How a Dhaka clinic cut no-shows by 40% with an AI voice agent

Most clinic owners don't realize how expensive an unanswered call is — until they count.

A small dental practice in Gulshan, Dhaka was running 6 chairs, 5 days a week. Their front desk was busy with walk-ins. After-hours calls went to voicemail. About 18 to 22 calls per week were missed — roughly 1,000 calls per year. Even if only 30% of those were new-patient booking intents, that's 300 lost patients a year.

For a clinic billing $40–$80 per visit, that's $12,000–$24,000 of foregone revenue, walking out the door every year.

The system we built

We didn't rip out their existing process. We plugged the leak.

1. AI voice agent (Vapi + ElevenLabs)

When a call comes in — at 2pm or 2am — the voice agent picks up in 1.2 seconds. It:

• Greets the caller in a calm, professional voice (the same voice, every time)
• Captures patient name, reason for visit, and preferred time slot
• Checks the clinic's Google Calendar in real-time
• Books the appointment if the slot is open
• Sends a confirmation via WhatsApp + SMS (Twilio)
• Schedules an automated reminder 2 hours before the appointment

If the patient asks a question the agent can't answer (insurance, complex procedures), it routes the call to the front desk during working hours — or queues it for the next morning.

2. The connective tissue (n8n)

Behind the voice agent, an n8n workflow does the heavy lifting:

• Receives the call transcript from Vapi's webhook
• Formats it into a HubSpot contact record (new patient or existing)
• Triggers the WhatsApp template message
• Adds the appointment to the Google Calendar
• 24 hours before: sends a reminder
• 2 hours before: sends a second reminder
• After the appointment: triggers a feedback request

The whole thing runs without a human in the loop for routine bookings.

The numbers (after 90 days)

• Missed calls: 18–22 per week → 0–2 per week (the 0–2 are callers who explicitly asked to speak to a human)
• No-show rate: 31% → 18% (a 42% relative drop) — automated reminders are the biggest single driver
• New patient bookings per week: 6 → 11 (the after-hours call capture made the difference)
• Front desk time saved: ~10 hours per week (reallocated to in-clinic patient experience)
• Revenue per chair per month: +11%

The system paid for itself in 5 weeks.

What it cost

Setup: $499 (Patient Journey Pro package — full voice agent build, n8n workflow, CRM sync)

Monthly: $99/month (monitoring, prompt tuning, monthly strategy review)

Total Year 1 cost: $499 + $1,188 = $1,687

Conservative revenue lift: ~$14,000/year (just from the no-show reduction + after-hours bookings)

ROI: ~730%.

Why this works

Two things, in order of importance:

1. The voice agent sounds human enough. ElevenLabs' latest models are at the inflection point where callers don't immediately hang up. That wasn't true 18 months ago. The bar has moved.

2. The workflow is deterministic. Every call follows the same path. Every booking triggers the same reminders. There's no "did the receptionist remember to send the SMS?" variable. That's where the no-show drop comes from — not the AI, but the consistency.

If you're running a clinic and your front desk is missing more than 5 calls per week, you should at least price out what an AI voice agent would cost vs. what those missed calls are costing you.

Usually the math isn't close.