Few healthcare processes feel more entrenched than prior authorization — the insurer approval protocol that’s become shorthand for delays and red tape that complicates care and goodwill between patients, providers and payers.
Now, Pittsburgh-based Highmark Health, which operates its multi-state insurance services firm Highmark and 14-hospital health system Allegheny Health Network, is collaborating with Abridge to compress what can be a monthslong process to the length of a patient visit, with ambient AI quietly doing the heavy lifting in the background to give patients clarity before they leave the exam room.
The agreement marks a few firsts. It changes the timing of prior authorizations, which have long occurred post-visit — even with tools meant to streamline the process. It marks a new frontier for ambient AI in clinical documentation, which is now capable of not only transforming free-form patient-clinician conversations into completed administrative tasks like prior auth, but can reorder those tasks to happen as part of the conversation.
It’s also a notable move in a landscape where the AI race between insurers and providers seemed poised to only make prior authorization worse. A partnership like this marks a plot twist Abridge founder and CEO Shiv Rao, MD, has had in mind for years. It’s a muscle the company can flex only because of the tens of millions patient-physician conversations it has recorded and documented.
“I think sometimes technology’s role is to trailblaze, to actually build solutions that solve for all the different pain points all of us in the industry and all of us as people are acutely aware of,” he said.
A ‘difficult’ and far-reaching pain point
Prior authorization began in the late 1960s and early 1970s as a cost-control mechanism for private insurers and early managed care organizations to avoid unnecessary hospitalizations. Initially, physicians needed insurers’ advance approval for a narrow set of high-cost procedures and drugs, often as part of utilization review. But as managed care expanded, so did prior authorization’s scope.
Over the last 15 years, the administrative process has become widespread and divisive. More than half of insured adults say their insurer required prior authorization for a needed hospitalization, service, treatment or medication in the past two years, a KFF poll found. An even greater share of the public — 73% — view insurers’ delays and denials as a “major problem.” While intended as a check on overuse, prior authorization has become a symbol to many of the system at its most obstructive.
“The pre-authorization approval process is… difficult, to say the least,” said Mark Sevco, president of Allegheny Health Network. “It’s time-consuming. It has high administrative costs, with multiple phone calls and conversations.”
Even in integrated delivery networks like Highmark Health, where the payer and provider arms share a parent, prior authorization remains cumbersome. The process is estimated to account for $35 billion of U.S. health care administrative spending, with an average cost per authorization of $40 to $50 per submission for private payers and $20 to $30 for providers.
Dr. Rao knows the prior auth slowdown firsthand. While rounding as a cardiologist, he recently ordered a cardiac MRI that set off “a number of paper cuts” over the following weeks — questions about eligibility, debates over whether a CT should come first and calls to other insurance-side clinicians to make his case. He doesn’t dispute the purpose prior authorization is intended to serve. But the means, he said, no longer keep pace with the end.
“When patients have to wait weeks or months for that cardiac MRI, it’s a disservice,” Dr. Rao said. “It’s certainly more clerical work for me and my team, but ultimately it’s doing the patient a disservice. And at the payer level, not only is the experience compromised, but the outcome could be compromised and certainly over time costs can go up, as well.”
From months to minutes at the point of conversation
Abridge’s new prior authorization solution, powered by its Contextual Reasoning Engine, leverages the clinical conversation, interprets its meaning and maps it to a patient’s health plan policies. Integrated into the documentation workflow, it can surface fine-print requirements for each diagnosis, order or prescription and prompt clinicians to fill any gaps required by insurers before the visit ends.
“I’ve never seen anything like this before,” said Richard Clarke, PhD, chief analytics officer at Highmark Health.
Speed is one differentiator here. The process must run in near real time, marking a shift from prior-auth solutions that accelerate approvals only after the patient leaves. It’s still not uncommon for patients to learn on the day of surgery whether or not their insurer covers a procedure. This review for AHN and Highmark patients now occurs during the visit, at the point of conversation, with the goal to “get it right the first time,” as Dr. Rao put it.
“The enablement of that seamless experience is not simple,” Mr. Clarke said. “There are actually a lot of complex questions that need to get answered.”
This is where science and collaboration — not just technology itself — come in.
Mr. Clarke said Highmark’s role in the partnership is to ensure accurate, specific policy information for each test, image, medication or procedure is available to Abridge’s AI technology in real time for the specific patient in front of the clinician. Delivering that information, he pointed out, has taken years of investment in coding, storing and structuring data so it can now integrate with a partner like Abridge.
From there, Abridge’s large language models must navigate Highmark-specific prior authorization policies at scale, determine approval eligibility, detect physicians’ recommended diagnostic or therapeutic from the clinical conversation already unfolding, understand how it needs to be authorized, and map it to a CPT code. It can then flag missing requirements — that one last question the physician did not yet ask — before the visit ends and then show its homework, sharing evidence with clinicians that shows how its AI reasoning aligns with the patient visit and health plan policy.
“We expose the AI’s reasoning and map each box it checks to the original policy,” Dr. Rao said. That can build trust at the point of conversation as well for the clinician that this technology really has their back and their patient’s back.”
Testing the model before taking it wider
As an integrated delivery network, AHN’s insurer-provider structure makes it an ideal testbed for Abridge’s capabilities, Mr. Sevco said — a way to trial innovations before potentially expanding them to Highmark’s other provider partners in New York, Delaware and West Virginia.
“Knowing that the product works, what we’re most excited about and the uniqueness of this is the Highmark strategic partnership to co-develop frontier products with Abridge and with a payer mindset,” Mr. Sevco noted. The relationship carries particular weight for AHN, which derives roughly half of its revenue from Highmark.
In Pennsylvania, AHN is going all-in.
“Most institutions just take it and use it in physician practice,” Mr. Sevco said. “We’re making a significant investment knowing physicians are overwhelmed and burned out.”
Abridge will support more than 50 million medical conversations in 2025 across 150-plus partners, with AHN now among them. The rollout will extend beyond physician practices to hospital medicine, emergency departments, nursing documentation, and home health.
Pilots with Abridge’s ambient documentation technology yielded notable results: 92% of patients said their providers were more attentive during visits, and clinicians reported dramatic reductions in after-hours charting, or “pajama time.”
Challenging the conventional wisdom
For Mr. Clarke, this collaboration defies a stubborn perception, often repeated by big tech firms, that healthcare lags far behind other industries. That belief was widespread when he joined the health insurer a decade ago after a career in financial services.
“There’s lots of assumptions all the time that all these other industries are moving at a different pace than healthcare,” he said. “Right now, that is actually not true.”
Ambient documentation technology has already posted quiet but significant wins for patients, care teams, and provider organizations. Its value lies in what happens behind the scenes: recording, interpreting and documenting so efficiently that clinicians can return their full attention to the patient.
Adding a payer into the mix broadens the conversation. Lawmakers in both parties have tried for years to curb prior authorization’s worst effects. Just in June, HHS met with more than 50 health insurers to secure pledges to streamline the process. But corporate declarations rarely translate to results. Health system CEOs remain skeptical, as the process often sits outside their control and remains a bargaining chip in some negotiations, as payers offer to drop certain requirements in exchange for lower reimbursement rates.
If AI can help solve this thorn in providers’ sides, it could reclaim the most precious resource in healthcare: time. That’s what Mr. Sevco sees as the bigger arc.
“Our dream and ambition at AHN is wait time zero, so that patients get care when and where they need it,” he said. “Leveraging AI to cut through bureaucracy, eliminate waste, and follow clear clinical guidelines means we can do it right the first time — in a way that lifts clinician satisfaction, patient satisfaction and lowers the cost of care.”
For Abridge, their solution is anchored in the conversation and subsequent documentation, and integrates with other systems of record in the market that may handle downstream parts of a prior auth workflow.
More importantly, the work marks the start of a different kind of conversation: a departure from the AI arms race providers have feared, in which each side adopts algorithms to outmaneuver the other.
“What does that future look like? I think it’s clear that it’s not a good one, and ultimately it’s not a good one for the patient,” Dr. Rao says.
“Our work with Highmark and AHN shows that it doesn’t need to be that sort of negative-sum mindset of AI versus AI,” he says. “It doesn’t need to be a showdown. Instead it can be a partnership, and AI — pun intended — can be a tool that is a bridge between payer and provider.”
