CacheKetamine

The problem

A clinic’s front desk has one phone line and human hours. Patients call with the same questions — what the treatment is, what it costs, whether it’s right for them — and every missed call is a missed booking. Hiring more front-desk hours scales linearly and badly.

The build

A live voice AI agent that answers the clinic’s phone: it handles patient questions in natural conversation and books appointments during the call, around the clock.

Why voice and not a chatbot: this clinic’s patients pick up the phone — meeting them in the channel they already use beats teaching them a new one. Voice is also the harder, more valuable problem: latency, interruption handling, and natural turn-taking all have to work.

Why the scope stays operational: it’s a healthcare-adjacent build, so the agent handles scheduling and information — operational work — and is deliberately not a medical advice system.

Built on OpenAI voice with AWS hosting.

The method beat

Applied voice AI replacing a real human workflow — the clearest “this is not a demo” build in the set: it has a job, and it shows up for it every day.