Here’s something nobody really cares about and almost nobody knows is there: specialty pharmacy is one of the biggest, fastest-growing, most operationally broken parts of American healthcare, and it’s hiding in plain sight.
Let me give you the shape of it. In 2025, retail, mail, long-term care, and specialty pharmacies dispensed an estimated $293.4 billion in specialty pharmaceuticals — a 9.6% jump over the prior year. The broader specialty pharmaceuticals market sat around $285 billion in 2025 and is forecast to grow at a blistering clip into the 2030s, depending on whose model you believe. Oncology alone makes up roughly 47% of specialty drug spend, followed by blood disorders and inflammatory conditions. And between 2025 and 2027, more than 80 new specialty drugs are expected to launch across some 70 rare conditions, with annual costs running anywhere from $6,000 to $3.2 million per patient. A quarter of that spend is cell and gene therapy.
Now hold that next to how the work actually gets done. These aren’t pills you grab off a CVS shelf. The big retail chains own something like 90% of the prescription market by volume, but specialty drugs are a different animal entirely — genetic conditions, cancer, autoimmune disease, gene therapy. They’re expensive, they’re fragile, they’re often covered through a maze of Medicare, Medicaid, and commercial benefits, and every single one comes wrapped in a patient who needs to be called, verified, enrolled, reminded, refilled, and monitored. The operational surface area per prescription is enormous.
And how do specialty pharmacies handle that surface area today? Mostly by phone. Manually. I met an owner at a conference who told me his pharmacy runs about 5,000 active prescriptions and makes roughly 80,000 calls a month — by hand. When I heard that I genuinely didn’t know whether to laugh or start a company. (I did the second one.)
That’s the blind spot. Everybody in voice AI is fighting over restaurants, retail, insurance outbound — the crowded, commoditized stuff. Meanwhile here’s a vertical that is enormous, growing double digits, structurally dependent on outbound and inbound phone communication, and almost completely untouched by purpose-built automation. The market is also fragmented: most specialty pharmacies today are independent, so there’s no single gatekeeper you have to win, and every successful deployment becomes a case study that opens the next door.
Why it’s underserved, not just unserved
There’s a reason this hasn’t been solved yet, and it’s worth being honest about. Specialty pharmacy is hard. The conversations are clinical and emotional at the same time. The data lives in pharmacy management systems most generalist AI vendors have never heard of. There’s real regulation — accreditation standards, FDA reporting obligations, privacy rules. A generic “AI receptionist” bolted onto this will fail, loudly, and probably get someone hurt. So the easy money has stayed away, which is exactly why the opportunity is still sitting here.
Where voice AI actually fits
The work that’s eating specialty pharmacy alive is, conveniently, the work voice AI is now genuinely good at: structured, repetitive, high-volume phone conversations that still require a human touch. Refill reminders. Adherence check-ins. Order-status calls. Benefit verification follow-ups. Cold-chain delivery confirmations. Triage of clinical questions to a live pharmacist. A modern voice agent can place and take these calls with real empathy, look up a patient record mid-conversation, update data in the system of record, remember the last conversation, and even carry the thread over to SMS when the patient would rather text.
I’m not going to pretend voice AI replaces a clinical pharmacist. It doesn’t, and it shouldn’t. What it does is take the 80,000 manual calls and turn most of them into automated, logged, compliant conversations, so the humans can spend their time on the patients who actually need a human.
The thesis in one line
Specialty pharmacy is a $290-billion-and-climbing industry running its patient communication on manual phone calls, in a fragmented market, with no incumbent automation player. That’s not a feature gap. That’s a category waiting to be built. The rest of this collection is about the specific places where it breaks — and where the automation pays for itself.