What does Forbes mean by 'automating prior authorization backwards'?

In a June 30, 2026 article, Forbes published analysis from Ramya Ganti, CEO of Oprox, arguing that most healthcare AI tools treat prior authorization as a form-filling problem — gathering data, populating fields, and accelerating submission. What they don't do is replicate the reasoning process that determines whether a payer reviewer will approve or deny. The result: providers submit faster but experience the same denial rates. As Ganti put it, 'a faster fax machine is still a fax machine.'

What is the difference between form-filling PA automation and reasoning-first AI?

Form-filling PA automation accelerates the mechanics of submission — gathering clinical data, populating payer-specific forms, and transmitting requests faster. Reasoning-first AI replicates the payer reviewer's decision-making process before submission: analyzing clinical documentation against specific payer criteria, predicting approval likelihood, identifying documentation gaps, and pre-building the appeal case for any denial. The distinction is between speed (form-filling) and accuracy (reasoning).

Why do 88% of prior authorization denials go unchallenged?

According to KFF data on Medicare prior authorization, only 11.5% of denied PA requests are appealed. The primary reasons are: manual appeal processes that require clinical documentation assembly most billing staff aren't trained for, volume overwhelm (a practice generating 50+ denials per week can't manually appeal all of them), and lack of visibility into which denials are actually overturnable. The result is that 88% of recoverable revenue goes completely unchallenged — despite an 80.7% appeal overturn rate.

How much does prior authorization cost the healthcare system annually?

Prior authorization costs the US healthcare system approximately $35 billion annually, according to Health Affairs Scholar research cited in Forbes (June 30, 2026). The AMA reports that physicians spend 13 hours per week navigating PA requirements, and 78% of physicians say PA delays cause patients to abandon recommended treatment entirely. These costs are absorbed disproportionately by providers in staff time, delayed payments, abandoned treatments, and revenue that's denied but never appealed.

Can reasoning-first AI replace a billing company for prior authorization?

Yes. Billing companies bolt form-filling tools onto manual workflows — they accelerate submission but don't address the reasoning gap that causes denials. Reasoning-first AI owns the full chain: payer criteria analysis, clinical reasoning alignment, approval likelihood prediction, pre-submission correction, automated submission, and evidence-based appeal generation. The M&A signal confirms the shift: Experity acquired Exdion Healthcare (July 1, 2026) for AI-driven RCM automation, and the market is consolidating around full-stack AI platforms that replace fragmented billing company services.

Why Your Billing Company Is Automating Prior Authorization Backwards — And What Reasoning-First AI Actually Fixes

Most prior authorization automation has been solving the wrong problem. That's not a hot take — it's the conclusion of a Forbes analysis published June 30, 2026, authored by Ramya Ganti, CEO of Oprox. The core argument: the entire PA automation industry has been building faster fax machines while the actual problem — the reasoning that determines approval or denial — goes completely unaddressed.

The data backs it up. Only 11.5% of denied prior authorization requests are ever appealed. But when they are? 80.7% are overturned (KFF Medicare data). That means the vast majority of PA denials are wrong or overturnable — and 88% of that recoverable revenue vanishes because nobody does the work. The system-wide cost: $35 billion annually (Health Affairs Scholar). And your billing company? It's part of the problem.

The Forbes Diagnosis: Form-Filling Isn't the Work

Ganti's Forbes analysis cuts through a decade of PA automation marketing with one sentence:

"A faster fax machine is still a fax machine."

— Ramya Ganti, CEO of Oprox, Forbes (June 30, 2026)

The critique is specific and structural. Most PA automation tools — and most billing companies deploying them — treat the form as the work. They focus on gathering clinical data, populating payer-specific fields, and accelerating the submission process. What they don't do: replicate the reasoning process that determines whether a payer reviewer is likely to approve.

This is the distinction between speed and accuracy. A billing company that submits your PA requests 40% faster sounds like progress — until you realize the denial rate stays exactly the same. You're just getting denied faster.

Prior authorization is fundamentally two problems, not one:

A reasoning challenge at the front end — does this clinical case meet the specific criteria this specific payer applies for this specific service?
A recovery challenge at the back end — when the answer is no, can you assemble the evidence to prove the payer's decision was wrong?

Form-filling automation addresses neither. It optimizes the middle — the mechanics of submission — while leaving both the reasoning gap and the recovery gap untouched.

The $35 Billion Reasoning Gap

The scale of the problem is enormous and quantifiable:

$35B
Annual system-wide prior authorization cost (Health Affairs Scholar, cited in Forbes June 30, 2026)

That $35 billion isn't just payer administrative overhead. It's absorbed disproportionately by providers in the form of:

13 hours per week of physician time navigating PA requirements (AMA)
78% of physicians reporting PA delays that cause patients to abandon recommended treatment entirely (AMA)
Billing staff hours spent on phone holds, portal submissions, fax follow-ups, and manual appeal drafting
Invisible revenue loss from denied-then-abandoned claims that never enter the appeal pipeline

Your billing company absorbs some of this cost — and passes it back to you in the form of percentage-of-collections fees. They have no structural incentive to reduce denials because denied claims that generate appeal work are billable hours. The fee model rewards activity, not accuracy.

The 80.7% Proof: Your Denials Are Wrong

The KFF Medicare data on prior authorization creates a logical proof that's hard to argue with:

80.7%
Of appealed PA denials are overturned — but only 11.5% are ever appealed (KFF)

Follow the logic:

If denials were correct, appeals should rarely succeed
Appeals succeed 80.7% of the time
Therefore, the vast majority of denials are incorrect or overturnable
But only 11.5% of denials are appealed
88% of recoverable revenue goes completely unchallenged

This is not a documentation problem or a coding problem. It's a reasoning problem. The payer applied criteria. The provider's clinical case met those criteria. But nobody on the provider side did the reasoning work to demonstrate that alignment — either before submission (to prevent the denial) or after denial (to overturn it).

Your billing company doesn't do this reasoning work because they can't. Their PA workflow is: receive the order, fill the form, submit it, track the status, and escalate denials to your staff. The reasoning step — does this clinical case actually satisfy this payer's specific approval criteria for this specific service? — is a clinical intelligence function that requires analyzing payer criteria documents, mapping clinical evidence to specific approval thresholds, and pre-building the evidentiary case. Billing companies don't have the architecture for that.

What Reasoning-First AI Does Differently

Reasoning-first prior authorization AI inverts the entire workflow. Instead of starting with the form, it starts with the payer's decision criteria and works backward to determine whether the clinical case meets them — before anything gets submitted.

Pre-Submission Payer Criteria Analysis

The AI analyzes each payer's specific approval criteria for each service code. Not the generic criteria published in provider manuals — the actual decision rules that payer reviewers apply, including LCD/NCD policies, clinical pathway requirements, documentation thresholds, and step-therapy protocols. This analysis happens before the PA request is assembled, not after it's denied.

Clinical Reasoning Alignment

Once the payer criteria are mapped, the AI evaluates the patient's clinical documentation against those specific criteria. Does the chart note contain the clinical indicators the payer requires? Is the medical necessity language aligned with the payer's approval threshold? Are the diagnostic findings documented in the format the payer's reviewers expect? Any gaps are identified and flagged for the clinical team to address before submission — converting a likely denial into a clean approval.

Approval Likelihood Prediction

Based on historical approval patterns, payer-specific rules, and the current clinical case, the AI generates an approval probability score. Claims with high probability proceed through automated submission. Claims with moderate probability are routed for documentation enhancement. Claims with low probability trigger a pre-built appeal strategy — the appeal is prepared before the denial even arrives, cutting resolution time from weeks to hours.

Automated Evidence-Based Appeals

When denials do occur, reasoning-first AI doesn't start from scratch. The clinical evidence was already analyzed at the pre-submission stage. The specific payer criteria were already mapped. The AI assembles the appeal by matching the patient's documented clinical evidence to the exact criteria the payer cited in the denial — producing a specific, evidence-based appeal letter that addresses the payer's stated reasoning, not a generic template.

This is what converts the 11.5% appeal rate into something approaching 100% — and unlocks the 80.7% overturn rate at scale.

Why Billing Companies Can't Do This

The structural limitation isn't talent or technology — it's the business model. Billing companies are organized around transaction processing: receive claims, scrub them, submit them, follow up on unpaid ones. PA is handled as a subset of that workflow — another form to fill, another status to track.

Reasoning-first PA requires a fundamentally different architecture:

Capability	Billing Company	Reasoning-First AI
Payer criteria analysis	Generic checklists	Payer-specific decision rule mapping
Clinical reasoning	None — form-filling only	Chart-to-criteria alignment scoring
Denial prediction	Reactive — after denial arrives	Pre-submission probability scoring
Appeal generation	Manual — template letters	Auto-generated, evidence-mapped
Learning loop	None — same process each time	Continuous improvement from outcomes
Coverage	Business hours, staff-limited	24/7, unlimited volume

Billing companies bolt form-filling tools onto manual workflows. Reasoning-first AI owns the full reasoning chain from payer criteria analysis through submission to appeal. One is an incremental improvement on a broken process. The other replaces the process entirely.

The M&A Signal: Full-Stack AI Is Consolidating

The market is voting with capital. In the first week of July 2026 alone:

Experity acquired Exdion Healthcare (July 1, 2026) — Experity serves approximately half of all US urgent care clinics and acquired an AI-driven SaaS platform for RCM automation. This isn't a point-solution acquisition; it's a platform play to embed AI across the entire revenue cycle.
Greenway Health CCO Troy Wasilefsky described the shift "from fragmented point solutions to unified AI-powered platforms that manage the full revenue cycle end-to-end" (Healthcare IT Today, July 2, 2026)
Inovalon President Karly Rowe stated that "no individual or team can keep pace with growing complexity of payer rules" — the explicit case for AI reasoning over human processing (Healthcare IT Today, July 2, 2026)
TruBridge quantified the prevention advantage: providers can "identify and stop deniable claims before they are created — preventing $50-$60 per claim in costly rework"
HFS Research reported Optum PreCheck Prior Authorization delivering "near-instant approvals, reducing median drug approval time to under 30 seconds" (July 3, 2026)

The consolidation pattern is clear: the market is moving from fragmented billing company services to integrated AI platforms that handle the reasoning, not just the forms. Companies that only fill forms are acquisition targets. Companies that do the reasoning are acquirers.

The Upstream Shift: Moving AI Left

The most telling signal comes from HFS Research's analysis of where AI intervention creates the most value. The finding: health plans must "shift left" — moving AI intervention upstream, before the claim is submitted, before the PA request is generated, before the denial exists.

This is the exact opposite of how billing companies operate. Billing companies are downstream by design. They receive claims after the clinical encounter, process PA requests after the order is placed, and handle denials after the damage is done. Every step is reactive.

Reasoning-first AI operates upstream:

Before scheduling: Verify eligibility and benefits, identify PA requirements for planned services
Before ordering: Analyze whether the documented clinical case meets the specific payer's approval criteria
Before submitting: Score approval likelihood, flag documentation gaps, pre-build the appeal case
At submission: Route through the optimal channel with complete, criteria-aligned documentation
After denial: Deploy the pre-built appeal within hours, not weeks

By the time a billing company even sees the PA request, reasoning-first AI has already determined whether it'll be approved, fixed the gaps that would cause denial, and prepared the backup plan. The billing company's involvement becomes redundant.

The CodaMetrix Signal: Reasoning, Not Just Automation

CodaMetrix's CTO stated explicitly that AI optimizes the revenue cycle "through reasoning, not just automation" (Healthcare IT Today, July 2, 2026). This framing matters because it validates the structural distinction: the healthcare AI market is bifurcating between tools that automate faster and systems that reason better.

Form-filling automation is a commodity. Every major EHR vendor, billing company, and RCM platform offers some version of "submit PA requests faster." The differentiation has collapsed to zero because the underlying approach — speed up the form — has hit its ceiling. You can't fix a 30% denial rate by submitting 40% faster. The math doesn't work.

Reasoning-first AI creates a structural advantage because it changes the output: fewer denials, higher approval rates, faster appeals, and a learning loop that improves with every decision. That's not a feature improvement on the billing company model. It's a replacement.

What This Means for Your Practice in Q3 2026

Three decisions converge right now:

Your billing company contract is a depreciating asset. Every quarter you pay percentage-of-collections fees for form-filling PA automation, you're paying premium rates for a commodity service that the market has already decided is insufficient. The Forbes analysis, the M&A wave, and the CodaMetrix/Greenway/Inovalon statements all confirm: form-filling is yesterday's architecture.
The reasoning gap is quantifiable. Take your practice's PA denial rate, multiply by your average PA denial value, and calculate 80.7% of that — that's the revenue you'd recover if you appealed every denial. Then multiply by 88% — that's the revenue you're currently leaving on the table. For a mid-size medical practice, this is typically $200K-$500K annually.
The competitive window is closing. Optum is delivering sub-30-second PA approvals. Experity just acquired full-stack AI RCM capability for half the urgent care market. Payer-side AI is accelerating. Practices that deploy reasoning-first AI in Q3 2026 build the pattern library, train the learning loop, and compound the advantage. Practices that renew their billing company contract spend another year bleeding revenue to a gap that widens monthly.

The billing company model was built for a world where PA was a paper process and human reviewers made decisions at human speed. That world is gone. The question isn't whether reasoning-first AI replaces billing companies — the M&A market has already answered that. The question is whether your practice makes the switch before or after the competitive gap becomes permanent.