The Appeals Arms Race: AI on Both Sides of the Table

Research completed Feb 12, 2026

---

1. Provider-Side AI Tools Generating Appeals

The provider side has a ~1-year head start on payers. Multiple funded startups and incumbents now automate appeal letter generation.

Waystar — AltitudeAI / AltitudeCreate

• Launched: January 2025, rolled out at no additional cost to existing denial/appeal management customers.
• Scale: Targets the 450 million annually denied claims in the US.
• Performance: Creates hundreds of appeal packages simultaneously. 90%+ faster — cuts time from 38 hours to 2 hours per batch. Equivalent of redeploying 13 FTEs at a mid-size health system. Early adopters overturning 40% more denials.
• How: Pulls claim data, denial reason codes, clinical documentation from EHR/clearinghouse integrations. Generative AI drafts payer-specific appeal letters citing medical policy and clinical evidence.
• Market position: Public company (NYSE: WAY), dominant in RCM. At-scale rollout, not a startup experiment.
• Sources: Waystar announcement, CNBC Jan 2025

Cofactor AI

• What: AI platform for hospitals using a proprietary "medical-native foundation model" to review clinical evidence, coding guidelines, peer-reviewed studies, and find discrepancies between claims, documentation, payer policies, and provider contracts.
• Founded: August 2023. Funding: $4M seed led by Drive Capital (Nov 2024).
• Differentiator: Integrates with EHR and clearinghouses. Claims to "instantly find the exact contextual information required to generate winning appeals."
• Sources: MobiHealthNews, Healthcare IT Today

Longtail Health — AppealAssist

• Batch appeal generation. Upload hundreds of denial cases via XML/CSV/text. Auto-ranks by urgency, financial impact, or overturn likelihood. Generates appeal letters for all cases in a single workflow.
• Payer compliance checking built in.

Claimocity

• AI-assisted RCM platform for hospitalists. Automates charge capture, eligibility verification, prior auth, and coding. Less specialized in appeals than Waystar/Cofactor — more full-stack RCM with denial prevention emphasis.

Patient-Facing AI Appeal Tools (new category, 2024-2025)

• Counterforce Health launched early 2025 in Durham, NC.
• Fight Health Insurance built by a former IBM/Apple/Google/Netflix engineer. Open-source, on-device OCR.
• These tools cut appeal prep time from hours to under 60 seconds.
• Sources: PBS NewsHour, Stateline

Apella (NOT appeals-focused)

Apella Health raised $80M Series B in January 2026 but focuses on OR optimization with ambient AI and computer vision. Not an appeals company — included here to prevent name confusion.

---

2. Payer Response: Counter-Systems and Payment Integrity AI

The Explicit Arms Race

Centene CEO Sarah London, July 2025 earnings call: "As [hospitals] integrate AI into the revenue cycle, we're integrating AI into payment integrity to make sure that we are sort of keeping pace with all of that." — Source: Healthcare Brew

UnitedHealth / NaviHealth — nH Predict

• Algorithmic tool predicting appropriate post-acute care stays. UHC set a goal to keep rehabilitation stays within 1% of nH Predict prediction.
• UHC post-acute denial rate: 8.7% (2019) to 22.7% (2022).
• 90% error rate — 9 of 10 appealed denials ultimately reversed.
• Class action proceeding (Estate of Gene B. Lokken v. UnitedHealth Group). Court ruled Feb 2025 that breach of contract and good faith claims can proceed. If plan documents promise physician review, AI gatekeeper = breach of contract.

CVS Health — Post-Acute Analytics

• AI project launched 2021 to reduce SNF spending. Initially projected $10-15M savings over 3 years. Revised upward to $77.3M savings.
• Source: Senate PSI Report

Cohere Health — Payer-Side PA Automation

• Funding: $90M Series C (May 2025, led by Temasek). Total ~$200M.
• Automates up to 90% of prior authorizations. 85% handled in real time. 9 million FHIR-based authorizations processed in 2024.
• Cohere Align (2025): Analyzes provider historical behavior to identify trusted clinicians. 80% of PA submissions streamlined for pre-approved providers.
• Sits upstream of appeals — if PA is auto-approved correctly, fewer appeals.

MHK — AI-Augmented UM Platform

• LLM and Agentic AI "plug-in" capabilities. Integrates medical policy, MCG, InterQual, CMS NCD/LCD criteria within UM workflow.

---

3. Payer-Side Appeals & Grievances Workflow Software

Existing systems that payer-side AI must integrate with or replace:

---

4. Appeals Workflow from the Payer Side

Appeal Types and SLAs

Medicare Appeal Levels

1. Plan-Level Reconsideration (30 days / 72 hours)
2. Independent Review Entity (IRE) (30 days / 72 hours)
3. OMHA — Administrative Law Judge
4. Medicare Appeals Council
5. Federal District Court

Typical Processing Steps and Automation Potential

---

5. Where AI Adds the Most Value for Payers

Ranked by impact:

1. New Information Detection — The threshold question: "did the provider submit new clinical information?" Currently done by manual page-by-page comparison across hundreds of pages. NLP-based semantic diff is the single highest-value application.

2. Document Parsing & Information Extraction — Appeals arrive as faxed PDFs, scanned records, handwritten notes, EHR exports. OCR + NLP to extract diagnoses, procedures, medications, lab values, imaging results. Everything downstream depends on this.

3. Criteria Matching — Automated matching of extracted clinical data against digitized criteria (MCG, InterQual, CMS NCD/LCD, plan medical policy). Reduces reviewer variability, accelerates review from 45-60 min to 10-15 min per case.

4. Response Drafting — Generative AI producing decision rationale letters with clinical citations and regulatory-compliant language. Reduces letter generation from 20-30 min to 2-3 min.

5. Reviewer Routing — Classify clinical domain, match to reviewer expertise and availability, prioritize by SLA deadline and overturn risk.

6. Overturn Risk Scoring — Score each appeal for overturn probability. Focus expensive MD time on cases where it matters. Auto-approve high-probability-of-overturn cases to reduce waste.

---

6. Financial and Volume Data

Scale

Appeal and Overturn Rates

Market Growth

• AI prior authorization spending: $10M (2024) to $100M (2025) — 10x YoY
• 71% of health insurers using AI for UM (2025)
• 84% of large health insurers using AI for some operational purpose (NAIC 2024 survey)
• Insurance AI spend expected to grow 25%+ in 2026

---

7. Regulatory Context

CMS-0057-F (Effective January 1, 2026)

• Expedited PA response: 72 hours. Standard: 7 calendar days.
• Must provide specific denial reasons (not boilerplate).
• Public PA metrics reporting starting March 31, 2026.
• Four APIs required by January 2027 (Patient Access, Provider Access, Payer-to-Payer, Prior Auth).

Colorado AI Act (SB24-205) — Enforcement June 30, 2026

• First state to regulate AI in healthcare decisions.
• Requires disclosure when AI used in consequential decisions.
• Must allow appeals of AI-generated adverse decisions.
• Requires clinician review of automated decisions.
• Prohibits fully automated denials.

ERISA Implications

• ERISA preempts state laws for self-insured plans — creating a regulatory gap where AI denials are most aggressive.
• Key cases: Lokken v. UnitedHealth (breach of contract for AI replacing physician review), Cigna algorithm case (state UCL claims survived ERISA preemption via savings clause).
• Pattern: if plan documents promise physician review, using AI as gatekeeper = breach of contract.

NAIC Survey (2024)

• 84% of large health insurers using AI operationally. 37% for PA, 44% for claims adjudication, 56% for UM.

Senate PSI Report (October 2024)

• Documented UHC, CVS, Humana using AI/predictive tech to increase post-acute care denial rates.
• UHC denial rate: 8.7% to 22.7%. CVS projected $77.3M savings from AI denials.

---

8. Implications for DaisyAI at Premera

The Cycle

1. Payers deploy AI to increase denial rates (nH Predict, Post-Acute Analytics)
2. Providers deploy AI to generate appeals faster (Waystar, Cofactor) — appeal volume and quality increase, 40% more overturns
3. Payers need AI to process the incoming wave of AI-generated appeals — this is where DaisyAI enters
4. Regulatory response forces clinician oversight, specific denial reasons, public reporting

Competitive Gap

Nobody is building AI-native clinical appeals review for payers. Existing tools handle workflow (routing, SLAs, correspondence). Nobody does clinical reasoning — new information detection, criteria matching, evidence summarization — at the appeal review step.

Our Wedge

Document parsing + new information detection + criteria matching. The highest-value steps that are currently 100% manual. We assist nurse reviewers and medical directors — decision support, not automated decisions. This architecture is regulation-safe (Colorado Act, CMS requirements, ERISA litigation patterns).

The $18 Billion Argument

70% of denials that are appealed get overturned. $18 billion annually is wasted arguing over claims that should have been paid. AI that helps payers make better initial decisions (fewer bad denials) and process appeals faster and more accurately is worth real money to every payer in the country.