By Avi Mukherjee, Chief Product & Technology Officer, Verato
AI pilots in healthcare are in full swing, with the goal of creating less confusion and complexity in the system and more focus and time for patient care. Healthcare has never been an early adopter of technology—the sensitive nature of our data begs for caution because lives are literally on the line. Healthcare AI is still in its early stages, and this is good news, because it means it’s not too late to build the foundation we need to realize its true value.
Today, with the introduction of agentic AI, the possibilities for improving staff productivity and patient outcomes are unprecedented. The question healthcare leaders should be asking is no longer, Should we invest in AI? But rather Which AI initiatives will actually deliver lasting value? And how can we ensure successful outcomes and not science experiments?
Where AI Is Delivering Today
Here are a few areas where I’ve seen AI currently serving providers, payers, and patients.
Providers: Reducing Clinician Burnout
Physician burnout rates dropped below 45% in 2025, reaching the lowest level since before COVID-19, according to the American Medical Association (AMA). This is a promising sign that efforts to improve physician well-being are gaining traction after reaching an all-time high during the peak of the pandemic. Still, fewer healthcare workers are serving a growing patient population, as demand for physicians and nurses is projected to outpace recent graduates.
AI-enabled ambient listening during patient visits allows doctors to spend less time documenting visits and more time with patients. By automating administrative tasks for physicians and nurses through ambient listening, it’s estimated that physicians can gain back 16 minutes of documentation time and spend 13 fewer minutes in the medical record for every eight hours of patient care. The results are promising, but like everything we do in healthcare technology, you need the “human in the loop” to ensure accuracy.
This upside goes beyond increasing time with patients, it directly contributes to reducing the burnout quotient that has been hampering clinician interest. Let’s hope we see an uptick in medical school enrollments once the tide turns and AI-enabled charting workflows become standard.
Payers: Working Smarter to Reduce Delayed Payments
Prior authorization, claims adjudication, revenue cycle management, and medical coding all involve repetitive tasks that consume significant time and resources. AI can accelerate these payer workflows by organizing information, identifying missing documentation, surfacing relevant clinical evidence, and reducing manual review for routine cases.
According to a 2025 AMA survey, 95% of patients report delays due to prior authorization hold ups, and 88% of physicians report that it interferes with continuity of care. AI is helping payers become more proactive instead of transactional.
The anguish of a sick patient waiting to find out if their insurance will cover a rare cancer treatment can be significantly reduced thanks to AI matching records. The JAMA Network Open found a 33.9% reduction in authorization time resulting from the integration of prior authorization workflow technology, from 4.2 business days to 2.8 business days.
The most successful payers are using agentic workflows for prior authorization and care coordination with identity infrastructure in place from the start. Physician coding knowledge will be largely replaced by AI, analogous to how computer vision disrupted radiology by enabling faster and more accurate scanning to determine radiation locations.
Another area where AI is showing great promise for payers is revenue cycle management (RCM). By automating the administrative processes that occur before, during, and after patient care—from eligibility verification and coding to claims submission, denial management, and payment collection—payers can process claims efficiently, manage risk, ensure accurate payments, and reduce administrative costs.
Patients: Improving Customer Service
Agentic AI refers to intelligent software that can reason through a series of tasks, make decisions within defined guardrails, and complete work across multiple systems with limited human intervention. One area where this is making incredible strides for improved productivity is call centers.
Voice agents cut average handle time by 30–50% through instant data lookups, parallel task execution, and consistent call flow. AI agents can trace a patient’s history in seconds, without putting them on hold. Overall, this means better navigation of healthcare services, and the latest bots even have personal names to make callers feel like they are actually talking to a human.
AI excels at information aggregation, analysis, and decision support—tasks where speed and accuracy matter more than human judgment—freeing customer service professionals to focus on the moments where human expertise is the only alternative.
Gaining Steam on the AI Horizon
Other use cases rapidly gaining steam that will benefit the entire ecosystem of providers, payers, and patients include:
- Preparing patient records before appointments
- Coordinating care across providers and fragmented systems (enabling decisions like optimal clinic placement, identifying provider leakage to competitor health systems, and post-acute care gap analysis)
- Identifying missing patient and provider information before it delays treatment
- Recommending next-best actions for clinicians
- Monitoring workflows in the background and resolving routine tasks automatically
- Transforming healthcare from managing information to managing patient care
What it Will Take to Claim AI Victory in Healthcare
The organizations that realize the true benefits from AI most won’t necessarily have the newest or the fastest AI models. They’ll be the ones that build trustworthy data, clear governance, and workflows where humans and intelligent agents work together.
Before Healthcare Organizations Can Claim Victory in the AI Race, Two Prerequisites Must be Realized.
1. Quality Data as the Foundation of All AI Integrations
Clear and accurate information is foundational to a successful experience because an LLM can only function on the data it’s fed—garbage in will yield garbage out.
Better clinical documentation or improved customer service, faster prior authorization, and the like, will not be possible without a complete picture of a patient as a person. This includes knowing who is and is not covered under a family policy, as well as patient preferences for receiving information and consent to data release.
Similarly, if the data is not correct, none of these agentic processes will work because mistakes will just beget more mistakes. If you don’t have the right records and attributes regarding a patient’s history, AI will hallucinate, potentially leading to disastrous consequences. Consider a disparate set of patient records where only one entry lists an allergy to a certain medication. If these records are not unified and accurate, AI could mistakenly list this drug in an intake sheet.
Lastly, if AI takes a piece of information out of context, for example, a household may consist of four people, but if the children move out and the parents get divorced, the context of how they are related changes. If these contextual data points are missed, one claim’s mistake can be carried over to multiple transactions, building on agentic processes that repeat and create a snowball effect of misinformation that could take months or even years to fix.
2. The “H” in Healthcare is for “Human”
Humans will always be needed to play the following roles in healthcare:
- Showing Empathy: Patients have feelings and need emotional support. AI lacks empathy, perspective, and perception.
- Difficult Conversations: AI should never have to deliver the news that someone has tested positive for a life-threatening disease.
- Shared Decision-Making: It takes a team—doctors, family, pharmacists—to make decisions on a course of treatment. AI can lay out the options, even recommendations, but it can’t be put in a position to decide on care or treatment.
- Building Patient Trust: Patient relationships evolve over time and clinical teams build trust by understanding each patient’s unique situation. Bots can make things up.
- Clinical Judgment in Complex Cases: Bots don’t go to med school, doctors do.
Since human-to-human interaction in healthcare cannot be fully replaced, AI should be used to
augment, not eliminate, these essential human touchpoints.
The Bottom Line:
Health systems and payers tend to be late adopters when it comes to the implementation of new technologies, waiting until roughly 50% of peers have adopted before committing, even when they publicly state innovation goals.
In the age of rapid AI deployment, this is a smart strategy because foundational investment must run in parallel with AI projects, not sequentially. This intelligent collaboration of data cleansing and AI workflow development must co-exist if the benefits of AI are to lead to systematic improvements and not seismic catastrophes.
Each AI project should contribute budget to a shared foundational infrastructure pool to ensure that mistakes won’t be made and carried by an algorithm into multiple data streams going forward. Thoughtfulness, vision, and disciplined experimentation are what will separate the winners from the losers in the next frontier of healthcare AI development.
The groundbreaking opportunities will only be realized if we establish a solid identity layer as the foundation, ensure human in the loop collaborative intelligence, and continually measure outcomes.
