Artificial intelligence

The industry is looking down the barrel of what the insurers say is a billion-dollar question: Are AI-backed billing tools driving up healthcare costs? 

Autonomous medical coding, AI-assisted documentation and ambient listening tools for notetaking all promise to reduce administrative burden in healthcare. But that has come at a price for Blue Cross Blue Shield Association of about $942 million over two years, according to research the insurer released.

BCBSA said the rise in spending is coming without a corresponding increase in treatment, which the insurer said it would expect if patients were truly sicker. Hospitals are increasingly billing inpatient stays as more complex than before, claiming more patients have a greater number of medical diagnoses that require higher reimbursement. 

That means AI-backed billing technologies could be contributing to higher healthcare spending, the insurer said.

“If patients are truly sicker, we'd expect to see more treatment,” said Luke Chalker, BCBSA’s senior vice president of product and data science, in a statement. 

“For example, we're seeing significantly more anemia diagnoses at these hospitals without a corresponding increase in transfusions,” he said. “The disconnect between diagnoses and treatment suggests that AI is identifying more billable conditions, not sicker patients.”

However, the companies that create those billing technologies are pushing back on that accusation, insisting their AI is getting providers what they are owed under the existing rules dictating reimbursement for care between payers and providers.

“Making a claim like this is making it in the context of the current paradigm,” said Dr. Travis Bias, deputy chief medical officer of health information systems at Solventum, whose coding platform is used by more than 80% of U.S. hospitals. “We live in a fee-for-service system, we pay for volume, not for value. So yes, if you collect more codes and do more procedures, the payments will increase.”

The BCBSA isn't the only payer facing higher costs. Analysts at PwC attribute a 9% increase in insurers' medical costs next year to AI-backed billing tools.

Payers are worried that these tools are enabling upcoding, when payers or providers submit inflated diagnostic codes to garner more reimbursement.

Correction, not upcoding

Hamid Tabatabaie, president and CEO of Codametrix, a healthcare technology company that uses AI to automate coding for over 500 hospitals and health systems, doesn't dispute BCBSA's claims about rising healthcare costs. But he does take issue with the blame being assigned to AI.

Autonomous coding tools are increasing costs for payers, he told Healthcare Dive. But it's not for nefarious reasons, although he couldn't rule out very rare instances of healthcare fraud, waste or abuse. Rather, the tools are overwhelmingly capturing what providers couldn’t without technical assistance.

“When last year they submitted their claims, if they weren't using a valid system, they were missing it,” Tabatabaie said. “And now this year, they have addressed it.”

And in the current fee-for-service environment, capturing a more complete picture of a clinical encounter will lead to higher costs, explained Bias. 

Under a fee-for-service system, payers reimburse providers based on the volume of services they render and the complexity of care. This incentivizes providers to deliver more care to maximize revenue and payers to limit care to save costs. 

But AI is now making it easier for providers to better capture what they are entitled to under existing rules, raising alarms for payers, according to Tabatabaie.

“Any optimization that a provider does that causes them to receive payment for otherwise what they weren't receiving — they don't like that,” Tabatabaie said. “Payers don't want to pay any more than they absolutely have to.”

Tabatabaie argues that medical codes were never meant to be about billing in the first place. ”Codes are supposed to be codes of clinical concepts,” he said. “They got hijacked by payers because that was the most convenient way to do claims processing.”

Both executives say their tools are designed to capture a patient’s medical reality more completely, which has benefits beyond reimbursement. AI can help identify community health needs and allocate resources more appropriately, according to Bias.

To prevent inaccurate coding, both companies also emphasized the strict guardrails they've built into their systems, touting coding compliance guidelines to ensure accuracy. 

A fix requires systemic change

The concerns expressed by payers may not last forever, though, Bias explained. New incentives and shifting behaviors for providers could address some of the cost challenges associated with using AI-backed billing platforms..

The healthcare industry, for instance, has been earnestly transitioning to value-based care since the Affordable Care Act was enacted in 2010. Value-based models reimburse providers based on patient outcomes rather than volume, incentivizing lower-cost, higher-quality care.

These payment models have yet to overtake fee-for-service, though. But it's something payers should care about more, Tabatabaie said.

With greater adoption of these models, payers wouldn’t be reimbursing providers for every code they submit. A more complete picture of a patient’s encounter also supports value-based payment since the models emphasize medical necessity and appropriate utilization more than fee-for-service. This also shifts the incentives for providers, he said.

Bias also predicts providers and payers will change the way they document, code and pay for care as revenue cycle AI tools continue to develop. Much of this will boil down to AI's potential to reduce administrative waste between payers and providers.

“It's going to take a realignment of systemic incentives,” Bias acknowledged. “That's a much bigger conversation... at a societal level.”

That conversation might have to take place sooner rather than later. Questions still remain about the disconnect identified in BCBSA's analysis around rising diagnosis codes and flat treatment patterns. 

Still, revenue cycle AI is inevitably here to stay despite these worries, as most providers use some form of AI for clinical documentation and coding.

This begs the question: Who pays the price? When payers shell out more for care, premiums increase and those costs cascade to employers and patients. Administrative burden may also escalate with more audits, denials and appeals. Bias noted that already, “large health systems are wasting billions of dollars simply rebutting, appealing denials.”

While vendors and payers debate who's right about coding accuracy, the meter keeps running.

CMS Administrator Dr. Mehmet Oz delivered a candid assessment of artificial intelligence's impact on healthcare costs in September: It's going to get worse before it gets better.

"Short term, AI is going to be inflationary because it's going to turbocharge the ability of the current billing systems to work more effectively," Oz said at Oracle's annual health and life sciences summit in Orlando. 

It’s the latest signal from the industry that AI could be inflating healthcare costs — even as its proponents argue the technology will lower them over time. 

Healthcare has been rapidly adopting AI across clinical and administrative realms. The tech has shown promise for reducing clinicians’ time spent on paperwork, helping doctors with clinical documentation and streamlining healthcare payments. 

But researchers have observed rising healthcare costs tied to AI’s use in the coding and billing space that they say could become a systemic issue if not addressed.

For providers, AI-powered coding is enabling them to bill insurers for more conditions when they treat patients. They say the technology helps them more accurately bill for care, while insurers say doctors are using AI to charge more money without any actual increase in medical care.

A study released by the insurer group Blue Cross Blue Shield Association found that the increase of billing has added nearly $1 billion in healthcare costs between 2023 and 2025. The group said the study “raises concerns” that AI tools are contributing to higher spending.

The Trump administration has been bullish on deploying AI in healthcare, even as it’s rolled back regulations on the technology. The CMS stood up a new office dedicated to health tech and AI implementation in June and has weaved an AI focus into several initiatives, such as the health tech ecosystem and its anti-fraud efforts. 

Although Oz noted the increase in healthcare costs, he said AI was still essential in the industry.

“I guarantee you we will lose lives if we don't use AI in the day-to-day trench warfare of fighting disease in America,” he said.

Oz said he hoped AI would lower healthcare costs in the long run, and that the government was relying on one particular area to make that happen: accountable care organizations.

ACOs are groups of providers that assume responsibility to provide care for a specific group of patients, such as Medicare enrollees. The hope is that a shift away from a fee-for-service model, where providers bill for each service, toward models that track health outcomes will incentivize providers to use AI as a tool for clinical efficiency rather than an engine for generating bills.

The CMS has launched several initiatives aimed at boosting ACO participation, claiming it saves billions for Medicare and drives down spending. 

Critics of the ACO model say it poses financial risks for providers if they fail to lower spending while maintaining quality care and adds a new layer of bureaucracy to their jobs.

Sponsored content

By Starburst Data

Agentic AI is moving quickly from concept to deployment in healthcare. In a September 2026 survey of 250 U.S. healthcare leaders responsible for AI strategy or identity security, 28% said their organizations already had agentic AI in production and another 44% were piloting or running proofs of concept.

Unlike AI tools designed mainly to summarize information, generate content or assist with a narrow task, agents can carry out multi-step work across systems. For payers, that could mean supporting prior authorization, claims operations, payment integrity, member service or care management. In pharma, agents could assist with clinical development, regulatory intelligence, pharmacovigilance, manufacturing and medical affairs. Health technology companies are exploring how agents can make their products and platforms more intelligent.

But the move from generating answers to taking action changes the data challenge.

Generative AI forced healthcare leaders to ask whether their data was ready to be queried. Agentic AI forces a harder question: Is the data ready to be acted upon?

The new challenge is context

Consider prior authorization. An AI agent supporting the process may need to review benefit information, coverage criteria, claims history, clinical documentation, provider information and the status of an existing authorization. It may also need to request missing information or trigger the next step in a workflow.

Each individual dataset can be accurate and the agent can still fail if the overall context is incomplete, outdated or inconsistent.

A pharma agent assisting a clinical development team faces a similar challenge. It may need to reason across trial data, site performance, operational metrics, regulatory information, safety data and internal research. McKinsey estimates that 75% to 85% of pharmaceutical workflows contain tasks that could be enhanced or automated by agents.

Those workflows rarely depend on a single system or neatly curated repository.

For an agent, data quality is only part of the picture. It also needs to know what the data means, where it came from, whether it is current, how it has been transformed and whether it is allowed to use it for the task at hand.

Healthcare's data environment was not designed for autonomous actors

Healthcare data is distributed by nature. Payers operate claims platforms, member systems, provider directories, utilization management applications, warehouses, lakehouses and third-party platforms. Pharma companies work across research repositories, clinical systems, regulatory platforms, manufacturing environments, commercial systems and external datasets. Health tech companies frequently need to connect data across their own platforms and their customers' environments.

Some of that information is structured. Some lives in notes, documents, messages, PDFs and other unstructured formats. Different systems may use different definitions, access policies, update schedules and identifiers.

People have learned to work around many of those gaps. They know which system is usually most current, which field needs interpretation or when to check another source before making a decision.

An AI agent does not automatically inherit that institutional knowledge.

And because agents can operate at machine speed, a data problem can become an operational problem much faster. An agent working from stale provider data might initiate the wrong next step. An authorization agent missing recent documentation might request information that has already been submitted. An agent supporting clinical development might draw on information without understanding whether it is current, approved for that use or traceable to its source.

Once AI begins acting across workflows, provenance, freshness, permissions and auditability are no longer back-office data concerns. They become part of the workflow itself.

Centralizing everything cannot be the prerequisite

One response is to assume that organizations must consolidate all relevant data before deploying agents. For most healthcare organizations, that is neither realistic nor necessary.

The industry has spent years building data warehouses, cloud platforms, data lakes, lakehouses, SaaS applications and specialized operational systems. Mergers, partnerships, regulatory requirements and new technologies will continue adding to that landscape.

At the same time, interoperability requirements are expanding the need to exchange data across organizational boundaries. Under the CMS Interoperability and Prior Authorization Final Rule, impacted payers generally must implement new Provider Access, Payer-to-Payer and Prior Authorization APIs beginning Jan. 1, 2027. Those requirements expand access to claims, encounter, prior authorization and other healthcare data.

The practical goal should be to give AI systems governed access to the data they need across the environments where it already lives.

That means separating data location from data accessibility. An agent does not need every dataset to be copied into one platform before it can use it. It does need a consistent way to find, access, interpret and combine information without bypassing the policies that govern it.

Every agent needs a trusted path to data

Before scaling an agentic AI use case, healthcare leaders should look closely at the data behind each action the agent may take.

Start with the action itself. What can the agent recommend, initiate, approve, route, change or escalate? A low-risk administrative task and a decision that affects coverage, payment, safety or compliance should not have the same data requirements.

Then map the information required to support that action. Which internal and external sources are involved? Which source is authoritative? How often does the information change?

Next, look at access. Can the agent reach the data it needs without unnecessary duplication? Are identity, role, purpose, consent and other policies applied consistently across systems?

Finally, make the process traceable. Leaders should be able to determine which data informed an action, where it came from, how it changed along the way and whether it was current at the time.

These capabilities can also be reused and create value beyond a single AI project. Reliable claims, eligibility, provider, clinical, research or product data can support multiple agents as well as analytics and other AI use cases.

Data readiness becomes AI readiness

Healthcare organizations will keep experimenting with AI agents because the potential applications are too broad to ignore. The harder part will be moving from a controlled pilot to an agent that can operate reliably across real workflows, systems and data.

At that point, the challenge becomes less about what the model can do and more about what the agent can access. Does it have the information it needs? Is that information current? Can it distinguish authoritative data from everything else? Are the right access controls in place?

For payers, pharma companies and health tech organizations, answering those questions does not require moving every dataset into one place. It requires a way to work across distributed data without losing visibility, governance or control.

That becomes more important as agents take on work that has real operational consequences. An agent acting on outdated eligibility information, an incomplete claims history or the wrong permissions can create errors that ripple through a process and have to be corrected later.

Before expanding an agent's role, healthcare organizations need to know exactly what data it is relying on and whether that data is fit for the action it is about to take. If they cannot establish that, the agent may be ready to act, but the organization is not ready to let it.

Health systems are hustling to implement artificial intelligence across their organizations, but how to measure returns on those investments isn’t always clear, experts said at the HLTH 2025 conference.

AI technology, which includes ambient documentation scribes, summarization tools and revenue cycle management products, costs money and time to put in place. And it’s not always obvious how AI produces hard-dollar ROI, which is important to justify expenses and measure success, experts say.

“The use of AI to simplify administration, the use of AI to assist with productivity is absolutely out there,” said Thom Bales, principal and health services advisory leader at PwC. “There still are questions around the ROI on it and hitting the bottom line.”

At the same time, health systems are also facing increased financial pressures. This summer, President Donald Trump signed a massive tax and policy bill that makes historic cuts to Medicaid, likely culling millions of people from the safety-net insurance program.

Meanwhile, more generous financial assistance for people who buy coverage on the Affordable Care Act exchanges — the issue at the center of a weeks-long government shutdown — is set to expire at the end of the year, also pushing more people off coverage.

More uninsured Americans means less revenue for providers and an increase in uncompensated care.

Still, the healthcare sector is hopeful that AI will move the needle on some of its biggest challenges, including workforce shortages, provider burnout and lengthy administrative tasks. Amid the challenging economic environment, healthcare companies are prioritizing AI investments most likely to improve profit margins and demonstrate clear returns, according to a survey published by Klas Research and Bain & Company.

But given those returns can be hard to show, hospitals may have to get creative in measuring the value of AI, including by looking at time savings for providers, stronger employee retention or improved patient satisfaction, experts say. Those softer measures could then be linked to financial improvement over time.

“I do think ultimately we can trace the majority of these things back to a bottom line, but you may have to connect more dots,” said Jennifer Goldsack, CEO of  industry association the Digital Medicine Society.

Piecing together ROI

Measuring financial returns is always tricky, given the complex interplay of factors that complicate attributing eventual savings to a specific product. This is also true when it comes to AI, where the direct financial impact of a product depends on the tool, the costs needed to adopt it and each provider’s specific priorities, experts said at HLTH.

“In healthcare, particularly in the healthcare delivery side, it’s really hard to apply line of sight to what we might think of as ROI,” Micky Tripathi, chief AI implementation officer at the Mayo Clinic, said during a Monday panel. “I don’t think that any of us should really expect you’re going to see a cost reduction that a CFO could look at and say, ‘Oh, I see it there.’”

Measuring financial returns is easier with certain AI products. For example, revenue cycle management tools might allow health systems to see metrics like time to collections improve, according to Sandra Johnson, senior vice president of client services at EHR vendor CliniComp.

But the impact of other AI tools is less obvious. There’s limited evidence that AI documentation assistants, which record physicians’ conversations with patients and create a clinical note, are affecting productivity and financial performance, according to a report published by the Peterson Health Technology Institute. 

However, the scribes are likely reducing clinician burnout, the report found — a significant concern that could be exacerbated by the amount of time and effort needed to document care and do other work in EHRs.

If AI scribes prevent a clinician from leaving the health system, that could be a major cost saver, experts say. The expense of replacing a physician can be two to three times the physician’s annual salary, studies suggest. 

And there are other potential financial and efficiency gains from ambient scribes, said Dr. Nele Jessel, the chief medical officer at health technology firm and EHR vendor Athenahealth.

For example, providers could complete notes more quickly and possibly see more patients. They also might be able to document services more accurately and bill for a higher level of services.

Patient satisfaction is another important metric for health systems. Patients tend to like ambient scribes in part because their doctors can be more present during appointments, instead of staring at a computer screen while they take notes. Documentation assistants might also help patients understand their care plans better, since physicians have to speak out loud during an exam to ensure the scribe picks up the information, Jessel said. 

“Providers are getting much better, for example, [at] vocalizing the physical exam, talking about the plan, right?” she said. “So it increases the transparency and also makes it a much more enjoyable experience for patients.”

How health systems are measuring AI’s impact

To implement AI successfully, health systems need to think critically about the biggest problems they want to solve and their main priorities, experts say. That plan should include how they’re going to evaluate the AI tool, which can help them determine if the project was a win — whether financially or by other metrics.

“To me, step zero is being very, very clear about, what is it that we’re doing? Why are we doing it? And what does success look like for that?” Mouneer Odeh, chief data and AI officer at Cedars-Sinai, said during a Sunday panel. 

Cleveland Clinic evaluates AI using two types of metrics, according to Sonja O’Malley, the general manager of business development and licensing at Cleveland Clinic Innovations, the system’s commercialization and innovation arm. 

Those include quantitative measures — like reducing no-show rates or cutting documentation time, compared to a baseline metric — as well as more qualitative criteria, like feedback on patient or clinician experience. 

Some tools will never be revenue generators. But they might do other important things, like allowing a clinician to spend time on tasks that require their full training and expertise, O’Malley said.

When an AI tool is brought before Ardent Health, metrics of success have to be clearly defined and measurable alongside a baseline, according to Anika Gardenhire, the health system’s chief digital and transformation officer. 

Financial return on investment is almost always important, but it might be correlative rather than direct causation, she said.

If a tool only breaks even, the system needs to consider other associated costs, such as implementation expenses or the time clinicians spent on the project when they could have been working elsewhere.

“If you get to a place where you’re just like, ‘Hey, this thing didn’t even break even. It doesn’t even pay for itself.’ Unless it’s doing something absolutely miraculous for patients, it’s out,” Gardenhire said.

Artificial intelligence tools are improving rapidly, leaving regulators and lawmakers the challenging task of determining how — and when — to oversee the technology in healthcare, experts said at the 2026 HIMSS conference. 

Regulations for AI use in the sector are fragmented. The Trump administration has largely moved to limit rules that could slow the adoption of the technology, leaving healthcare organizations with less guidance on how to implement AI.

Meanwhile, some states are attempting to legislate how AI is used in healthcare, creating a convoluted regulatory environment for providers and payers who operate across state lines. 

“It remains a very complex environment, with few guardrails for the use of AI in healthcare, and still a lot of work to do to really create an environment that is going to produce safe, reliable artificial intelligence for clinical use,” Tina Joros, the vice president of policy and innovation at health IT company Veradigm, said in an interview. 

Still, the HHS has done a lot of work to understand what the sector needs from the federal government, Arman Sharma, the deputy chief AI officer and senior advisor in the office of the secretary at the HHS, said during a panel discussion. 

The department has moved to align AI work across its agencies and make sure they’re working in the same direction, Sharma said. Previously, agencies might start their own separate AI projects without clear visibility into what their colleagues were doing.

To help the massive department figure out how it could advance AI in clinical care across its divisions, the HHS released a request for information in December. 

The RFI, which received nearly 450 comments, asked the industry how regulations might need to change to include AI tools, how the department could simplify payment to encourage the use of the technology and what research and development investments could offer best practices for adoption.

“We recognize that this is a space that is very volatile. It’s moving very quickly,” Sharma said. “It’s very important that the industry gets clear signals from the agency about what we care about.”

Adapting to a fast-moving technology

Inside the HHS, the Food and Drug Administration is trying to set policy that works in 2026 but will also adapt as the technology evolves, Jared Seehafer, senior advisor in the office of commissioner at the agency, said during a panel. 

The agency has approved more than 1,300 AI medical devices since 1995, according to a tracker maintained by MedTech Dive. The number of submissions has spiked in recent years as investment in the technology has increased. 

But the advent of agentic AI, or AI systems that could act more autonomously and potentially improve themselves, adds new challenges to the FDA’s regulatory framework, given developers typically have to notify the agency about their update plans, Seehafer noted. 

AI tools also might need a more balanced approach between reviewing them before they hit the market and keeping an eye on the devices in operation, he added. The performance of AI models can degrade over time, so the tools could need continuous oversight to ensure they remain accurate.

“That is going to require of us a new regulatory framework, but it’s also going to require from the entire community — and all stakeholders of different types — data sharing in a way that heretofore does not really happen,” Seehafer said. 

Another challenge is an agentic system that relies fully on human oversight likely won’t be able to scale, given the amount of human resources needed to monitor all of the AI’s output, said Dr. Haider Warraich, a program manager at the Advanced Research Projects Agency for Health, or ARPA-H.

For example, ARPA-H, a relatively new agency that supports biomedical and health research, is working on a project that aims to develop clinical AI agents that can autonomously guide cardiovascular disease care, Warraich said.

The program also plans to build a supervisory agent that can oversee the clinical tools to ensure they continue to perform safety and effectively. 

“I think we’re still in the early innings. So a system that has no human oversight would be unacceptable, I think, not just to patients and health systems, but also to clinicians like myself,” Warraich said. “And really the goal of the supervisory agent is, how do you create a system that can optimize human oversight, especially in the absence of ground truth?”

A complex regulatory environment

Meanwhile, as AI technology advances, the current regulatory scheme remains complicated. For example, few states have one regulatory body responsible for AI, so rules could fall under a number of agencies, like state departments of insurance or attorneys general. 

Plus, state legislation can be slow-moving, so guardrails could become stale and out-of-date by the time the bill advances, Veradigm’s Joros said.

For now, the adoption of AI agents or tools for diagnosis and treatment might take time, Danny Sama, vice president of digital platform and strategy for information services at Northwestern Medicine, said in an interview. 

“Frankly, I think a lot of it’s going to depend on regulatory and legal,” he said. “At some point, some agent is going to make a mistake, and there will be a malpractice lawsuit, and the question is, who’s liable?”

Healthcare is not known for its amazing customer experience. The nation’s health system is notoriously complex, requiring plenty of administrative work from patients to sift through their benefits, find available providers or collect their health data.

But artificial intelligence could begin to ameliorate some of the sector’s complexity, according to Nathan Frank, chief digital and technology officer at Aetna. The CVS Health-owned insurer has rolled out new digital tools that aim to help beneficiaries navigate their care, like a conversational AI assistant and a product called Care Paths that guides patients with specific conditions through their benefits and next steps.

“We believe that it’s going to change the consumer healthcare experience,” he said. “We think it’s going to continue to advance the clinical experience and really improve the ability for us to meet our members and provide better clinical outcomes.”

Frank sat down with Healthcare Dive to discuss how Aetna is using AI to engage members, how the insurer thinks about risks associated with the technology, and the importance of monitoring the products’ performance and soliciting feedback.

HEALTHCARE DIVE: How is Aetna thinking about using AI to communicate with your members? I know you recently rolled out a conversational AI assistant. 

NATHAN FRANK: So, we took a little bit of a different approach to that. We also have the ability to have a chat experience through our digital properties, but we actually embedded the Ask Anything capability. So no matter where you’re at in our app or web experience, you’re going to get the ability to ask anything in natural language. You’re going to get a full response that might provide you a map to your provider. It might provide a next best action. It might start to give you information that you might not even be thinking about. And then we’re using something called GenUI, which actually modifies that experience personally for you.

There’s a real world advantage to that. Less calls to our call centers. If members can get an answer very quickly on our web and app properties, they don’t want to have that phone call. We also know that healthcare is deeply personal. A lot of our members, if they have a chronic condition or are recently diagnosed with cancer, they want to call us. They want to get someone on the phone, they want to talk to a member services rep or a care manager.

Now we have the opportunity on the back end to provide [our member services representatives] with all of the same AI-powered tools we provide our members on the front end. So our member services reps know who the caller is. They know if they went to digital [properties], they know the question that they asked. And then we start to anticipate the answers for our member services folks so they can actually focus on being empathetic to the member during these situations, versus scrambling around looking for four different docs or the answer to a complex benefit question. 

How do you handle the potential risks of using AI, like providing inaccurate information to a member?

We want to make sure that AI is not a black box — fully explainable, fully transparent where we’re using AI. We’re spending a lot of time and effort putting the guardrails around no bias and hallucinations, and making sure we monitor and measure that constantly. 

Our experts in the business segments — that might be a clinician, that might be a nurse — are partnering with us, sitting in a small group of people developing these solutions. They also have the ability to provide instant feedback. If we’re summarizing clinical information using AI, we want to rely not only just on the technology to make sure that the guardrails are there, but we want to make sure those 15,000 nurses [that work for us] have the ability to say, “Hey, that didn’t work properly” or, more importantly, after an interaction: “That worked awesome.” And so we get that instant feedback, and that’s helped us make sure we’re providing the best experience.

So we have a robust AI governance model at scale that we’ve put into place. It makes sure that we're putting the right guardrails in there, we monitor it, and then we want to make sure, from an experience perspective, we have that connection back to the people that are using it every day.

You mentioned the importance of monitoring AI. What does the process look like? How do you think about the consistent monitoring and the workforce you need to maintain to make sure the AI is working well?

That’s sort of the beauty of technology. We have a really amazing set of product managers and engineers who are constantly monitoring and tuning the back end. That’s real time monitoring through technology tools, but also on a regular basis, retrospective monitoring of the data and how it’s being used. 

And then from a front-end experience perspective, you want to make sure that we’re providing the best experience so they can provide feedback there. So once you combine those two, you’re talking about using the actual technology and data and then the experience on the front end.

I'll just give you an example: clinical call summarization. We started out with a relatively small group to make sure we got it right. This would be a test and learn experience. We would run that for a period of time, make sure that we have the right transparency, monitoring, measurement, and we’re comfortable. Then we start to expand it, and then that’s where the tuning comes in to give us that optimized experience.

What has beneficiaries’ reaction to AI tools been like? I imagine some members might not want to talk to an AI, but they could value getting information more quickly.

Every member is different, right? Some members want to have a conversation via phone call. Some never want to call. Some want to be able to do it digitally. And so we look at success in a couple of different ways. With Care Paths, we’ve been able to reduce calls around maternity, diabetes and [musculoskeletal conditions] by [8.6%]. And the reason why is because they’re having a really good experience in the app, so they don’t have to call. 

For those three conditions, we’ve actually increased the digital engagement by 38%. So you’re pregnant, you want to understand what your benefits are. You opt into a Care Path for maternity on our app, we’re going to give you a personalized, condition-focused view of what’s next for you, what you might need to do, what questions you want to ask your provider. We also give you the instant ability — if you want to have that live connection — to click on a button and talk to a care manager on your team. 

We’ve seen our customer [satisfaction] continue to go up because the digital experience is improving. AI is helping with that. 

Health systems are rushing to adopt artificial intelligence tools, hoping the technology can help them cut costs and alleviate strain on overworked providers. 

But AI isn’t always easy to implement, particularly at organizations with few resources, and clinicians have plenty of concerns about inaccuracies and biased responses. Experts say that governance protocols are key to ensure models are used correctly and perform up to standard, but many providers say they’re unaware of their organization’s AI policies.

The Joint Commission is hoping its new voluntary certification program can provide a blueprint for overseeing AI. The healthcare accrediting and standards group rolled out the certification early this month in a bid to recognize organizations that can demonstrate effective data management, risk and bias reduction, monitoring and safety evaluation, and education and training. 

The certification “can be understood, adopted and enabled from anyone on the spectrum of maturity in regards to where they are with governance,” said William Walders, executive vice president and chief digital and information officer at the Joint Commission.

Walders and Ken Grubbs, the commission’s chief nursing executive and executive vice president of accreditation and certification operations, sat down with Healthcare Dive to discuss the certification, its impact and how the program works for low-resource health systems. 

HEALTHCARE DIVE: Tell me about the Responsible Use of AI in Healthcare certification. What are some of the things hospitals and health systems need to do to be certified?

KEN GRUBBS: Number one, making sure that there’s an appropriate governance structure in place. It really allows organizations to determine what that looks like and how they formalize it. It really is focused on patients and ensuring that it’s protecting patient safety, while also the workforce’s trust in using the technology.

There has to be effective data management, and just making sure that that data is protected, that it’s not accessed in an unauthorized manner. It really gets into that ongoing monitoring — making sure organizations have some type of registry of what products they’re using, making sure that they track changes to that, and then all of the monitoring that would go along with any program from a quality, performance improvement, and safety standpoint.

Again, it’s focused on patients. And it’s focused on making sure that there’s the ability to identify any risks, like bias or anything of that nature.

WILLIAM WALDERS: Embedded in some of [the certification] is education and training. Having role- and use case-specific training to ensure safe adoption. I think that’s overarching. It’s something health systems do well. They keep top of mind how their clinicians and others practice care, and this is a tool in their toolbox, no different than the other ones that they have to really enable and affirm safe and quality healthcare.

What do you think the impact of the certification program will be? Is this something that you’ve had a lot of interest in from health systems?

WALDERS: We’ve gotten significant digital engagement and outreach from folks seeing a need and curious how to fill it. Not only was the certification written to survive a lot of pace of change, it was written intentionally to survive the evolution of AI — flexible. Similarly around organization size and also around use case and type. 

A lot of that outreach has been targeted. I’ve been asked this question in the past: “We have AI in our washing machines in the laundry. Does that need to go to governance?” It’s up to you. I mean, if you ask me, no. My washing machine is not providing a clinical decision. But that’s probably the first question you should ask as part of your governance process. Is it applicable? 

That’s an organizational interpretation, but that’s probably your first check. Toothbrush? No. Washing machine? No. CT scanner? Yeah, probably. And so a lot of the outreach has been there.

But let’s be clear, there’s been a bit of a gap. And we feel the obligation and commitment to safety and quality to fill that gap, and so our focus is on where AI overlaps with safety and quality. And so a lot of folks have been grateful that we’re doing something and we’re here to help.

How do you think about ensuring low-resource hospitals — small, rural, dependent on Medicaid — can effectively adopt AI? Is that something you were thinking about when developing the certification?

GRUBBS: Part of it was making sure that we were not too prescriptive, that we got the principles of safe use of AI from a governance perspective and implementation standpoint within the certification, but we allow organizations to determine what that looks like. So that allows adaptability.

The other thing is, a part of the field review was to ask organizations to give feedback to make sure that we had an understanding of what the lift would be for this certification. Based off of that feedback, we made changes as well. So I think those two things together were extremely meaningful in making sure that we were not overburdensome, but we put the absolute concepts of safety within the certification and allowed organizations to determine what model works or what framework works best for them based off of those principles. 

Could you give examples of something that you thought you would have included initially, but then changed to keep the certification applicable to a broader group of providers?

WALDERS: When we first looked at this, we thought we could be a little more prescriptive on governance composition. Can we have a model? Do you look at something like an org chart? And then you realize that no, sometimes the chief operating officer is the same person who’s fixing the boiler, who sometimes is the phlebotomist, right? And so we didn’t want to be as prescriptive as deciding roles and responsibilities.

And other threads like that, just pressure testing. Put yourself in that clinic in Oklahoma. Does this work for them? And it needs to. They’re getting an AI ultrasound machine because they need ultrasound in the clinic. Could it fit that process? Is it overly prescriptive that you need seven different titles with seven unique expertises to vet the radiology aspect of a GE device? There were some deep dive discussions, and getting back to reality and accessibility was where we landed.

Adopting artificial intelligence tools could be a boon for safety-net health organizations, which frequently operate at low margins with fewer resources — and face more turmoil as significant Medicaid cuts loom, experts said at the 2026 HIMSS conference.

Implementing the technology is more challenging for the providers, which serve low-income populations and offer care regardless of patients’ ability to pay, given their workforce challenges and sometimes limited technical capabilities. 

Now, safety-net organizations will face new headwinds. Last summer, President Donald Trump signed sweeping tax and policy legislation into law that includes significant funding reductions to the insurance program Medicaid. 

The “Big Beautiful Bill” will likely cull millions of people from coverage and cut billions from state Medicaid programs.  

“[The law] presents an existential crisis to safety net providers,” Kara Carter, senior vice president of strategy and programs at the nonprofit California Health Care Foundation, said during a panel. “There’s no question about it. That is red states, blue states, purple states. It does not matter.”

However, the challenges brought on by the policies could be an opportunity for safety-net organizations to take advantage of AI to ameliorate some of the damage, experts said. 

In one of the law’s most significant impacts, many Medicaid beneficiaries will need to log 80 hours of work, school or volunteer hours per month to stay enrolled in the safety-net insurance program. 

That will likely create a lot of administrative work for enrollees and Medicaid programs. For example, before Arkansas’ work requirement policy was struck down during the first Trump administration, many beneficiaries didn’t know about the requirements and struggled to report their compliance. 

Additionally, the law passed last year requires states to check enrollees’ eligibility every six months instead of annually. 

AI could help with redeterminations and enrollment calls, Carter said. Some AI agents can handle outreach at the rate of 40 to 100 calls a second, she added. 

Many of the law’s provisions don’t go into effect immediately, giving states and providers some time to prepare, said David Ford, CEO of CMA Physician Services, the practice transformation subsidiary of the California Medical Association. 

“Before state legislatures start doing the really awful, terrible things we know they’re going to do — kicking people off the program, limiting benefits — the hope is that AI can generate some efficiencies and give us some savings and allow us to maintain some of that, even in a resource-strapped environment,” he said. 

Plus, the Rural Health Transformation Program — a $50 billion fund included in the law to improve care in rural communities — could give some states a lot of money to adopt AI and other technologies, said Adimika Meadows Arthur, executive director of the nonprofit Health Tech 4 Medicaid.

AI might be able to fill some gaps among providers too. If clinics need to cut back on staff, they’ll still need to make calls to patients — and AI could be one option to help, said Dr. Jeffrey Arroyo, associate director of clinical informatics at AltaMed Health Services, the nation’s largest federally qualified health center. 

“We’re already short staffed, but we’re probably going to lose more,” he said. “That doesn’t mean the work is going to go away. So it’s going to be put more on IT to take advantage of these AI solutions that are seen outside of healthcare and make magic happen.”

Generative artificial intelligence has exploded in the healthcare sector in recent years, driven by hopes the technology could take on a variety of tasks — from clinical documentation to data analysis — and lessen the industry’s long-standing workforce challenges.

At the same time, healthcare organizations often struggle to manage cybersecurity, burdened by frequent cyberattack attempts as the sector adopts more internet-connected tools.

AI products could be another target for cybercriminals. Meanwhile, hackers can use their own AI to launch cyberattacks. That should create new work and security disciplines for healthcare cyber teams, said Taylor Lehmann, a director of the office of the chief information security officer at Google Cloud.

Lehmann sat down with Healthcare Dive to discuss how the advent of generative AI tools will impact healthcare cybersecurity and what organizations need to do to prepare for an increasingly AI-augmented security workforce. 

HEALTHCARE DIVE: Would you say these generative AI tools are more ripe for attack than any other piece of tech hospitals use? Or is that just another vendor they have to consider when they're thinking about cybersecurity?

TAYLOR LEHMANN: This is where I am concerned about the future. Today we’re talking with AI systems, tomorrow we’re watching AI systems do stuff — like the agentic push we’re still on the cusp of — so this answer might change a bit. But number one, it’s going to be very hard to detect when AI is wrong and whether that being wrong is due to it being manipulated by a nefarious actor. 

Detecting wrongness is already a challenge. We’ve been working on this for years. No one’s been able to completely eliminate hallucinations or inaccuracy. But I do think detecting more than just hallucinations, but functionality that has been introduced by a bad guy sitting in the middle to get a certain outcome — it’s going to be really hard to detect and increasingly difficult to detect. 

So organizations need to make sure that they have provenance for basically everything from the model being served to you, the technology that has been done with, all the way down to the provider of the model and the data they use to train it. That visibility is going to be critical. 

At Google we use model cards. We use practices like cryptographic binary signing, where we can tell exactly where the code came from that is running that model. We can tell you exactly where the data came from that trained that model. We can trace every record that went into that model’s training from birth to death. And organizations are going to need to be able to see that in order to manage those risks.

You mentioned the importance of being able to tell what models are trained on and what data is flowing in, and how that could help you figure out if the tool has been infiltrated. How should health systems be thinking about implementing workforces now to deal with those concerns?

There’s definitely new security disciplines coming. This is actually some of the work that we’re doing right now in the office of the CISO. Google is looking at some of these new roles and new capabilities, and then putting guidance together for folks on what to do.

The first thing I would say is, there are ways and methods that are needed to secure AI systems out of the gate. We feel very strongly that any reasonable approach to securing AI and AI systems needs to involve having strong identity controls in place and making sure that we know who’s using the model. We know what the model’s identity is, we know what the user’s identity is and we can differentiate between those two things. We also have radical transparency in everything the model does. Right or wrong, we can see it.

One of the new big areas is this concept of AI “red teaming.” Organizations deploy a team of people that do nothing but try to make these things break — basically try to get them to produce harmful content, take actions that were not intended — to test the limits of the safeguards, as well as evaluate how well the models are trained, whether they’re over- or under-fit for purpose.

I think you’re starting to see AI governance becoming hugely important. It’s always been important, but understanding the risks associated with AI — especially in regulated industries or in safety-critical use cases — requires a combination of technical skill as well as understanding of how AI works and is built, as well as the regulatory or the business context to determine what is an important risk.

Having really robust AI governance and people with these right skills is kind of a difficult place today. Because right now, you’ve got people who are great engineers, you’ve got people who are great doctors. You don't necessarily have both, with a risk hat on. So we’re going to see the rise of industry-specific, even practice-specific, governance professionals. 

Hackers are increasingly using AI to improve their intrusion attempts. How should health systems be thinking about the increased risks posed by cyber criminals now that they have access to AI?

Number one is, we have to wrangle how organizations provision and manage identities, like the identities of humans, the identities of machines, the identities of agents. 

I do not think your average hospital really thinks about identity more than how to identify the people walking in the door and what their user ID and password are. Identity is going to be the piece of digital evidence or digital artifact that is going to tie everything together. Based on identity is where we will apply controls with appropriate context.

The second is speed. AI systems move fast. They move faster than humans. Weaponized AI systems, especially those that have been trained to do very specific things, will move at lightning speed. We’re not going to say in the future, “Hey, you have less than an hour to fix a ransomware-positive infection of a piece of code.” You have seconds, if milliseconds, now. And if you can't operate with that same speed to detect and eliminate, you’re going to be in trouble. 

The strategy I have is, look at your architecture, look at the systems, look at your applications. Look at how quickly you can deploy new controls. Look how quickly you can rebuild the system from scratch, look how quickly you can deploy a patch. Look how quickly you can detect an event and take action on it, and drive down the things and eliminate the things that stand in your way from doing that — whether it be from hours to minutes or minutes to seconds. But prioritize those things that help you deploy corrections and detect and respond to issues extremely quickly. 

Lured by promises of time savings and more face-to-face interactions with patients, a growing number of doctors are using artificial intelligence-backed scribes for clinical documentation. 

But experts say there’s a hidden cost: the risk of medical malpractice.

As AI scribes move from limited pilots to routine use in exam rooms, the technology can make documentation mistakes that are hard to detect, leading to potential medical errors and malpractice lawsuits as hospitals increase their use, experts say. And clinicians are ultimately on the hook for those mistakes, regardless of whether they used AI to generate notes.

While there are no public malpractice cases tied to AI scribes, “that doesn’t mean there are zero,” said Bill Satterwhite, a practicing physician and licensed attorney who serves as principal of healthcare performance improvement and physician enterprise at Huron Consulting Group. That’s because lawsuits often take a while to work their way through the legal system, especially those dealing with sensitive personal information like medical malpractice.

The risks are increasing, as more than 1 in 4 physicians use AI to document billing codes, medical charts or visit notes, as well as generate chart summaries, according to a survey by the American Medical Association. Another 7 in 10 respondents see opportunities for AI to automate clinical and administrative tasks contributing to burnout, the survey found.

“We've seen a lot of benefits in terms of work-life balance for providers,” said Julie Massey, senior partner of digital and technology transformation and co-leader of the clinical technology innovation practice at healthcare consulting firm Chartis. 

Still, research shows that AI scribes produce significantly worse quality clinical documentation than humans, which could raise the risk of medical malpractice if patients receive improper care. 

A patient, for example, could experience a negative drug interaction if an AI scribe omits information about the medication a patient is already taking. 

In a study published in the Annals of Internal Medicine, human graders scored primary care notes produced by 11 commercial AI scribe tools significantly lower than notes written by 18 human clinicians

The notes written with AI scored lower in all 10 quality domains the graders assessed, with the largest deficits between human and AI-generated notes in thoroughness, organization and usefulness. 

“Although ambient AI scribes hold promise for reducing clinician burden, independent, vendor-neutral evaluations of note quality are essential before large-scale clinical deployment,” the study says.

Potential errors raise malpractice risk

AI models are largely less-than-reliable notetakers because they operate via prediction, generating outputs by selecting the statistically most likely next word or character based on prior context. 

Because it's based on probability, the outputs are unstable. If a doctor were to use the same AI scribe for the same clinical visit, it would produce different notes each time.

That makes it far more difficult to audit than traditional software, which fails in predictable rather than unpredictable ways, Satterfield said.

AI scribes are especially prone to a number of mistakes that raise the risk of medical malpractice, experts say. 

These include missing clinical information discussed during a visit, transcribing medical terminology incorrectly, fabricating information about a patient to provide an answer in an EHR template or filling in an information gap — often called hallucinating.

They may also omit non-verbal observations, such as facial expressions or tone, misinterpret vague statements, drop the context behind a clinical decision or introduce racial bias.

Even worse, AI scribes won’t notify the clinician when such an error occurs. 

For example, if a patient says something unclear, an AI scribe will resolve the ambiguity silently rather than inform the provider. A human, on the other hand, would be more “comfortable raising a question,” said Jennifer Geetter, a partner at McDermott, Will & Schulte.

Accuracy can also degrade over time.

“It’s like a knife that gets dull,” Geetter said.

Although providers should review, approve and sign off on AI-assisted documentation, they may spend less time and effort reviewing notes as reliance on the scribes grows, according to the Texas Medical Liability Trust, a malpractice insurer. Providers could rubber-stamp notes without critical review, assume all important information was captured or avoid thinking critically about what was documented.

In fact, nearly 9 in 10 physicians worry that using AI will lead to skill loss, especially among those early in their careers, according to the AMA.

How providers can limit their liability

Fortunately, hospitals and health systems can take steps to limit the liability risks associated with AI scribes.

Experts recommend starting with a pilot to ensure there are no major issues before deploying the technology systemwide.

Other precaution measures include:

• Approving products centrally rather than leaving the choice to individual clinicians.
• Including AI scribe use in consent forms and procedures.
• Establishing strict review protocols by treating notes generated by an AI scribe as a draft the clinician must verify before signing.
• Setting documentation timetables, so notes get reviewed while the visit is still fresh in the clinician’s mind.
• Training staff on AI use and documentation, with refreshers whenever the system is updated.
• Using safeguards built into EHR systems, such as prompts to confirm medications, dosages and allergies.
• Auditing notes on an ongoing basis, including after the pilot ends, to catch accuracy drift.

Ultimately, however, the simplest precaution is reminding providers that they’re responsible for patient care, even if technology can make it easier.

“You need to think of AI scribing as being a copilot, not the pilot, not autopilot,” Satterwhite said.

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 last year 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 last year 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 this week. 

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 published last month 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.