Helix Specialty PharmacyWinning the prior-authorization fight.

Specialty drugs almost always require prior authorization, the payer's sign-off before a therapy can be dispensed. For Helix, that one process gates revenue, delays patients, and consumes the staff. Every new referral lands as a pile of unstructured documents, and the clock starts ticking. The faster a PA moves, the faster the patient starts therapy and the faster Helix gets paid. It is, unambiguously, the constraint on growth.

We don't automate all of prior authorization on day one. We start at the narrowest, most painful point: intake. Referrals arrive by fax, portal, phone, and PDF (prescriber notes, lab results, diagnosis codes, insurance cards) in no consistent format. Staff spend hours reading, sorting, and keying before any clinical work begins.

An AI-native intake surface ingests every referral whatever the format, extracts the patient, prescriber, drug, diagnosis, and insurance details, checks for what's missing, and assembles a clean, structured PA case. The clerks stop transcribing and handle only the exceptions the system flags.

That first surface is funded the same way as everywhere: by redirecting the budget for the next intake hire Helix was about to make. Small, rolling investment. The freed staff guide what comes next.

Clean intake exposes the next constraint: everything downstream. So the team builds outward, one control surface at a time: eligibility and benefits checks, matching the clinical packet against each payer's criteria, generating and submitting the authorization, tracking status and appeals. Each piece lifts the next bottleneck the moment intake stops being the binding one.

Built one at a time, the surfaces share a data layer. Once intake, eligibility, criteria, submission, and status all write to the same place, the combined system develops abilities no single step was built for. In a regulated setting, those are exactly the abilities that protect both revenue and patients.

The accumulated case history lets the engine score approval likelihood before submission, so weak packets get strengthened first. Knowing each payer's behavior enables payer-specific routing, submitting the way each plan actually approves. And every denial feeds denial-pattern learning that pre-empts the next one. None of these were the goal of any single build. They emerged from the combination.

Turnaround on a prior authorization drops from days to hours. First-pass approval rates climb because packets go out stronger. The same clinical staff handle far more volume, reviewing exceptions instead of assembling paperwork. Helix takes on more prescribers and more therapies without the PA backlog that used to cap its growth. And, as everywhere, each stage paid for the next.