Prior Authorization is Healthcare's Hardest AI Problem, says Sagility's Anand Biradar

A patient waits. A treatment has been ordered, but before it can happen, the provider needs approval from the insurer. That process, prior authorization, can take days or weeks depending on the case. Nearly 95% of physicians say it delays access to necessary care, according to the American Medical Association's 2025 survey.

Anand Biradar, senior vice president of GTM strategy, enablement and enterprise transformation at Sagility, has spent two decades in healthcare operations. "Sometimes PAs can take a week, two weeks[,] or sometimes even more, depending on the case," Biradar said to AIM Media House.

AI is now being tested across the industry to speed that process up. Prior authorization remains, in his view, the hardest piece of the puzzle to solve. Here's why.

What Prior Authorization Actually Involves

Prior authorization is tied to the specifics of a person's health plan. An employer negotiates coverage terms with an insurer, and depending on the plan, certain procedures require sign-off before a provider can deliver them. If a service falls into that bucket, the hospital or clinic has to secure approval first.

"[Prior authorization] is really to avoid unnecessary waste of procedures," Biradar said.

The burden this creates is significant. Physicians complete an average of 40 prior authorization requests per week, and 32% say those requests are often or always denied, per the AMA survey. Nearly 79% report that patients abandon treatment altogether because of the delays.

This is the point in the healthcare system where administrative process and patient outcomes meet directly.

Why AI Hasn't Solved It

Enterprise AI adoption broadly has struggled to move past pilot stage. Research from MIT's NANDA initiative found that 95% of generative AI pilots across industries failed to produce measurable financial impact, largely because tools weren't built to adapt to real organizational workflows.

Biradar sees a version of that same problem in healthcare, magnified by the stakes involved. "[AI companies] understand the AI world better, but they don't understand the operating world," Biradar said. "There were a lot of design mistakes. I'm talking Fortune 50 companies that we work with. Absolute design error."

Healthcare has leaned instead toward what Biradar calls assist-level AI, where a system supports a claims reviewer or case manager rather than replacing their judgment. Regulatory variation between states makes a wrong call too costly to risk without a human checkpoint, he said.

That caution has policy backing. In 2025, four states passed laws restricting insurers from using AI alone to deny medical necessity or prior authorization claims, including Arizona's HB 2175 and Maryland's HB 820. The federal government's own pilot has drawn similar scrutiny.

CMS launched its WISeR model in January 2026, testing AI-assisted prior authorization in traditional Medicare across six states. Participating vendors earn a share of what CMS calls averted expenditures, a structure critics say could favor denials. The Senate voted in July 2026 to keep the pilot running despite those concerns.

The Path Forward Is Incremental

Biradar doesn't expect prior authorization to disappear. "Eliminating prior auth is not a solve. Reducing [it] is happening, as you can see," Biradar said.

Over the next year, he expects visible but partial progress. Some plans are already using AI for smaller, contained tasks, such as reaching out to patients about outdated provider information or handling basic triage before a case reaches a human reviewer.

Larger, end-to-end redesigns will take longer for national payers managing scale across millions of members. Regional and smaller plans, working with more contained systems, are likely to move faster on full automation, he said.

Some of that pressure is coming from a shift toward accountable care, where insurers and providers operate as a single entity rather than separate parties negotiating over approvals. Kaiser Permanente is the most established example of this model, commonly known as a payvider. Biradar pointed to Kaiser as one of the few organizations executing it well.

Not every company attempting the same structure has matched its results, he said. The current administration has also pushed insurers toward more accountable care arrangements, a trend Biradar expects to keep shaping how utilization management, and by extension prior authorization, gets redesigned.

The broader market reflects that same uneven pace. Vendors including Innovaccer and Infinx have built tools that compress certain prior authorization steps from minutes to seconds, according to a 2026 market review. Those gains tend to apply to narrow, well-defined workflows.

Prior authorization sits where cost control, regulation and patient care meet. The caution slowing AI's rollout reflects an industry aware of what's at stake when the outcome is a person's access to care.