How to build a smart telephone agent

Questions
, ,
1

:bullseye: What is your goal?

I’m a new business selling AI agents and learning as I go. I have a potential client who is an electrical contractor who is interested in the idea of having an automated phone system that can answer and screen his phone calls between personal and business, and potentially book his appointments for him while he is on a job site. Additionally, it can also act as a 24hr emergency answering system.

:thinking: What is the problem & what have you tried?

I have taken nearly all of the courses in the Make Academy, and I’m not finding a course that helps with the scenario and agent I wish to build. I’m still learning. Because the number of modules seem endless, I’m struggling with where to start my search for the trigger module/webhook and how the connections are made with the client’s phone provider. I want to build and test the system on my own phone, as I have a similar setup to this potential client. I’ve started my agent system prompt with this: "You are a customer service agent whose primary goal is to answer phone calls, to determine who the caller wants, and direct them either to the business, (I inserted my business name here) and book appointments, or to personal phone calls for (I inserted my personal name here). Always respond with a friendly and empathetic tone.

1 Like
4

You can’t do this with make alone, in a typical setup you’ll need a voice platform like vapi, retellai or elevenlabs to build your voice agent. The voice agent can be connected to a phone number (through sip trunk in something like twillio) and can be assigned tools to perform tasks (i.e. book meeting). The tools can be build in make.com and voice agent calls them through webhook.
I have an example video on my youtube channel.
Hope that helps.

5

Thanks so much, that’s awesome! Will check it out. Your help is greatly appreciated!:grinning_face:

2 Likes
6

Hi @Kim_A ,

Hi,

Based on your query, we understand that the agent you’re looking to build is an inbound AI phone agent that can:

  • Answer incoming calls
  • Screen calls between personal and business
  • Act as a booking agent
  • Function as a 24/7 emergency answering system

To build and test this type of solution, we recommend using platforms such as Vapi or Retell. These platforms are well-suited for creating voice-based AI agents and provide good documentation and tutorials. You can explore YouTube walkthroughs as well as the official documentation to understand the overall setup and flow.

At a high level, the process looks like this:

  1. Set up the voice AI platform (Vapi or Retell).
  2. Choose appropriate models for:
  • Speech-to-text (transcription)
  • Text-to-speech
  • The AI/agentic model (e.g., OpenAI models)
  1. Prompt the agent clearly for your use case (call screening, booking, emergencies).
  • You’ll need to prompt the agent to collect key booking details such as name, phone number, address, preferred date, and time.
  • For reference, here is Vapi’s prompting guide: Voice AI Prompting Guide | Vapi

Once the agent collects these details during a phone call, you can configure a custom tool with a webhook that sends the collected parameters to Make.com. In Make, you can then design the workflow to:

  • Create or update appointments in the appropriate booking system
  • Handle notifications or follow-ups as needed

After the booking is completed, Make can return a webhook response back to Vapi/Retell, allowing the agent to confirm the successful booking directly with the caller during the same phone call.

For general inquiries or emergency questions, you can set up a knowledge base within Vapi or Retell. The agent can reference this knowledge base to respond accurately, based on how your prompt is designed.

Regarding phone numbers:

  • You can use a free Vapi phone number for initial testing (note that international calls are not supported outside the USA).
  • Alternatively, you can purchase and configure a phone number via Twilio and connect it to the voice AI platform. This allows you to test the system by calling the number from your own phone.

Hope this clarifies the approach and helps you understand where to start. If you have any further questions or need clarification on any step, please feel free to reach out. We’re happy to help.

Best regards,
Msquare Automation
Platinum Partner of Make
@Msquare_Automation