Building Suzana: The AI Phone Agent

The idea of creating Suzana was driven by a simple yet powerful question: Can AI effectively answer phone calls and provide useful information to customers? The goal was to design an AI agent that could handle basic inquiries and provide follow-up actions through text messages.

Prototype Development:

1. Voice Recognition and Response: Suzana was equipped with advanced voice recognition technology to understand and respond to common customer queries.
2. Information Retrieval: Suzana was programmed with key information about SAMO Technologies, such as our services, operating hours, and contact details.
3. Follow-Up via Text: At the end of each call, Suzana would send a text message containing relevant links and further information based on the caller's needs.

Testing Suzana:

The initial tests were promising. Suzana managed to handle several inquiries, directing callers to appropriate resources without human intervention. For example, after answering a call about our services, Suzana would send a text message with a link to our service page and contact information for follow-up questions.

Test it yourself

I might change this a bit more in the future after publishing this blog. But you can call the number below and talk to Suzana. By calling the number below, you agree to receive a text message after the call.

The Role of Prompt Engineering in AI

A crucial aspect of making AI agents like Suzana effective is prompt engineering. Prompt engineering involves crafting precise and contextually relevant prompts that guide the AI to provide accurate and helpful responses. Domain expertise plays a significant role here; by leveraging in-depth knowledge of the specific industry, we can create prompts that better reflect the nuances and specific needs of the business. For example, understanding the typical queries a restaurant might receive allows us to tailor the AI's responses to be more relevant and efficient, ultimately enhancing the overall customer experience.

Below is a screenshot of part of the prompt I fed chatgpt API in order to create the turn base conversation. As you can see, for production use, the prompt would be more detailed and specific to the business, than the ones everyday user would use with the Chat-GPT chat app. This is where the domain expertise and lenguistic proficiency comes into play

After the call is over, I used other services' APIs such as Sendgrid and Twilio to send to myself and e-mail with the call outcome, and to the user a SMS with a link to SAMO's website

Email

This is the screenshot of the email I received after the call was over. The email contains the call outcome, the call recording, and the transcript of the call. It can use some more styling, and better formatting, but it gets the job done. This was a proof of concept for myself anyway. Simple Sendgrid API call and dynammic template.

SMS

The SMS was sent using Twilio. Simple, can also be greatly improve.

Cost

Time cost: It is time consuming to setup. I used 4 different companies Twilio, Sendgrid, ChatGpt, and Microsoft. After the initial setup the maintenance is greatly reduced and you are left with the cost of the AI turn base conversation (Chat GPT) and Twilio metered cost of voice and SMS. However, it is a fraction of the cost of a human. In 10 test calls each lasting around 5 minutes the average cost was around $0.50 to $1.00 per call.

Use Cases for AI Phone Agents

Business Offices:

In business settings, AI agents can assist with various administrative tasks such as directing calls, providing information about services, and scheduling meetings.

Example

Law Firms:

For law firms, AI agents can manage initial consultations by collecting basic information and scheduling appointments. They can also provide details about services offered and direct callers to specific departments.

Example

1. Caller: "Do you handle family law cases? "
2. Suzana: "Yes, we do. I can send you a link with more information about our family law services. Would you like to schedule a consultation?"

1. Caller: "Can I speak to someone about your IT services? "
2. Suzana: "I can connect you with our IT department. Meanwhile, I"ll text you a link with detailed information about our services."

Restaurants:

For restaurants, an AI phone agent like Suzana can significantly enhance customer service by providing information such as operating hours, address, and menu details. The AI can also handle reservations by sending a text link to an online booking system or even taking reservations directly during the call. This not only saves time for the staff but also ensures that customers receive timely and accurate information.

Example

1. Caller: "What are your operating hours?"
2. Suzana: "We are open from 8 AM to 10 PM every day. Would you like to make a reservation?"
3. Caller: "Yes."
4. Suzana: "Great! I'll text you a link to our reservation system."

Doctor Offices:

In medical practices, AI phone agents can handle appointment scheduling, provide information on office hours, and address common patient queries. This can reduce the workload for receptionists, allowing them to focus on patient care.

Example

1. Caller: "Can I schedule an appointment?"
2. Suzana: "Sure! I can help you with that. I'll text you a link where you can choose your preferred time and date."

Law Firms:

For law firms, AI agents can manage initial consultations by collecting basic information and scheduling appointments. They can also provide details about services offered and direct callers to specific departments.

Example

1. Caller: "Do you handle family law cases? "
2. Suzana: "Yes, we do. I can send you a link with more information about our family law services. Would you like to schedule a consultation?"

The prototype of Suzana at SAMO Technologies demonstrated that AI can indeed handle phone calls effectively, providing valuable information and follow-up actions to customers. As AI technology continues to evolve, its potential applications in customer service across various industries are immense. From handling reservations at restaurants to scheduling appointments at medical offices, AI phone agents like Suzana can streamline operations and enhance customer satisfaction.

Embracing AI for phone answering can free up valuable time for your staff, allowing them to focus on more complex and critical tasks. As we continue to refine and expand the capabilities of AI agents, the future of customer service looks incredibly promising. If you haven't considered integrating AI into your phone answering system, now might be the perfect time to explore this transformative technology.