An AI receptionist answers enquiries, books appointments, and handles routine customer questions around the clock - so a missed call or an after-hours email stops being a lost customer. This is a step-by-step guide to creating one (or a customer support agent, which is the same build with a different job) on cto AI Business: from capturing enquiries in the Inbox, to booking and escalation, to the approvals that keep it safe. Free to pilot.
TL;DR. Build it as a small AI team: a Team Lead that owns the conversation, a Member that handles bookings and lookups, a knowledge base of your business info, and approvals on the high-stakes actions. cto gives you the team, the Inbox to receive written enquiries, and MCP connections to your calendar and (for phone) a call gateway. Start on the free tier with approvals on everything, then loosen.
Receptionist vs support agent - same build
Both are the same shape - an agent that fields inbound messages, answers from your information, does a few actions (book, look up, route), and escalates what it can't handle. The difference is the job:
- AI receptionist - answers enquiries, qualifies, and books appointments. See the AI receptionist use case.
- AI customer support agent - answers product/service questions and resolves routine issues. See the AI customer support agent use case.
This guide is the how-to; those pages are the worked examples.
What you'll need
- A place for enquiries to land. cto gives each team a dedicated Inbox (email address) for written enquiries. For phone, connect a call gateway (Twilio, Vonage, or similar) via MCP so the team can hear and answer calls.
- Your business info - hours, services, pricing, policies, FAQs - so the agent answers correctly.
- Your calendar or booking tool connected as an integration, if it books.
Step by step
- Create the AI business. Write a Plan: "Answer inbound enquiries 24/7, answer common questions from our info, book appointments, and route anything urgent or unusual to a human."
- Give it your knowledge. Add your hours, services, pricing, and FAQs so replies are accurate - this is what separates a helpful agent from a generic one.
- Connect the channels. Point your enquiry email at the Inbox; for phone, wire a call gateway via MCP.
- Hire the Member(s). Capability for booking (calendar write), customer/order lookup, knowledge-base Q&A, and escalation. One well-scoped Member can cover several; split them if you want strict separation.
- Set approvals. In week one, have bookings and outbound replies wait for your sign-off, and set anything unusual to escalate. Loosen once the drafts are reliable.
- Pilot free, then scale. The free tier covers low volume; premium handles production.
What it does live
A question arrives - by email or call - "Are you open Saturday and can you fit a boiler service?" The agent answers from your info, checks the calendar, offers two slots, books the pick, and sends a confirmation. A question it can't answer, or an angry customer, gets routed to you with the context attached. Routine work handled; judgment calls escalated.
Keep the guardrails on
- Escalate the hard stuff - complaints, refunds, anything off-script - to a human with context.
- Approve bookings and outbound messages until the drafts have earned your trust.
- Give it accurate info - a receptionist confidently quoting the wrong price is worse than a missed call. Keep the knowledge base current.
Build vs buy
If you'd rather not build, packaged AI receptionist services (Smith.ai, Goodcall, Rosie) exist and are covered in the AI receptionist use case. Building on cto AI Business gives you more control - your own team, models, tools, and approval rules - and a free tier to pilot on. Packaged services are faster to switch on but less flexible.
FAQ
How do I create an AI receptionist?
Build it as a small AI team on cto AI Business: write a Plan, add your business info, connect your enquiry channels (email Inbox, or a phone gateway via MCP) and calendar, hire a Member to handle bookings and Q&A, and set approvals on bookings and outbound messages. Free to pilot.
Can an AI receptionist answer phone calls?
Yes - connect a call gateway (Twilio, Vonage, or similar) via MCP and the team can answer calls. For written enquiries, cto's built-in Inbox catches email without any extra setup.
What's the difference between an AI receptionist and a support agent?
Same build, different job: a receptionist answers enquiries and books appointments; a support agent answers questions and resolves routine issues. Both field inbound messages, act, and escalate what they can't handle.
Is an AI receptionist reliable enough to trust with customers?
With guardrails, yes: keep approvals on bookings and outbound replies, escalate anything unusual to a human, and keep its knowledge base accurate. Start with approvals on everything and loosen as it proves out.
How much does an AI receptionist cost to build?
On cto AI Business, free to pilot (ad-supported, with rolling limits) - enough to run a small business's volume. Premium handles production volume. Cheaper than a packaged answering service or a hire.