Ask three pharmacy automation vendors what they charge and you will probably sit through three demos before you get a number.

That is a budgeting problem, not a sales annoyance. Independent pharmacy gross margin fell to 18.2 percent in 2024, according to the NCPA Digest. At that margin, an operating line item you cannot forecast is a line item you cannot approve.

So here is the arithmetic, in the open.

A task is not a script

Your pharmacy management system vendor charges you per script. That unit makes sense for them, because it maps to what the software does.

Automation priced per task uses a different unit, and the difference matters when you build a forecast. A task is a defined outcome plus the steps required to reach it. You define where it ends.

• A refill ends where you say it ends. The pharmacist review queue, the print queue, or the point where a human takes over. Whichever end point you write into the logic is the completed task.
• A cycle fill for one patient is one task. Twenty prescriptions in that patient's cycle is still one set of instructions and one defined outcome, so it counts once, not twenty times.
• A prescription placed on hold today and future filled ten days from now is two tasks. Each run starts fresh with no memory of the last one, so each is separately defined and separately completed.

The practical consequence: your task count does not track your script count. Pharmacies running heavy cycle fill and medication synchronization programs often generate far fewer tasks than their script volume would suggest.

How the monthly minimum works

The structure has two numbers and one rule.

• $0.25 per completed task.
• $750 per month minimum, which is the equivalent of 3,000 tasks.
• You pay whichever is greater, never both.

Run 1,000 tasks in a month and the bill is $750. Run 4,000 tasks and the bill is $1,000.

The minimum applies per deployment, not per location. A three-store chain under common ownership runs one deployment: a secure connection at each store, one place to build and manage task logic, and a single $750 monthly floor across all three locations. Separate ownership structures are the exception and are set up separately.

Terms are month to month with 30 days notice. Pilots run for the first two weeks at no charge and carry no obligation. The full model is laid out on the task and pricing page.

What a month actually looks like

Aggregate numbers hide the part that matters, which is how different kinds of work count. Here is an illustrative month for a three-location chain running refills, e-Rx intake, and a cycle fill program.

Note what the cycle fill line does to the total. Those 260 patients might represent several thousand prescriptions, and they cost $65 for the month because the work is defined once per patient rather than once per prescription. That single distinction is usually the difference between a forecast that lands and one that does not.

The floor never enters the picture here, because 6,300 tasks is well past 3,000. It matters at the other end of the scale. A single location running 900 refills and 300 e-Rx entries generates 1,200 tasks, worth $300 of task value, so the bill is the $750 minimum.

If you want a starting estimate before you pull your own reports, the average independent pharmacy fills 217 prescriptions per day, per the NCPA Digest. Your own refill and e-Rx queue counts from last month are a better input, and most pharmacy management systems will produce them in a few minutes.

Three ways to cap the number before anything runs

Predictability is a design question, not a billing question. Three controls sit in front of the invoice.

1. The queue you build sets the ceiling. The AI Technician works from a view you define. If that view holds refills and nothing else, refills are all it can touch, which makes your maximum task count knowable in advance rather than discovered at the end of the month.

2. Nothing loads or schedules without your approval. Task logic is written in plain English, not code. You read it, change it, and approve it before anything is scheduled to run.

3. You can set a hard ceiling. Cap the spend or the task count. Work stops at the cap and you get a notice asking whether to continue, in the same spirit as a data plan alert.

There is also a completion condition worth understanding: billing triggers on reaching the end point you defined. If the end point you defined is to get to a certain step and then hand the work to a pharmacist, then reaching that step is a completed task, because that was the outcome you asked for.

The cost that never shows up on an invoice

The line item is only half of the comparison. The other half is rework, because when automation fails quietly the cleanup lands on a pharmacist and appears on no vendor invoice, which is why the exception path deserves as much scrutiny as the rate.

Two related reads on that: how logic capture keeps the queue working your way in There's No Such Thing as a Standard Refill Queue, and what to press vendors on architecturally in Bots Are Not AI Agents.

Run your own numbers. The task and pricing model shows the full structure, and the calculator at pharmacy.sonet.io works from your actual refill and e-Rx volumes. If you are weighing more than one approach, the Pharmacy Automation Comparison Guide lays out the tradeoffs. For the mechanics of how the AI Technician drives your existing system through the screen, see Vision Agents.

FAQs

How much does PAT cost?
$0.25 per completed task, with a $750 per month minimum equivalent to 3,000 tasks. You pay whichever is greater, never both. Terms are month to month.

Is the $750 minimum charged per pharmacy location?
No. It applies per deployment. A multi-location chain under common ownership runs one deployment with a secure connection at each location and a single monthly minimum across all of them.

Does a cycle fill count as one task or one task per prescription?
One task per patient cycle. A patient with 20 prescriptions in a cycle is one defined outcome and one set of instructions, so it bills once.

Is there an extra charge to build or change task logic?
No. Task creation and iteration are included, and the logic is written in plain English so your own staff can adapt it over time.

How long does a pilot take, and what does it cost?
Pilots run about two weeks at no charge, on one location and one or two workflows, with no contractual obligation.