What an AI receptionist should and shouldn't do
An AI receptionist can answer enquiries, qualify customers and book appointments around the clock. Here is how to plan and build one responsibly.
For: Service businesses that miss calls and messages
Lorenzo DevereuxFounder & CEO, Devereux Digitals

Many enquiries arrive when nobody is available: evenings, weekends, lunch breaks and busy days on site. Each missed call or unanswered message is a potential customer who may simply contact a competitor instead.
An AI receptionist answers those enquiries instantly, on your website, on WhatsApp or by phone. Built well, it gives accurate answers, collects the right details and books appointments straight into your calendar.
What an AI receptionist actually does
- Greets the customer
- Answers questions
- Asks qualifying details
- Checks availability
- Books & confirms
- Hands over if needed
Choose your channels
- Website chat: the easiest place to start.
- WhatsApp: ideal where customers already message you.
- Phone: voice AI can answer calls, which suits trades, clinics and salons that miss many calls.
Build its knowledge base
The assistant should answer from your own information rather than general knowledge from the internet. Gather your services, prices or price ranges, opening hours, service areas, policies and the questions customers ask most. Modern AI platforms use this content to ground their answers.
Define qualification and booking rules
- Qualifying questionsDecide what you need before booking: service type, location, urgency and budget range.
- Booking logicConnect to your calendar or booking system, with rules for duration, buffers and who can be booked.
- ConfirmationSend confirmations and reminders by email, SMS or WhatsApp.
- CRM loggingSave the conversation summary and details to your CRM automatically.
Design the human handover
Decide which situations always go to a person: complaints, complex quotes, urgent issues and anything sensitive. The handover should pass along the conversation history so the customer doesn’t have to repeat themselves.
Privacy and transparency
Tell customers they’re talking to an AI assistant, collect only the data you need, and make sure your providers handle data in line with applicable privacy law such as GDPR. Keep a clear record of what data is stored and where.
Launch, review and improve
Review transcripts weekly at first. Look for questions it couldn’t answer, places where customers dropped off and any incorrect responses, then update the knowledge base and rules.
Illustrative example: a plumbing business that misses calls on the job
Imagine a plumbing business whose engineers can’t answer the phone while working. Many callers hang up and try the next plumber, and evening and weekend calls go to voicemail.
An AI receptionist could answer every call and website chat, identify whether the issue is an emergency and collect the address and a description. It could then book a standard visit into the calendar or immediately alert the on call engineer to an urgent job. It would answer common questions about call out charges and coverage areas from the business’s own price guide, and say so honestly when a question needs a person.
The owner reviews call summaries each morning instead of returning a long list of voicemails.
Common mistakes to avoid
- Letting the AI answer without your own knowledge base.
- Hiding the fact that customers are talking to an AI.
- No escalation path for urgent or sensitive cases.
- Launching without reviewing real conversations.
How to measure success
Track the share of enquiries handled without human help, bookings made by the assistant, handover rate, customer satisfaction and the accuracy of answers based on transcript reviews.
Also measure missed calls or unanswered messages before and after launch. This is often where the clearest value appears.
Your action checklist
- Choose the first channel: chat, WhatsApp or phone.
- Write a knowledge base from your real FAQs.
- Define qualification and booking rules.
- Design the human handover.
- Check privacy and data handling.
- Review transcripts weekly.
Questions to ask before you start
- When do we currently miss enquiries?
- Which questions could be answered from our existing information?
- What must always go to a person?
- Which calendar or booking system should it connect to?
Tools and technologies
- OpenAI or Anthropic models
- Voiceflow or Botpress
- Vapi or Retell (voice)
- Calendly or Cal.com
- WhatsApp Business Platform
- HubSpot
Frequently asked questions
Conclusion
An AI receptionist helps every enquiry get a prompt, accurate response and puts qualified customers in your calendar, even when your team is busy or offline.


