Home›Blog›AI in Pharma Hub Services: Benefits Verification and PA
Published: November 20, 2026Last updated: November 20, 2026

AI in Pharma Hub Services: Benefits Verification and PA

VA
Vibhor AgnihotriHead of US Business, Zelthy
Share:
AI in Pharma Hub Services: a prescription passes benefits verification and prior authorization gates to a specialty drug vial

A new specialty prescription can sit for days or weeks while a hub team confirms coverage, chases a prior authorization and finds the patient a way to pay. That wait has a cost. IQVIA reports that only 30% of new-to-brand commercial prescriptions were approved within 7 days of being written in 2025, down from over 50% in 2018, and in the American Medical Association's 2025 survey, nearly four in five physicians said they had seen patients abandon treatment because of prior authorization barriers.

Hub services are one component of a wider patient support program (Zelthy's complete guide to patient support programs covers the full model). Zelthy, an AI-native platform built on the open-source Zango framework, is used by pharmaceutical teams to build patient services workflows such as enrollment, benefits verification and financial assistance, with programs deployed across 12+ countries.

This post covers what hub teams do in benefits verification (BV) and prior authorization (PA), why drug PA sits outside new federal API rules, where AI helps today, and where case managers stay in charge.

What do hub services do between the prescription and the first dose?

Hub services are manufacturer-funded teams that take a new prescription, confirm coverage, help the prescriber's office secure any required authorization, arrange financial assistance, and route the prescription to the right pharmacy. They do not make clinical decisions.

In practice, a benefits verification for a specialty drug has six steps:

  1. Intake. An enrollment form, prescription and insurance card arrive, often by fax or e-referral, and are matched to a patient record. (Zelthy's companion post on AI for fax enrollment forms goes deeper on this step.)
  2. Eligibility check. Active coverage is confirmed, typically through the HIPAA standard X12 270/271 eligibility inquiry and response (version 5010) for medical plans; pharmacy transactions use the NCPDP standard instead.
  3. Benefit determination. The team works out whether the drug falls under the medical benefit (often physician-administered infusions or injections) or the pharmacy benefit (often self-administered drugs dispensed by a specialty pharmacy). The answer changes the payer contact, the PA route and the patient's cost share.
  4. Coverage rules. Formulary status, PA requirements, step therapy and quantity limits are checked against the payer's current policy.
  5. Cost share. Deductible, coinsurance and out-of-pocket maximum are estimated so the team can screen for copay or foundation support.
  6. Routing. The prescription goes to an in-network or limited-distribution specialty pharmacy, or the site of care is confirmed for buy-and-bill.

A 270/271 response confirms coverage, but it often does not answer the specialty questions in steps 3 to 5. That gap is why hubs still phone payers. An AJMC study of oral cancer drugs found that 73% of prescriptions needed two or more phone calls and 40% needed five or more, with a median of 12 days from prescription to receipt.

Why prior authorization still sets the pace of time to therapy in 2026

Prior authorization remains the least predictable step because volume is high, criteria vary by payer, and most drug PA still moves through portals, fax and phone. The main federal rule that sets PA deadlines and API requirements, CMS-0057-F, does not yet apply to drugs.

In the AMA's 2025 survey of 1,000 physicians, respondents reported about 40 prior authorization requests and around 13 hours of physician and staff time per week, and more than a quarter said PA had led to a serious adverse event for a patient.

Payer data shows how much of the delay is administrative. A PhRMA-sponsored IQVIA analysis of Medicare Part D found the initial denial rate for first attempts at new branded medicines rose from 37% in 2021 to 47% in 2025, about half of denied attempts were approved within 30 days, and 60% of approvals occurred on the same day as the initial denial. Average time to approval was roughly 11 days, and about 1 in 10 patients waited more than 30 days. Same-day reversals usually point to missing information rather than clinical disagreement, which is friction better data assembly can remove.

The 2025 CAQH Index puts medical prior authorization at 40% electronic adoption in 2025, up from 31% in 2023, compared with 96% for eligibility and benefit verification.

The regulatory split hub teams need to understand

The CMS fact sheet on CMS-0057-F sets out the 72-hour and 7-day timeframes and the January 2027 API deadline for impacted payers. The CMS FAQ explains that drugs were excluded "because the standards, processes, and decision timeframes for issuing prior authorizations for drugs differ," while noting that payers are not prohibited from including drugs covered under a medical benefit. For Part D, CMS requires the NCPDP SCRIPT standard for electronic prior authorization, adopted under section 6062 of the SUPPORT Act. The voluntary commitments are set out in the June 2025 HHS and CMS announcement.

The practical takeaway: for most specialty drugs, hub teams cannot wait for payer APIs. CMS has proposed extending electronic PA, shorter decision timeframes and transparency requirements to drugs (CMS-0062-P, April 2026), but that rule was still a proposal at the time of writing. Near-term gains come from working the manual steps faster and more accurately.

Drug PA split: medical services pass a CMS-0057-F API gate to a 7-day decision, while drug PA runs through NCPDP, portal and fax
CMS-0057-F speeds medical items and services. Drug PA follows a separate path through NCPDP SCRIPT (Part D), payer portals, fax and phone.

Where AI helps in benefits verification and prior authorization today

AI is most useful in hub services where work is repetitive, document-heavy and rule-based: reading intake documents, matching cases to payer policy, pre-filling forms, chasing status and drafting letters. The AI prepares work for a person to check; it does not make coverage or clinical decisions.

The evidence on drafting is encouraging but specific. A 2024 study in JCO Clinical Cancer Informatics found that general-purpose large language models produced clear appeal letters for denied radiotherapy services without confabulating patient histories, but all models were poor at supporting claims with correctly cited literature, and a fine-tuned model performed worse. The lesson for hubs: AI drafts save time, and a clinician still checks every reference.

Copay and foundation screening follow similar patterns.

Where case managers stay in charge, and what payer-side AI controversies teach

Humans stay responsible for clinical judgment, unusual payer conversations and every patient-facing decision. Payer-side experience shows why: when automation drives denials without clinical review, regulators and lawmakers respond.

In October 2024, a U.S. Senate Permanent Subcommittee on Investigations staff report found that in 2022, UnitedHealthcare and CVS denied post-acute care PA requests at about three times their overall denial rates, and Humana at more than 16 times, and recommended limits on how much predictive technology can influence human decisions.

Policy is moving toward human review. California's SB 1120, effective 1 January 2025, requires a licensed physician or qualified provider to decide medical-necessity denials, according to a statement from its author, Senator Josh Becker. Even CMS's WISeR model, which tests AI-assisted PA in traditional Medicare from 2026 to 2031, states that recommendations for non-payment are determined by appropriately licensed clinicians.

Hub teams sit on the other side of the table, but the principle carries over. For case managers, AI means less re-keying and hold time and more time on complex cases and patients, the same pattern seen across AI-enabled patient support programs.

Implementation checklist for AI in BV and PA

  • Confidence routing. Send low-confidence outputs to a person and track override rates.
  • Source traceability. Link every AI output to its source document, 271 response or payer policy.
  • Audit trail. Log who approved what and when.
  • Privacy and security. Hub data is PHI; apply the controls in Zelthy's HIPAA Security Rule guide to any AI vendor or model.
  • Adverse event capture. Call summaries and document extraction should feed pharmacovigilance intake, as covered in Zelthy's post on PSP pharmacovigilance.
  • Metrics. Measure days from referral to BV complete, BV to PA submission, PA submission to decision, decision to first fill, and abandonment, before and after rollout.
Confidence routing: high-confidence cases go straight to the drug, while cases needing review route to a case manager
Clean cases move straight through; uncertain ones go to a case manager.

What this looks like for programs deployed on Zelthy

Zelthy is a platform builder, not a hub operator. Manufacturers and their partners use it to configure patient services workflows, including enrollment, benefits investigation, PA tracking and financial assistance, and to connect them with pharmacy, claims and EHR systems. Teams can configure confidence-based routing, human review steps and audit trails for AI-assisted tasks on Zelthy.

Reported results from programs deployed on Zelthy include a patient support program with 60% faster onboarding, 30% lower cost and a 45% improvement in adherence. A compassionate access program deployed on Zelthy reported 65% faster approvals and 40% lower operational costs, and went live in 5 weeks. These figures come from specific deployments; results depend on program design, payer mix and therapy area. At platform level, Zelthy has supported 1M+ patients across 12+ countries, with 300+ applications deployed. More detail is on the Zelthy Patient Services page.

This article is for general information and is not legal or regulatory advice.

Frequently asked questions

Can AI do prior authorization for specialty drugs?

AI can prepare most of the work: extracting data from referrals, matching cases to payer criteria, pre-filling forms and tracking status. The prescriber still signs and submits the request, and the payer makes the decision. Clinical rationale and appeals need clinician review.

Does the CMS prior authorization rule apply to drugs?

No. CMS-0057-F excludes drugs from its prior authorization API and decision-timeframe requirements because drug PA uses different standards and processes. Payers may voluntarily include medical-benefit drugs in their APIs, and Medicare Part D uses the NCPDP SCRIPT standard for electronic prior authorization. In April 2026 CMS proposed extending the requirements to drugs (CMS-0062-P).

What is benefits verification in pharma hub services?

Benefits verification confirms a patient's coverage for a specific drug, whether it falls under the medical or pharmacy benefit, what PA or step therapy applies, and the expected out-of-pocket cost. Hubs use the result to route the prescription and screen for financial assistance.

How long does prior authorization take for new specialty prescriptions?

It varies widely. IQVIA reports that only 30% of new-to-brand commercial prescriptions were approved within 7 days of being written in 2025. For Medicare Part D, IQVIA found average time to approval of roughly 11 days, with about 1 in 10 patients waiting more than 30 days.

Will AI replace hub case managers?

No. AI removes re-keying, hold time and status checks, which shifts case managers toward complex cases, payer escalations and patient conversations. Regulators and lawmakers are also moving toward requiring licensed humans for medical-necessity decisions.

If you are mapping where AI fits in your hub workflows, book a demo with the Zelthy team.

Sources

  1. "Latest prior auth survey shows promised reform remains elusive," American Medical Association, June 2026.
  2. "CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F)" fact sheet, CMS, January 2024.
  3. Interoperability frequently asked questions (general), CMS.
  4. "E-Prescribing Standards and Requirements," CMS.
  5. CMS-0062-P proposed rule, "Interoperability Standards and Prior Authorization for Drugs," Federal Register, 14 April 2026.
  6. CMS-4189-F final rule, Federal Register, 31 December 2020.
  7. "Health Plan Eligibility Benefit Inquiry and Response," CMS.
  8. "HHS Secretary Kennedy, CMS Administrator Oz secure industry pledge to fix broken prior authorization system," CMS, June 2025.
  9. 2025 CAQH Index Executive Report, CAQH, 2026.
  10. "Beyond payer rejections: creating and sustaining demand," IQVIA, September 2026.
  11. "Increasing payer control for Medicare patients initiating branded medicines," IQVIA (sponsored by PhRMA), March 2026.
  12. Geynisman DM et al., "Provider and patient burdens of obtaining oral anticancer medications," American Journal of Managed Care, April 2018.
  13. Kiser KJ et al., "Large Language Models to Help Appeal Denied Radiotherapy Services," JCO Clinical Cancer Informatics 8:e2400129, September 2024.
  14. "Insurers' AI denials of postacute care face Senate scrutiny," AJMC, October 2024.
  15. "Landmark law prohibits health insurance companies from using AI to deny healthcare coverage," Office of Senator Josh Becker, December 2024.
  16. "WISeR (Wasteful and Inappropriate Service Reduction) Model," CMS Innovation Center.

Similar Blogs