The 3 cost layers of an AI receptionist
Most vendors quote a monthly subscription fee and leave the rest implied. The real cost structure has three distinct layers — and confusing them is how companies end up with sticker shock 6 months in.
Layer 1: Infrastructure (per-minute)
Every AI phone call burns compute in real time: speech-to-text (STT) converts the caller's voice to text, a large language model (LLM) generates the response, text-to-speech (TTS) converts it back to audio, and a telephony layer handles the PSTN call itself. Combined, these run $0.07–$0.33 per minute.
On managed platforms like Retell AI or Vapi with bring-your-own-key (BYOK) pricing, you're looking at $0.21–$0.33/min. If you run 3,000 minutes of calls per month — about 1,500 two-minute calls — that's $630–$990/mo in infrastructure alone. Self-hosted LiveKit at 50K+ minutes/month cuts that to $0.07–$0.15/min.
Layer 2: Build cost (one-time)
The AI doesn't come pre-configured to know your hours, your staff, your booking system, or your escalation rules. Building that — conversation design, integration with your scheduling software, CRM logging, error handling, HIPAA compliance if needed — is a one-time engineering investment.
- Proof of Concept (demo on your real calls): $5K–$15K, 2–4 weeks
- Production build with appointment booking + CRM integration: $25K–$40K
- HIPAA-compliant build with multi-location scheduling: $40K–$75K
- Enterprise-scale with compliance, multiple languages, complex orchestration: $75K–$150K
Layer 3: Ongoing operations (monthly)
After the build, you're paying for platform fees, telephony numbers ($1–3/number/month), call infrastructure, monitoring, and periodic prompt and integration updates as your business evolves. Budget $200–$800/mo depending on call volume and complexity. This is where the ROI math lives.
AI receptionist vs human receptionist vs answering service
| Factor | AI Receptionist | Human Receptionist | Answering Service |
|---|---|---|---|
| Monthly cost | $200–$800/mo | $2,900–$4,200/mo | $200–$1,200/mo |
| Annual cost (all-in) | $2.4K–$9.6K/yr | $47K–$63K/yr | $2.4K–$14.4K/yr |
| Available hours | 24/7 | Business hours | 24/7 |
| Books appointments | Yes | Yes | No (message only) |
| CRM integration | Yes | Sometimes | No |
| HIPAA-ready | Yes (with proper setup) | Yes | Varies |
| Handles 10+ simultaneous calls | Yes | No | No |
The ROI math for a dental practice
Here's a real-world example. A dental practice with 200 inbound calls per month and a 22% miss rate:
The break-even point is about 1.5 recovered appointments per month. Most practices see 10–15. That's not a marginal improvement — it's a structural one.
What affects your AI receptionist cost
Frequently asked questions
How much does an AI receptionist cost per month?
$200–$800/mo in ongoing costs once built. The build itself is a one-time cost of $5K–$40K depending on integration complexity. A Proof of Concept to test the use case on your real calls runs $5K–$15K and delivers in 2–4 weeks.
Is an AI receptionist cheaper than a human receptionist?
Significantly. A human receptionist costs $35K–$45K in salary alone — plus $12K–$18K in payroll taxes, benefits, and overhead — totaling $47K–$63K/yr. An AI receptionist costs $2.4K–$9.6K/yr to operate after the initial build. That's an 80–90% reduction in ongoing cost.
What is the per-minute cost of an AI phone receptionist?
Between $0.07 and $0.33 per minute. Managed platforms like Retell or Vapi with your own LLM keys (BYOK) run $0.21–$0.33/min. Self-hosted infrastructure on LiveKit at 50K+ minutes/month drops to $0.07–$0.15/min. For most practices running 2K–10K minutes/month, managed platforms are cheaper when you factor in DevOps overhead.
How much does it cost to build a custom AI receptionist?
A production-ready AI receptionist with appointment booking, CRM integration, and call routing runs $25K–$40K for most businesses. HIPAA-compliant builds with multi-location scheduling (dental groups, medical networks) run $40K–$75K. We also offer a fixed-scope PoC at $5K–$15K — a working AI receptionist on your real calls in 2–4 weeks, before you commit to a full build.
What's the difference between an AI receptionist and an answering service?
An answering service is staffed by humans — often offshore — who take messages and sometimes forward calls. It doesn't book appointments, integrate with your scheduling system, or handle 50 simultaneous callers. An AI receptionist does all three, at comparable monthly cost but dramatically more functionality. Answering services work for simple message-taking; they fall short for practices that need live appointment booking.
What can an AI receptionist do?
Answer every inbound call 24/7 (no hold music, no voicemail), book appointments directly into your scheduling system, handle FAQ queries (hours, pricing, directions, insurance, availability), route urgent matters to a human, send SMS confirmations, and log every interaction to your CRM. It handles 70–80% of inbound call volume without human involvement. Exceptions: document signing, complex complaints, and situations requiring genuine human judgment.