How to Build Your First AI Phone Agent in 15 Minutes
David Shemang · 22 June 2026
What we're building
By the end of this guide, you'll have a live AI phone agent that:
- Answers calls from a Nigerian +234 number - Greets callers in Nigerian English - Handles appointment booking for a dental clinic - Transfers complex cases to a human receptionist - Sends you a notification when a booking is made
Total time: about 15 minutes. No code required.
Step 1: Create your agent (3 minutes)
Sign in to [Magana](https://magana-app.devit.i.ng) and click New agent on the dashboard.
Give your agent a name — we'll call ours "Bright Smile Dental Reception".
In the System prompt field, write your agent's instructions. Think of this like the briefing you'd give a new receptionist on their first day:
``
You are the AI receptionist for Bright Smile Dental Clinic, Lagos.
Your job is to handle appointment bookings.
When a caller contacts you: 1. Greet them warmly in Nigerian English 2. Ask if they're a new or returning patient 3. Ask what type of appointment they need (check-up, emergency, specific treatment) 4. Ask for their preferred date and time 5. Confirm their name and phone number 6. Let them know the clinic will confirm within 1 hour via WhatsApp
Clinic hours: Monday to Saturday, 8am to 6pm. Address: 15 Marina Road, Lagos Island. Phone: 01-234-5678
If a caller has a dental emergency, transfer them immediately to Dr. Adaeze at 0803-XXX-XXXX.
If you can't answer a question, say "Let me connect you with our receptionist."
`
Step 2: Choose a voice (1 minute)
Go to the Voice tab in your agent settings.
Browse the Deepgram Aura-2 voice library. For a professional dental clinic in Lagos, we recommend Aura Asteria (clear, warm, Nigerian-English friendly) or Aura Luna (slightly warmer, more conversational).
You can preview each voice with a sample sentence. Pick one, adjust the speed to 1.0, and save.
Step 3: Add a webhook action (5 minutes)
To actually record bookings, we need to send data somewhere when a caller books an appointment. We'll use a simple webhook.
Go to the Actions tab and click Add action.
- Name: Record appointment - Type: Custom Webhook - URL: Your booking system's webhook URL (or a tool like Make.com / n8n) - Description (for the AI): "Use this action to record an appointment booking when a caller confirms their preferred date, time, and treatment type."
Add the fields you want extracted:
- patient_name — Caller's full name
- phone_number — Caller's phone number
- appointment_type — Type of appointment (check-up, cleaning, emergency, other)
- preferred_date — Preferred appointment date
- preferred_time` — Preferred time slot
Click Test to verify the webhook works with sample data, then Save.
Step 4: Assign a phone number (2 minutes)
Go to Phone Numbers in the sidebar and click Get a number.
Select a Nigerian (+234) number — Lagos area codes are available. The number costs ₦22,000/month and is included free on all Magana plans.
Assign the number to your Bright Smile agent.
Step 5: Test your agent (3 minutes)
Before going live, use the Test Call tab in your agent dashboard.
Click Start call and speak to your agent from your browser. Try a realistic conversation:
> "Hello, I'd like to book a check-up for next Thursday afternoon."
Watch the transcript update in real time. Check that: - The agent greets you warmly - It asks the right qualifying questions - It tries to trigger the webhook action when you confirm a booking - The voice sounds natural
If anything is off, adjust your system prompt and re-test.
Step 6: Go live
Once you're happy with the test call, your agent is live. Calls to your +234 number will connect immediately.
You'll see every call in the Calls section — with transcripts, recordings, and AI-generated analysis of how each call went.
What's next
Now that your basic agent is running, here are some things to try next:
- Add a knowledge base — Upload your clinic's FAQ and price list so the agent can answer detailed questions - Enable call analysis — Magana automatically analyses every call for sentiment, success, and topics — no manual review needed - Set up outbound reminders — Build a second agent that calls patients 24 hours before their appointment
*Start building your agent now — [it's free to try →](https://magana-app.devit.i.ng/sign-up)*