AI Disruption in Pharma Services: Market Structure Implications

Thesis

The market-structure thesis is hybridization. AI-native entrants will win wedges where the workflow is standardized and the buyer can measure transaction-level ROI, but incumbents with pharma contracts, clinical escalation teams, data rights, and compliance infrastructure will remain the natural prime contractor for complex specialty programs. The highest-probability outcome is therefore not a clean startup displacement cycle; it is a repricing of FTE-heavy hub work, with incumbents buying, partnering with, or embedding AI point solutions while manufacturers demand cost-per-transaction transparency.

The Capital Wave

AI investment in healthcare automation — particularly prior authorization, benefit verification, and patient access workflows — experienced explosive growth in 2024-2025. The capital deployed signals that investors believe AI can fundamentally restructure the economics of pharma services call centers.

Company	Category	Total Funding	Key Round	Primary Technology	Status
Tandem AI	PA/Rx automation	$137M	$100M round (Jan 2026), led by Accel/Thrive/General Catalyst. $1B valuation (unicorn).	Automates paperwork and phone calls for PA and pharmacy routing. Free to providers.	Growth stage; unicorn
Infinitus Systems	Voice AI for PA	$103M	Series C $51.5M (Oct 2024), led by a16z	”Eva” voice AI agent: 6M+ calls, 100M+ minutes automated, 150K+ providers, 1,400+ payers. Patented hallucination-prevention.	Growth stage; production-proven
Tennr	Document AI	$150M+	Series C $101M (2025), $605M valuation	RaeLM vision-language model trained on 100M healthcare documents, 2.3B data fields, 8K payer criteria sets. 10M documents/month.	Growth stage
Cohere Health	Payer-side PA AI	$90M+	Series C $90M (May 2025), led by Temasek	Payer-side AI: 12M PA requests/year, auto-approves up to 90%. 660K+ providers. Oct 2025 Microsoft partnership for ambient PA.	Growth stage; payer-focused
SuperDial	Voice AI for PA	$20M+	Series A $15M (Jun 2025), led by SignalFire	1M+ calls completed. 4x productivity gains for billing teams. Mar 2026 Omega Healthcare partnership. Cartesia voice AI.	Early growth
Neon Health	AI-native hub	$6M	Seed $6M, led by NFX	Full-stack specialty drug access automation (BV, PA, copay, adherence). Claims 2x speed, 80% cost reduction. ~6 pharma brands.	Early stage
Coral AI	AI PA automation	Undisclosed (Lightspeed seed)	Seed (2024)	500K monthly workflows (2025 milestone); revenues scaled 8x. Reads referral packets, checks eligibility, reasons through clinical criteria.	Early growth

Total AI PA/hub investment (2023-2026): $500M+ in disclosed funding. PA AI spending grew from $10M (2024) to $100M (2025) — 10x YoY. Healthcare AI spending hit $1.4B in 2025 (nearly 3x 2024), with healthcare adopting AI 2.2x faster than the broader economy (Menlo Ventures).

Cautionary precedent: Olive AI — Valued at $4B after $400M raise (July 2021). Shut down October 31, 2023. Promised to replace humans in healthcare admin, but complex workflows broke the AI. Sold pieces to Waystar and Humata Health.

What AI Actually Automates Today

Prior Authorization (Highest AI Penetration)

The PA workflow has the most standardized, repetitive structure — making it the most amenable to automation:

Information gathering: Extracting patient demographics, diagnosis codes, medication details, and prior therapy history from EHR/claims data → AI document extraction (Tennr)
Payer rules matching: Identifying the specific PA criteria for the patient’s insurance plan and drug → Rules engines and NLP
Phone-based PA submission: Calling payer phone lines, navigating IVR systems, providing clinical information to payer representatives → Voice AI (Infinitus, SuperDial)
Fax-based PA submission: Filling and faxing PA forms with correct clinical documentation → Document AI (Tennr, Coral AI)
Status checking: Calling payers to check PA determination status → Voice AI
Appeals: Generating letters of medical necessity, scheduling peer-to-peer reviews → AI-assisted but still requiring clinical judgment

Current automation rates (vendor claims, unaudited):

Infinitus: “millions of calls” processed; claims significant time and cost savings
Neon Health: Claims 98% automation rate, 80% cost reduction, 2x speed (unverified)
SuperDial: Focus on provider office workflows; claims 60-80% reduction in staff time for PA calls
100ms: Voice-first automation spanning intake, benefits verification, and PA-adjacent workflows; relevant when buyers want one layer across specialty-access operations instead of a single-call use case
RISA Labs: Oncology-specific orchestration across benefits, prior auth, denials, and status tracking; closer to a disease-line operating system than a generic PA bot
Traditional hub benchmarks: EVERSANA claims 92-96% ePA acceptance rate vs. 73% industry rejection rate

Benefits Verification (Medium AI Penetration)

Electronic BV (eBV) is already largely automated through platform-to-payer API connections
AI adds value in: interpreting complex benefit designs, identifying accumulator program exposure, predicting coverage outcomes
CareTria claims 85% of benefit investigations under 60 seconds
EVERSANA claims 90% eBV success rate vs. 43% industry average

Copay/Financial Assistance (Low AI Penetration)

Copay card adjudication is already automated at point-of-sale
AI opportunity: predicting accumulator/maximizer exposure, optimizing copay program design, identifying eligible patients from claims data
Phil’s digital-first model automates patient enrollment in 2 minutes (vs. days for traditional hub enrollment)

Adherence/Nurse Support (Lowest AI Penetration)

Clinical counseling, injection training, and side effect management remain human-intensive
AI opportunity: predicting non-adherence risk before it occurs (Amber’s ML model), automating routine refill reminders, triaging clinical escalations
This is the last hub function that will be significantly automated

Impact on Hub Services Call Center Models

The Bull Case for AI Disruption

If AI PA vendors’ claims are even directionally correct, the implications for traditional hub services are significant:

ConnectiveRx model (illustrative):

~1,400 employees across 12 locations on 3 continents (multiple sources confirm 1,098-1,404 range); grew headcount 6% last year
FTE-based revenue model at ~$55K loaded cost / ~$85-110K bill rate
CTO Steve Randall publicly presenting on “evaluating AI for patient outcomes in hub services”
Current strategy: “Apply automation and AI where it makes sense… then surround that by human touch” — augmentation, not replacement
If AI automates 50% of call center volume, ~500-700 FTEs become redundant
Revenue impact: $27-55M+ in annual FTE revenue at risk (for ConnectiveRx alone)

Industry-wide:

Global pharma hub and patient access support market: $3.6B (2025), projected to reach $7.6-10B by 2033-2035 (CAGR 10.8-11%)
Call center operations represent an estimated 50-60% of hub costs
McKinsey: AI-enabled PA can automate 50-75% of manual tasks; could save U.S. healthcare $200-360B annually (5-10% of total spend)
Front-office RCM total addressable market: $98 billion annually — software is only 3% currently (Menlo Ventures: “80% of the market is still completely untapped”)
AMA 2024 Survey: Physicians/staff spend 13 hours per week on 39 PA requests per physician. Manual PA call averages 24 minutes.
Transition from FTE-based to technology-based pricing fundamentally changes hub vendor economics

The Bear Case (Why Disruption Will Be Slower Than Investors Hope)

Payer system fragmentation: There are 1,000+ unique payer organizations in the US, each with different PA workflows, phone systems, fax requirements, and clinical criteria. AI must handle this fragmentation at scale.
Clinical judgment requirements: Complex PA (rare disease, gene therapy, oncology) requires clinical reasoning that current AI cannot reliably provide. The “easy” 60-70% of PA volume is automatable; the hard 30-40% still needs humans.
Regulatory uncertainty: CMS-0057 mandates electronic PA APIs by January 2027, which could make voice AI for phone-based PA partially obsolete (if payers build compliant APIs, you don’t need to call them).
Hub vendor response: Established hub vendors are investing in their own AI capabilities. ConnectiveRx, EVERSANA, and AssistRx are all building AI-powered PA and eBV tools. The AI startups may be acquired rather than displacing incumbents. A parallel pattern is emerging on the outsourcing side: Omega Healthcare is acting as a distribution layer for automation through its 2026 SuperDial partnership, suggesting BPOs may become AI channels rather than simple victims of automation.
Manufacturer switching costs: Switching hub vendors is operationally complex (data migration, patient communication, workflow redesign). Manufacturers may prefer their existing hub vendor adding AI capabilities over switching to an AI-native startup.
Audit and accountability: Manufacturers and payers need human accountability for PA decisions. AI-generated PAs that result in adverse patient outcomes create liability questions that haven’t been resolved.
The AI arms race risk: Stanford researchers published in Health Affairs (2025) on the “AI arms race in health insurance utilization review” — payers using AI to deny, providers using AI to submit/appeal, creating escalation loops that may increase administrative complexity rather than reduce it. 37% of insurers using or planning AI for PA within a year; 70% of large-employer groups exploring AI for PA. CMS launching WISeR pilot in 6 states for Medicare PA.
The Olive AI precedent: $4B valuation to shutdown in 2 years. Complex healthcare workflows broke the AI, requiring manual intervention that destroyed the economics. The current crop appears more grounded (Infinitus 6M+ calls, Tennr 10M documents/month), but Olive demands skepticism about hype cycles.

CMS-0057: The Regulatory Wildcard

The CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F) is the most significant regulatory development affecting PA technology:

Requirements:

Payers must implement four HL7 FHIR R4-based APIs: Patient Access API, Provider Access API, Payer-to-Payer API, and Prior Authorization API
Electronic PA submission, real-time status checking, and decision communication via FHIR
Applies to Medicare Advantage, Medicaid managed care, CHIP managed care, and Qualified Health Plans on ACA exchanges
Payers must publicly report PA metrics (approval rates, turnaround times, denial reasons)

Timeline:

January 1, 2026: Faster PA turnarounds (7 calendar days standard, 72 hours urgent/expedited). Specific denial reasons required. Public PA metrics reporting begins.
March 31, 2026: First public PA metrics due from payers
January 1, 2027: Full API compliance deadline — all four FHIR APIs must be in production

Current state: Only 35% of medical PAs are currently fully electronic (X12 278). CAQH estimates fully electronic PA could save the industry $454M annually. Manual PA costs $12.88/transaction vs. $0.05 electronic.

Impact on AI PA vendors:

Positive: Standardized APIs make electronic PA integration easier; FHIR reduces the need for phone/fax-based PA
Negative: If payer APIs work well, the value of voice AI for phone-based PA (Infinitus, SuperDial) decreases significantly — the problem they solve (navigating payer phone systems) becomes less relevant
Net: CMS-0057 benefits platform-based PA vendors (CoverMyMeds, Surescripts) and hub vendors with API integration capability over point-solution voice AI vendors

Market Structure Scenarios (2026-2030)

Scenario 1: AI Augments, Doesn’t Replace (Most Likely — 60% probability)

AI handles 40-60% of routine PA, BV, and status-check volume
Hub vendors integrate AI into existing platforms (build or acquire AI startup capabilities)
Specialist entrants like 100ms and RISA Labs win narrow but important wedges in specialty-pharmacy and oncology workflows rather than broad horizontal share
FTE headcount declines 20-30% over 5 years through attrition, not mass layoffs
Hub pricing shifts from pure FTE-based to blended (FTE + technology platform fee)
AI PA startups are acquired by hub incumbents (ConnectiveRx acquires Neon Health-type capabilities, EVERSANA builds in-house)
Manufacturer costs decrease 15-25% on a per-program basis

Scenario 2: AI Disrupts Traditional Hub Model (25% probability)

AI handles 70-80% of PA and BV volume
AI-native vendors (Neon Health, Infinitus) win significant market share from traditional hubs
Traditional hub vendors lose pricing power as AI alternatives demonstrate dramatically lower costs
Hub industry consolidation accelerates as weaker vendors can’t invest in AI
Manufacturer costs decrease 30-50% on PA/BV functions; clinical/nurse functions remain human-intensive
Hub vendor margin profiles shift dramatically (technology margins vs. labor margins)

Scenario 3: CMS-0057 Resolves the Problem (15% probability)

Payer FHIR APIs work well; electronic PA becomes standard without phone/fax
Voice AI vendors (Infinitus, SuperDial) lose primary use case
Hub vendors benefit from standardized electronic connectivity
PA technology market consolidates around platform vendors (CoverMyMeds, Surescripts)
AI investment shifts from PA automation to adherence prediction, patient finding, and clinical decision support

Implications

If you believe…	Then select…
AI will augment hubs	Established hub vendor (ConnectiveRx, EVERSANA) with AI investment roadmap
AI will disrupt hubs	AI-native vendor (Neon Health) or split strategy: AI vendor for PA/BV + traditional for clinical
CMS-0057 solves PA	Hub vendor with strong payer API connectivity (CoverMyMeds integration, Surescripts network access)
Uncertain	Established hub vendor with contractual AI performance improvement commitments and pricing flexibility

Key Takeaways

AI investment in PA automation is real and significant ($500M+ in 2023-2026). This capital will produce technology that changes hub vendor economics.
The “easy” 60-70% of PA volume is automatable today. Complex therapeutic areas (rare disease, gene therapy, oncology) require clinical judgment that AI can’t yet provide.
CMS-0057 is the regulatory wildcard. If payer FHIR APIs work, voice AI for phone-based PA becomes less relevant.
Hub vendors will adapt. The most likely outcome is AI augmentation of existing hub platforms, not wholesale replacement. Established vendors have customer relationships, data, and clinical infrastructure that AI startups lack.
Manufacturer action: Negotiate AI performance improvement clauses in hub contracts, benchmark AI-enabled vendors against traditional vendors on cost-per-PA and time-to-therapy, and monitor CMS-0057 implementation progress.

AI & Automation in Pharma Services
AI Prior Authorization Vendors Comparison
Hub Services Market Analysis
Hub Services Pricing Benchmarking
Prior Authorization in Specialty Pharma

Auto-generated cross-references closing audit-surfaced link gaps. Vendors named in the prose above without inline links are listed here so the wiki graph is queryable.