Skrypt Health

How Skrypt Health gave a New York clinic ~190 staff hours back and books ~70 appointments a week with Fish Audio.

Skrypt Health runs Ani, a HIPAA-compliant AI receptionist, on Fish Audio's voice runtime. From Sep 29 to Oct 30, 2026, Ani gave one family medicine practice in New York about 190 staff hours back, and now writes about 70 appointments a week straight into the clinic's EHR/PMS. Here is how the two layers fit, how Ani is sent out on outbound calls, and how clinics change her themselves.

Products used
Fish Audio voice runtime
Use Case
Healthcare

How Skrypt Health runs Ani on Fish Audio

Skrypt Health runs Ani, a HIPAA-compliant AI receptionist, on Fish Audio's voice runtime. From Sep 29 to Oct 30, 2026, Ani gave one family medicine practice in New York about 190 staff hours back, and now writes about 70 appointments a week straight into the clinic's EHR/PMS. Here is how the two layers fit, how Ani is sent out on outbound calls, and how clinics change her themselves.

"We're not a voice company. We build the AI that runs a clinic's operations. Fish Audio is the part that lets us stop worrying about the voice and spend our time on what the call has to actually do for the clinic."

Hasan Ahmed, Founder & CEO, Skrypt Health

About Skrypt Health

Skrypt Health builds the AI employees that run a clinic's day-to-day operations. Its receptionist, Ani, answers every call, books straight into the clinic's EHR/PMS, takes prescription refill requests, writes lab outcomes back into the EHR and follows up with patients on lab results. Callers who couldn't get through get a text; the front desk gets a clean daily queue in the Skrypt portal. Dental, veterinary and medical clinics across North America run on it, and every one of those actions touches patient health information.

The problem

A good receptionist is easy to describe and hard to run.

Before Ani, every call that came in after hours went to voicemail.

Skrypt owns the clinic side of a call: what the practice offers, when it's open, which requests are safe to book and which belong with a person. What it needed from a voice layer was simpler to say than to find. The voice had to sound like the practice, stay natural when a caller interrupts, wait quietly while a tool checks the schedule, and hand a warm transfer to the front desk without dropping the thread.

It also had to be something a ten-person team could ship against every day. Skrypt runs one configuration per clinic line, proves every change on a development line before it touches production, and holds some features back per clinic until the practice says go. A voice runtime that forced one global prompt, or couldn't be driven from code, would have turned every clinic change into a support ticket.

"Every clinic is a little different and they change their minds weekly. If the voice can't be configured like infrastructure, you end up hand-editing prompts at 11pm."

Hasan Ahmed

The solution

Ani runs on Fish Audio's voice runtime. Each clinic line is a configuration Skrypt owns in code: the prompt, the voice, the tools Ani may call and the rules that decide when it books, when it takes a message and when it transfers to a person. Development and production lines are separate, so a change is proven on a test line and published to the live line on purpose.

  • Answer in the practice's voice. Every inbound call is picked up around the clock with the clinic's hours, services and scheduling rules already loaded. Ani says what it's doing while a tool runs, so the caller never hears dead air.
  • Book straight into the schedule. When the request is eligible, Ani checks availability and writes the appointment into the clinic's EHR/PMS. It lands on the clinic's own schedule, where staff move or cancel it like any other visit.
  • Transfer with a briefing. When a caller needs a person, Ani warm-transfers to the front desk and briefs the staff member before connecting. The caller doesn't repeat the story.
  • Text back whoever got away. Callers who hang up early get a text from the clinic, and the request lands in the daily queue with the reason it needs a human.

Fish Audio's tool-call validation keeps the model honest: a request with a missing or malformed argument is rejected and Ani asks the caller instead of guessing. The clinical judgment stays with the clinic. Ani doesn't answer medical questions and doesn't improvise on anything the configuration didn't approve.

"We change a clinic's prompt in the morning, test it on the dev line, publish in the afternoon, and the front desk never knows anything moved. Fish Audio made the voice the easy part."

Hasan Ahmed

The result

From Sep 29 to Oct 30, 2026, with Ani running on Fish Audio, one family medicine practice in New York sent 4,714 calls from 1,930 different callers through Ani. 3,223 of them (68%) were handled end to end with no human in the loop, at an average of 85 seconds each.

That adds up to about 190 staff hours in one month. Ani spent 119 hours on the phone herself. On top of that sits the after-call work the front desk no longer does: 600 refill write-ups (about 30 hours), 355 after-hours and weekend calls that used to become voicemail callbacks (about 30 hours), bookings written into the EHR/PMS (about 12 hours) and 104 lab-result messages (about 5 hours).

In October, once Ani was answering first on every call, she wrote 308 bookings straight into the clinic's EHR/PMS: about 70 a week, or 10 every working day. 54 of them were new patients, about 12 a week. 29% of callers who wanted to book were booked on the spot, and 13 of the after-hours calls turned into bookings. At an estimated $180 per office visit, that is about $55k a month in booked visits.

600 prescription refill requests were captured and routed to the pharmacy desk. The practice signed at the end of July and took its first live calls on Aug 10, about two weeks later. Ani handled after-hours and missed calls in September and every call, first in line, by October. Over the five weeks on Fish Audio, a ten-person team shipped 33 production versions of the line's configuration, every one of them proven on the development line first.

Day to day the change is quieter than the numbers. The front desk starts from a booking on the schedule, a reschedule with a defined request, a transfer that arrives with context, or a message that says why it stopped. Nobody reconstructs a voicemail.

What it means for clinics

Inbound was the first job. Outbound is the second. Taking the call is one thing; having an employee on demand is another. The same Ani that answers the phone gets sent out: "We need more information on this medication before the doctor signs it. Send Ani." A lab result comes back and needs a patient conversation: Ani makes the call and writes the outcome back. The runtime that keeps an inbound conversation natural is the runtime that makes an outbound one trustworthy.

Self-serve, with the gaps made native. Fish Audio does the 90%: the voice, the turn-taking, the tool discipline. The industry's usual failure is the other 10%: every clinic has rules the platform didn't anticipate, and the fix is a pile of bolt-on tools that never quite fit. Skrypt's layer exists for exactly that 10%. Whatever a clinic needs that the runtime doesn't have, Skrypt makes native and puts behind a self-serve setting the practice controls. Hours, transfer targets, closures, what Ani may say about a service: the clinic changes it, Skrypt proves it on the dev line, it goes live. The control sits with the clinic, not with a vendor's roadmap.

Skrypt Health signs a Business Associate Agreement with every clinic it serves.

Conclusion

"Fish Audio gave us a voice our clinics trust from the first call. That frees a ten-person team to put every hour into what Ani does for patients — it is how we shipped 33 production versions in five weeks."

Hasan Ahmed, Founder & CEO, Skrypt Health

In one word: "Momentum."


Numbers are from Skrypt Health's own call ledger for one New York family medicine practice and cover calls answered on Fish Audio from Sep 29 to Oct 30, 2026. Weekly booking rates use October only, when Ani was answering first on every call. *Staff hours = Ani's own phone time (119 h) plus after-call work the desk no longer does, at about 3 minutes per refill write-up, booking and lab message and about 5 minutes per avoided voicemail callback (196 h in total). The 29% booking rate is for callers who wanted to book; bookings were partly switched off during the window, so the full-window floor is 27%. Visit value is estimated at $180, the average office visit price in New York. Clinic named only with its permission.

Read more from Skrypt Health

Read more customer stories

Fish Audio

Ready when you are

Talk to our team about your deployment. We'll come prepared.