Migrating a Pharmacy Voice Agent to ElevenLabs
A voice AI company running a pharmacy's patient phone line
A live pharmacy voice agent moved to ElevenLabs Agents with no backend changes, cutting cost per minute 70% at a 1.1-second response time.
70%
Lower platform and model cost per minute
1.1 s
Response time
400+
Patient calls handled a day
Challenge
A phone line patients rely on
Patients call the agent to refill and renew prescriptions, place orders, arrange deliveries and transfer prescriptions. The line had to stay up while the agent moved to a new platform.
Tools wired into an existing backend
The agent's tools called the client's existing AWS Lambda functions. The migration had to work with that backend as it was.
Heavy context on every call
Each call carried about 100k tokens of context, which weighs on both the cost per minute and how fast the agent can answer.
Approach
Five workflow steps in one agent
We moved the agent from Vapi to ElevenLabs Agents and rebuilt its five workflow steps inside a single agent.
Adapter Lambdas instead of backend changes
Adapter Lambdas sit in front of the client's existing Lambdas and expose eight tools to the agent, so nothing in the backend had to change.
A model chosen by testing
We tested six models and chose Claude Sonnet 5, and trimmed the context each call carries from about 100k tokens to 3k.
Tests before every release, review after every call
A 500-call text-mode test suite runs before each release, and a webhook sends every call to a call-review app.
Outcome
Lower cost, faster answers
Platform and model cost per minute fell 70%. The agent answers in 1.1 seconds with 30x less context per call.
No backend changes, no downtime
The switch needed no changes to the client's backend and caused no downtime. The agent now handles 400+ patient calls a day.
Focus Areas
Services Provided
Contact us
Have a Similar Challenge?
We build production AI systems end to end — from strategy and prototype to a scalable, secure platform. Let's talk about yours.