Eighteen months ago, Sonet partnered with a market-leading pharmacy management software provider to deliver their platform in the cloud. The cloud part was hard. The test was harder.

To validate the infrastructure under realistic load, we needed to simulate more than 300 pharmacy technicians using the system at once. Real workflows. Real database calls. Real queue depth. You can't hire 300 technicians for a 15-minute test. So we built AI agents to do it.

That's how Sonet's first AI agents were born. Not as a product. As a load generator.

What We Saw Next

The agents passed the test. The infrastructure held.

But we saw something the test wasn't designed to show us. Our agents had moved through real pharmacy workflows the way a technician would: filling queues, chasing rejections, working refills. The same work pharmacists tell us is eating 40 to 60 percent of their day. Research supports this: a peer-reviewed time-motion study found that pharmacists spend more than half of their time on tasks they themselves classify as non-value-added: prescription prep, administrative documentation, queue management. We started iterating.

Pharmacists are in the business of caring for people. That's the job. We wanted to do something similar for them: use AI to take the mundane work off their plate so a pharmacist can care for more patients, not fewer.

That was the moment we decided pharmacy deserved a product of its own.

The Question I Get Asked

I get asked all the time whether AI is going to replace pharmacists.

The question is fair. People are watching jobs change everywhere: in call centers, in radiology, in software engineering. The fear is rational, and most vendors meet it with the same exhausted phrases. Empower. Augment. Human in the loop. Words that mean nothing because they cost nothing.

I want to give a different answer. One specific enough to argue with.

The Contrarian Frame

Most AI in healthcare is being built to automate the cheapest labor in the system. That's why it's showing up at call centers and intake desks first. Volume work, low cost per hour, easy to defend in a board meeting.

We're building for the opposite end of the pharmacy.

The most trained, most expensive, most clinically valuable person in the building is the pharmacist. The median annual wage for a pharmacist in the United States is $137,480, and demand is growing. The pharmacist is also buried in repetitive work that has nothing to do with their training: insurance rejections, refill calls, queue triage, data entry.

The opportunity isn't to replace the pharmacist. It's to un-bury them.

What Caring Actually Looks Like

What does that work look like, the work only a pharmacist can do?

It's the counseling at the counter. The catch on a prescriber's mistake at 6pm before a patient takes the wrong dose home. The ten extra minutes with a confused Medicare patient who isn't sure which bottle is which. In a lot of communities, especially the ones independent pharmacies serve, the pharmacist is the most accessible healthcare professional a patient will see all year.

Sonet helps the people who help people. Anything that gives a pharmacist back time to do that work is, in a real sense, healthcare infrastructure.

The Line We're Drawing

Here's what PAT (Pharmacy AI Technician) does. And what it does not do.

PAT handles the repeatable, rules-based work that's eating your pharmacists' day: intake, data entry, queue triage, refill outreach, insurance rejection chasing. It works inside the same PMS your team already uses, drives it the way a technician would, and logs every action a pharmacist can review. You can see how PAT's underlying Vision Agent technology works.

PAT does not make clinical decisions. It does not exercise dispensing judgment. It does not counsel patients. Anything that requires a pharmacist's training and license stays with the pharmacist, with the pharmacist in the loop on every clinical decision.

That isn't a feature limitation. It's a product principle. We've spent 18 months working out which parts of a technician's day AI should do, and which parts it absolutely should not.

The Future I'm Betting On

Dispensing margins are compressing. Clinical revenue is where the next decade of independent pharmacy lives: MTM, immunizations, adherence programs, value-based care.

The pharmacy workforce is already strained. In 2024, 73 percent of practicing pharmacists rated their workload as "high" or "excessively high", up from 66 percent in 2014. The pharmacies that win the next decade will be the ones whose pharmacists spend more time on clinical work, not less. AI that handles the repetitive half of the day is the lever that makes that shift possible.

The pharmacist isn't being replaced. The pharmacist's job is being rebuilt around the work only a pharmacist can do.

We're early. We're working with our first design partners. We'll share more as we learn. If you're a pharmacy owner who's been wondering what AI actually means for your business, I'd rather have that conversation than write another post about it.