AI agents
How much does an AI receptionist cost?
AI receptionist cost has two parts: a one-time setup project and ongoing costs that grow with the minutes the agent spends on calls. When the agent runs in your own accounts, as ours do, you pay the AI, phone and speech providers directly and see every usage charge on your own bills.
By the KDS security engineering teamPublished 6 min read
Key takeaways
- AI receptionist cost has two parts: one-time setup work and ongoing costs that grow with call minutes.
- Setup covers call flows, integrations, the knowledge base, testing and attack testing before the agent goes live.
- When the agent runs in your own accounts, you pay AI, phone and speech usage directly to your providers.
- Call volume, integrations, languages, compliance needs and hand-off rules are the main cost drivers.
- Compare the total with the bookings you miss today and the staff hours routine calls take.
On this page
We don't publish prices, and a single number you see online rarely fits your business. The cost depends on your call volume, your systems and your rules. This guide explains each cost part so you can compare quotes and estimate the return with your own numbers.
The three parts of AI receptionist cost
- One-time setup: the project that designs, connects, tests and launches the agent.
- Ongoing service: monitoring, fixes and improvements after launch.
- Usage costs: AI, phone and speech services, billed by your providers based on how much the agent is used.
The first two pay for the work of whoever builds and runs the agent. The third goes to the providers whose services the agent uses on every call.
One-time setup: what the work includes
Setup is a project, and its size depends on how much the agent must do. In our projects, it includes:
- Discovery. We review your call types, call volumes, busy hours and the systems your front desk uses.
- Call flows. The greeting and AI disclosure, booking steps, the questions the agent may answer, transfer rules, after-hours rules and emergency rules.
- Integrations. Connections to your scheduling system, CRM, text messages and phone system. Each one needs access setup, data mapping, error handling and tests.
- Knowledge base. Hours, services, preparation instructions, policies and approved answers, written so the agent can use them.
- Testing. Test calls with realistic requests, accents, background noise and unusual cases, before real callers reach the agent.
- Attack testing. We try to make the agent reveal other callers' details or skip its transfer rules, including through prompt injection: tricking an AI with hidden instructions.
For one location and one scheduling system, a pilot takes 4–6 weeks in most cases, including testing with real calls. More locations, systems or languages add time and cost.
Ongoing costs after launch
- AI usage costs. The language model that runs the conversation is billed by usage, so this cost rises with the number and length of calls.
- Telephony. Your phone provider bills for the number, inbound minutes, forwarded calls and text messages.
- Speech services. Speech-to-text turns the caller's words into text, and text-to-speech gives the agent its voice. Both are usually billed by audio minute or by the amount of text.
- Monitoring and improvements. Someone reviews transcripts, fixes gaps in the knowledge base, updates rules when your services or hours change and reports the results each month.
Usage costs move with call volume. Ask any provider for an estimate at your real volume and at double that volume, and the difference shows how the bill will grow.
Why you pay some costs directly to your providers
We build and configure the AI receptionist on your own infrastructure: your AI provider account, your phone account and your cloud account. We don't host the agent or process your call data on our own systems.
This affects cost in three ways. You pay usage charges directly to each provider at the rates in your own contract, and you see them on your own invoices. You own the phone number, the recordings, the keys and the data, so you can change who maintains the agent without rebuilding it. And our invoice covers only our work: the setup project, then improvements on short contracts with no lock-in.
It also means a few more accounts to manage. We help you set them up and take only the access the work needs. You can remove our access when the work is done.
What raises or lowers the cost
| Cost driver | Lower cost | Higher cost |
|---|---|---|
| Call volume and length | Few, short calls | Many calls or long conversations, which raise usage costs |
| Integrations | One calendar | Several systems, or systems without an API |
| Languages | One language | Several languages, each tested separately |
| Compliance needs | No sensitive data on calls | Health, legal or financial data, with stricter contracts, logging and retention rules |
| Hand-off rules | Transfers to one number | Different routes by topic, location, staff member and time of day |
| Locations | One location | Several locations with different hours and services |
Compliance needs deserve a closer look. A dental or medical practice in the United States needs a written business associate agreement with each provider that handles protected health information on its behalf (45 CFR 164.502(e) (opens in a new tab)), and each agreement takes time to review. In the EU, the greeting must in most cases make clear that the caller is talking to an AI, under Article 50 of the AI Act (opens in a new tab).
Compare AI receptionist cost with missed calls
The useful comparison is not AI against zero. It is AI against what happens today: calls that go to voicemail, callbacks your staff make and hours spent on routine calls. Three simple formulas give you a monthly figure:
- Missed-call value = missed calls per month × share of those callers who would have booked × average value of a booking.
- Staff time value = hours per month spent on routine calls × share the AI can take over × full cost per hour of the person who answers.
- Monthly benefit = missed-call value plus staff time value. Subtract the ongoing monthly costs from your quote to get the net benefit, then divide the setup cost by the net benefit to get the payback period in months.
For a U.S. reference point on staff cost, the median wage for receptionists was $18.27 per hour in May 2025, according to the Bureau of Labor Statistics (opens in a new tab). Your full cost per hour is higher once you add payroll taxes, benefits and overhead, and it varies by region.
To run these numbers faster, use the missed-call cost calculator for lost bookings and the savings calculator on our home page for staff time.
Questions to ask before you sign
- Whose accounts does the agent run in, and who owns the phone number, recordings and transcripts?
- Are usage costs billed to you by the providers, or bundled into a price per minute?
- What does the monthly fee cover after launch: monitoring, fixes and content updates?
- What happens to the cost if your call volume doubles?
- Is attack testing done before the agent goes live?
- What do you keep if you end the contract?
How to get an accurate estimate
- Export 1–3 months of call logs from your phone system: total calls, missed calls, after-hours calls and average call length.
- List your most common call types and mark the ones you want the AI to handle.
- List the systems the AI must use, and check whether each one has an API.
- Write down your hand-off rules and any data the AI must not collect.
- Ask each provider to quote setup, ongoing service and estimated usage costs separately, at your volume.
In a free 30-minute assessment, we look at your call patterns and systems and tell you which cost drivers apply to you. See how the AI receptionist works, or compare it with a live service in AI receptionist vs answering service.
Sources
- 1.Occupational Outlook Handbook: Receptionists, U.S. Bureau of Labor Statistics, 2026-08-27 (opens in a new tab)
- 2.45 CFR § 164.502: Uses and disclosures of protected health information, general rules, Legal Information Institute, Cornell Law School (opens in a new tab)
- 3.Article 50: Transparency obligations for providers and deployers of certain AI systems, European Commission, AI Act Service Desk (opens in a new tab)
