Purpose-Built Workflow App

Private Medical Practice

Workflow automation for a solo Respiratory Therapist managing patient assignments from a phone — now with an AI agent that makes the calls.


A solo Respiratory Therapist managing 10 to 20 new patients a day — scheduling video setup sessions, logging call attempts, sending Zoom links. He was running it all from his inbox and memory. The work of tracking patients competed with the work of caring for them.

The practice

An independent RT practice focused on CPAP equipment setup — the critical first appointment that determines whether a patient actually succeeds with their therapy. Assignments arrive from referral partners twice daily, each one a name, a cell number, and a window to reach them. Miss the window, miss the patient. There's no front desk. There's no admin. There's just his phone.

The problem

Assignments came in by email with no consistent format. Pulling out names and numbers meant reading every email manually. Call history lived in a notes app or, worse, nowhere. Knowing who needed a Zoom link sent that morning required scanning back through a calendar. Three missed calls and a patient times out — a declined assignment and a referral partner who notices. The system wasn't broken. It just didn't exist.

What we built

A mobile-first workflow app that runs the full patient lifecycle from a single screen.

Gmail is connected and scanned automatically. Each new assignment email gets read — regardless of format — and the patient name, cell number, and landline are pulled out. Patients appear in the dashboard, ready to call. Each card tracks touch points: one tap for no answer, one tap for answered, one tap for voicemail. On a successful call, the RT picks an appointment time, selects FaceTime or Zoom based on the patient's platform, and the calendar event is created automatically. Zoom links go out roughly two hours before each appointment, no morning checklist required.

Then we let the agent dial

Once the manual workflow was solid, we added an optional AI voice agent. One tap on a patient card and the agent places the call, confirms the patient's identity, asks FaceTime or Zoom, offers open slots from the live calendar, and books straight in. Voicemails get a short, human-sounding scheduling message. Anything ambiguous hands cleanly back to the RT.

The card auto-refreshes while the call is on the wire. When it ends, the outcome lands as a touch point with an AI badge — same row, same workflow, just a different caller. The stats page shows calls placed, bookings made, estimated cost, and RT minutes saved.

What changed

Every patient has a logged status. Every attempt is timestamped. Declining after three no-answers is one tap. Rescheduling updates the calendar event in place — no duplicate invites. Patients who go quiet can be parked and resumed later. Nightly backups run on their own.

At a glance: who's new, who's in progress, who's scheduled, who's been declined, who the agent is calling right now.

The RT still runs the appointments. Everything else, the app handles — or the agent does, and the app keeps the record.


Nothing falls through. Nothing needs to be remembered.