we are sort of keeping pace with all of that,” London said.

Marc Douek, managing partner and co-owner at physiotherapy chain Renew Wellness, told us the chain cuts hours of work down to minutes by using a tool called Spry to automate prior authorization requests and appeals.

A 2024 American Medical Association survey found physicians and their staff spend an average of 13 hours a week completing prior authorization requests. It’s also an expensive process, as the healthcare industry 

 on prior authorization in 2023, according to the 2024 CAQH Index Report.

“Without using Spry in the automation…we wouldn’t be able to expand. We wouldn’t be able to treat our patients. We’d have to just turn people away,” Douek said.

Based on his conversations with insurers, Butt said providers have adopted AI faster—and insurers, which “haven’t adopted agentic [AI] as quickly,” now feel the tech has them “under the gun.”

“They’re left with old technology and humans responding to this deluge of appeals that are coming in from providers who just bought a software or an agent that essentially keeps appealing everything,” he said.

Butt is also co-founder of startup Penguin Ai, a platform intended to be a one-stop shop where both payers and providers can create AI agents. He believes both will need them in the future.

“What I predict is that it’s going to be agent wars galore,” he said.

Virtual Digital hand of Artificial Intelligence (AI). Healthcare technology and innovation concept. Hologram. 3d rendering

The AI arms race between insurers, providers has begun

Health plans have already been accused of using AI to deny patients. 

When you call up a company’s customer service line and are greeted by an off-kilter automated voice, how does that usually go?Personally, the call often ends with us yelling “Connect me to a person!” at a robot that just keeps asking the same questions over and over again.

If you relate, we have good news: That might soon be an interaction of the past.

 promise to flip that script by offering live assistants that remember and build off past queries to personalize and improve service. These agents can be online or even take on human-sounding voices to provide guidance over calls—and unlike chatbots, they can act autonomously without being prompted.

The even bigger news is organizations are already rolling these agents out in one of the domains where it’s most important to get things right: healthcare.

“I envision AI agents becoming the invisible infrastructure that handles background work…so clinicians can spend their time connecting with patients,” Max Solano, hospitalist and clinical informatics advisor at Jacksonville, Florida-based Ascension St. Vincent’s, told Healthcare Brew.

AI agents offer a “natural experience” and “open-ended conversation” compared to the sometimes frustrating 

 systems consumers have battled for decades, Israel Krush, co-founder and CEO of agentic AI company Hyro, told Healthcare Brew.

Patient scheduling company Zocdoc and financial platform Cedar are two healthcare companies that 

 agents earlier this year to assist with scheduling appointments and answering patients’ billing questions, respectively.

Sutter Health, a 27-hospital not-for-profit health system in California, is 

 health organizations to hop on the agent train, announcing Sept. 17 it will use Hyro’s tech to offer HIPAA-compliant AI agent-powered patient communications.

So how do you introduce an AI agent? Very slowly, Jennifer Bollinger, Sutter Health’s SVP and chief consumer and brand officer, told us.

It took about a year to choose a vendor and a bit over a month to integrate Hyro with Sutter’s 

, Epic, she said. Sutter is currently testing the agents for specific tasks, and Bollinger expects the first use case to become available to patients in about 60 to 90 days.

For now, she said the system will focus on using voice agents for basic tasks like authenticating identities, confirming appointments, and giving directions. Patients will be told they’ll be speaking with an AI agent and given the option to opt out, and more complex cases will be reserved for human staff.

Sutter eventually plans to bring Hyro’s agents online and have them take on duties like scheduling and rescheduling appointments, Bollinger added.

“Anything that we have standardized, that’s repeatable at scale across the organization, will be great use cases for the agentic AI,” she said.

To evaluate the tech’s success, Sutter will measure variables like time savings and whether the AI increases first-call resolution, Bollinger said.

These agents aren’t all good news, however.

Experts have warned AI agents could open up a health system to 

Agents are trained to serve, and hackers may try to order the programs to reveal private patient information, for instance, Healthcare Brew previously reported.

Krush said he understands those concerns and that the company kept this in mind when designing its agents, which he describes as “AI on a leash.”

“When you have guardrails in place, the AI agent basically knows what they’re allowed to do and what they’re not allowed to do,” he said.

Headphone of customer service. Support call, live advice, phone bot, chat headphone, service center, helpdesk advisor, virtual contact, care operator, microphone online, expert helpline 3d concept.

Healthcare’s new AI receptionist has arrived

California’s Sutter Health is one of the latest to adopt this new technology for patient communications.

Welcome back to AI 411, a monthly roundup of AI announcements from across the healthcare industry.

This month, agentic AI voice agents that can answer calls and facilitate back-office tasks, were all over the news. Here’s your September breakdown of health-related AI updates from around the industry.

On Sept. 2, Connecticut’s Hartford HealthCare selected 

Abridge’s ambient clinical intelligence platform

 for the health system after running a pilot program of the technology. The health system plans to use the company’s clinical documentation program to reduce administrative burden on clinicians.

Agentic AI company Assort Health announced Sept. 30 a $76 million Series B funding round led by Lightspeed Venture Partners, with support from investment companies Felicis, First Round Capital, Chemistry, A*, Liquid2, and Quiet Capital. The company has raised $102 million so far.

Health software company Athenahealth released the “next generation” of its electronic health record (

) platform, athenaClinicals, on Sept. 9. As part of the company’s push to release new AI tools across its portfolio, the EHR will implement personalized diagnosis information for patients and ambient note-taking for clinicians.

Venture firm Bain Capital announced Sept. 4 the completion of a merger between software companies HealthEdge and UST HealthProof. Bain purchased Boston’s HealthEdge for $2.6 billion plus debts in 2024 and has now purchased UST (financial information 

), joining the two companies under the HealthEdge name. Now, the company will establish a “single, end-to-end platform” for claims, payments, and care utilization management using AI, according to a Sept. 4 HealthEdge release.

Agentic AI company Hyro announced a new partnership with California health system Sutter Health Sept. 17. Under the partnership, Hyro’s technology will answer calls and messages to discuss scheduling, prescription management, and billing.

Agentic AI company Infinitus announced a partnership with IBM Consulting on Sept. 18. Together, the two tech companies hope to bring more agentic AI to speciality pharmacies for benefits verification and prior authorization.

On Sept. 4, AI voice company SoundHound AI announced a partnership with Ohio-based Primary Health Solutions. Under the agreement, SoundHound’s tool “Denise” will answer calls about appointment scheduling.

Agentic AI company Unity AI announced Sept. 23 an advanced version of its platform for ambulatory settings like radiology, dental, physical therapy, and behavioral health. Similar to other voice agents, its platform supports patient scheduling and patient intake.

AI mental health company Wysa acquired physical therapy company Kins on Sept. 30. While the financial information was not disclosed, Wysa will integrate its AI platform into Kins to help make care more accessible.

Healthcare Brew monthly series on AI Startups

AI 411: September 2025 

Your monthly roundup of AI announcements is here.

It seems like there’s an app for everything these days. You don’t have to leave your house for food, work, or social interaction (though we hope you still do on that last one).Now, we’re getting closer to testing drugs digitally, too.

 in npj Digital Medicine in May that used data from 59,000 Mayo Clinic patients’ 

 (EHRs), combined with computer modeling, to predict whether 17 existing drugs could help treat symptoms of heart failure.

The researchers checked their predictions against existing clinical trial results. They found their digital clinical trials predicted whether or not the drugs could improve several heart failure prognostic markers with about 89% accuracy, according to Nansu Zong, a biomedical informatician at Mayo Clinic and lead author of the study.

Because it uses existing data on real-world outcomes of drugs, an approach like this can’t predict outcomes of drug candidates that haven’t gone to market. But it could one day be a screening tool that helps researchers decide whether to repurpose an existing drug to potentially treat a new disease.

“I don’t think our intention will be to replace the actual clinical trial but to provide some signals so that it will be more effective and more efficient,” Cui Tao, the Nancy Peretsman and Robert Scully chair of the AI and informatics department and VP of Mayo Clinic Platform Informatics, told Healthcare Brew.

The researchers plan to test this model on other health conditions, too, and the “ultimate goal” is to predict how well

drugs will work on a disease, not just whether or not they’ll work, Nansu Zong, a biomedical informatician at Mayo Clinic and lead author of the study, told us. This can help them decide whether a clinical trial will be worth it.

“If you are low risk, low probability to achieve your expectation, maybe you’re going to change your drugs or you’re going to change your trials,” Zong said.

 of studies using observational data to simulate randomized trials, a process called emulation. Randomized clinical trials are considered the most rigorous way to determine a cause-and-effect relationship, but they’re expensive to conduct.

Doing a virtual trial run like this can help save money. The R&D for bringing a new drug to market can cost $1.5 billion to $2.5 billion over an average of 10+ years. R&D for repurposing an existing drug is around $300 million and takes an average of three years, according to a 2024 

 by Duke University research institute the Duke-Margolis Institute for Health Policy.

Emulation also allows researchers to track real-world drug outcomes after they’re approved and get data from patients typically 

 in clinical trials, like older people, minorities, or people with comorbidities, Tao said.

 According to the study, this research went beyond other clinical trial emulations by incorporating drug-target prediction, an analysis that predicts the probabilities a drug will affect specific genes.

In this case, the researchers predicted the probability each drug would work on genes associated with heart failure.

The researchers found this was a more accurate method than EHR-based emulation alone, which had about 70% predictive accuracy in this study according to Zong. The paper notes, however, “drawing definitive conclusions about the method’s effectiveness at this stage would be premature.”

 The research “offers a great direction for helping to design more targeted trials,” Jimeng Sun, professor at University of Illinois Urbana-Champaign’s school of computing and data science and co-founder of Keiji AI—a generative AI platform for clinical research—told us.

“I think similar approaches can be expanded to other conditions like 

 if sufficient EHR data can be obtained,” Sun, who was not involved in this study, said.

One limitation, he added, is that outcome measures have to be data EHRs regularly include.

“Diseases with slow progression and without very clear outcome endpoints, like Alzheimer’s, will be difficult to adopt with these approaches,” he said.

The study’s researchers said they want to explore whether advanced simulation methods like AI can allow them to evaluate new, untested drugs as well.

“We are currently investigating the use of advanced simulation technologies to generate synthetic data, which may extend the framework to evaluating new drug candidates,” Zong said.

How Mayo Clinic researchers are using algorithms to predict drug efficacy

The tool may aid the selection of drug candidates in the not-too-distant future.

Maia Anderson

If you’re a clinician, your peers may be judging you for using generative AI.

A study published in August in the medical journal npj Digital Medicine found clinicians negatively viewed their peers who use generative AI in clinical decision-making. The study, conducted by Johns Hopkins researchers, involved 276 practicing clinicians at an unnamed health system.

The clinicians viewed those who use AI to help them make patient care decisions as having a “lack of clinical skill and overall competence, resulting in a diminished perceived quality of patient care,” according to a press release from Johns Hopkins.

Tinglong Dai, Bernard T. Ferrari Professor of Business at Johns Hopkins and co-corresponding author of the study, said in a statement the stigma around AI “may be an obstacle to better care.”

However, while study participants negatively viewed fellow clinicians using the technology for clinical decision-making, the majority said they still believe AI is a beneficial tool, particularly when it’s tailored to a specific health system’s needs.

As of the end of 2024, roughly 85% of healthcare leaders surveyed by consulting firm McKinsey said they were either already using generative AI or were exploring the technology’s use.

Additionally, clinicians who framed their use of the technology as a verification tool rather than one for primary decision-making were viewed more positively, the study found. But those who didn’t use generative AI at all were viewed most positively.

Quinn Waeiss, a postdoctoral fellow at Stanford’s Center for Biomedical Ethics and a research associate with the university’s McCoy Family Center for Ethics and Society, told Healthcare Brew the study’s findings align with their research concerning the “uncritical use” of AI.

“I’ve heard both from patients and providers, this concern that providers go to medical school—we have standards for what their host of years and years of training looks like in their capacity of providing care to patients. We expect that they’re drawing on that expertise,” they said.

However, Waeiss added it shouldn’t be up to clinicians themselves to decide how to 

“Without incorporating the healthcare system perspective, we place a lot of that responsibility on individual providers, and I have concerns that we’re going to end up placing more and more expectations on [providers] as we shift the responsibilities,” they said.

A photo collage of a row of healthcare professionals with their arms crossed next to a doctor on a computer whose screen shows a healthcare cross made of binary code.

Doctors using AI for clinical decision-making viewed negatively by peers, study finds 

But clinicians still view AI positively overall. 

AI is hailed by many in healthcare as the future. But some worry as the tech advances, rural areas may get left behind.

A 2025 survey of 972 rural healthcare staff and leaders by independent research firm Black Book Research found just 8% of rural community and critical access hospitals use AI-driven analytics for predictive healthcare. Compare this to an 

 of 2023 American Hospital Association survey data, which found 65% of 2,547 hospitals across the US used AI or other predictive models.

Carrie Cochran-McClain, chief policy officer for the National Rural Health Association, told Healthcare Brew many rural facilities don’t have the funds or infrastructure to adopt AI and other new tech. It’s already a “battle” to help more rural facilities implement telehealth amid interoperability issues and areas without broadband internet, she added.

“About half of our rural hospitals right now in the country have negative operating margins, which means there’s not a cushion to invest in new software,” she said. “We’ve talked to clinicians who are starting to integrate pieces of [AI] and seeing a lot of benefit, but it’s very much still new to that setting.”

 Though AI comes with potential drawbacks—like 

 concerns—studies also suggest predictive analytics can identify high-risk patients for earlier intervention, potentially 

Groups ranging from industry organization the 

 have pointed to AI as a way for rural hospitals to 

About 66% of primary care health professional shortage areas and 62% of mental health shortage areas were in rural regions as of March 31, 

 the Health Resources and Services Administration.

“Clinicians could use AI to help inform their decision-making where they may not have a deep bench of providers in their clinical practice,” Cochran-McClain said.

Rural systems’ low adoption of AI is something Cochran-McClain and other experts hope could soon change.

Though July’s reconciliation bill had provisions that were decried as potential 

 for rural health, it also allocated $50 billion of funds to a newly created Rural Health Transformation Program.

One of the approved uses for these funds is “providing training and technical assistance for the development and adoption of technology-enabled solutions that improve care delivery in rural hospitals,” including AI tools, according to Centers for Medicare and Medicaid Services (CMS).

Cost estimates vary, but California-based software development company Folio3 estimates adding AI can cost anywhere from $40,000 for a simple AI tool to $100,000+ for a fully customized system.

Without a pathway for long-term Medicare and Medicaid reimbursement, however, pouring money into these technologies doesn’t make sense, Cochran-McClain said.

“This funding could be a great way to infuse the startup or implementation of some of these technologies. But there needs to be a sustainable path to support them outside of the transformation fund,” she said.

CMS reimburses some AI devices through mechanisms like the new technology add-on payment 

, which allows payment for recently FDA-approved technology to be bundled with payments for other healthcare services.

But devices are only eligible for that pathway when they’re “new,” usually meaning the first few years after FDA approval. When that period ends, the path to reimbursement can become uncertain, and reimbursement amounts may decrease or go away.

There’s no predictable or permanent AI reimbursement pathway, Cybil Roehrenbeck, a partner at law firm Hogan Lovells who leads the firm’s healthcare lobbying practice and serves as executive director of the AI Healthcare Coalition, told us.

Roehrenbeck wants a pathway for software-as-a-service and AI medical services within the Medicare Physician Fee Schedule. She says they should be categorized as medical services under Medicare Part B and billed, like other services in that category.

CMS noted in its proposed 2026 Physician Fee Schedule that many in the industry have sought a similar pathway, and asked for public comment through Sept. 12.

“It’s easier for providers to onboard the technology with that type of certainty,” Roehrenbeck said.

United States technology modern innovation circuit board computer map shape.

AI