BIG UPDATE: AI Agent Now Calls And Book Appointments - OpenAI Realtime API
Science & Technology
Introduction
A few weeks ago, I launched my real-time API AI agent designed to call various companies and gather information, particularly to assist with booking appointments. Recently, I released an exciting update that I will elaborate on and demonstrate through some call examples.
During today’s demonstration, I tested my AI agent by trying to book a dental appointment. Below is a quick recap of a 30-second sample from one of the calls made to Lumia Dental.
Call Transcript:
Agent: Thank you for calling Lumia Dental, this is Ta, how may I help you?
Caller: Hi there, is this Lumia Dental?
Agent: Yes, it is.
Caller: Cool, I’d love to know if there are any appointments available for Friday.
Agent: Would this be for a cleaning and a checkup, or did you have a specific concern?
Caller: No, it’s just for a cleaning and a checkup. Fridays work best for me.
Agent: Yeah, this coming Friday, we have a 9:45 or a 12:30.
Caller: Oh, that’s great. I’ll call back to confirm.
Agent: Yeah, thank you so much for your time. Take care.
Caller: You’re welcome.
As you can see, the AI successfully captured valuable data, including the company name and appointment times. Let's dive deeper into the code and explore additional conversations.
Recent Changes and Improvements
The latest updates to the OpenAI Realtime API bring exciting features, including the ability to create speech-to-speech experiences using five new voices. If you compare these new voices to those from previous releases, you’ll notice they sound much more natural and expressive—something I specifically requested.
Another major update is the reduction in costs associated with using the service. By implementing prompt caching, text inputs are now discounted by 50%, while inputs can receive discounts of up to 80%. Such changes significantly enhance the platform's viability.
While I won’t go through every new voice in this video, I can confirm they are of high quality. There's no requirement for you to modify your existing setup to take advantage of prompt caching; it operates autonomously.
Setup and Data Capture
Here’s a quick overview of setups I have in place. In my configuration, I enter the name of the company, my contact number, and there’s a structured system message that guides the AI’s interactions. For instance, I've programmed the agent to introduce itself, ask for the appointment, and manage the response based on the company's answers.
This time, when reaching out to companies, I captured various significant details, such as the availability of appointments. Out of the eight calls made, six were successful.
Performance Review and Conversations
Let’s listen to a call made to Studio Smiles, which was fully booked.
Call to Studio Smiles
Agent: Good morning, Studio Smiles Dental Office. Orch speaking.
Caller: Hey, I was just wondering if you have any appointments available for Friday.
Agent: Let me see. At the moment, Friday is fully booked.
Caller: No worries at all. Thank you for checking! I'll have a look at my schedule and maybe I’ll call back to book something else. Thanks again!
Agent: You’re very welcome!
While the conversation was polite, we gathered the necessary information about the unavailability of appointments.
Call to Expert Dental
Here is my favorite call of the day, which was made to Expert Dental.
Agent: Your call may be monitored or recorded for quality assurance purposes.
Caller: Hi there, is this Expert Dental?
Agent: Yes, speaking. How may I help you?
Caller: I was wondering if you have any appointments available for Friday.
Agent: Are you a current patient?
Caller: I'm actually a new patient. Do you have anything available?
Agent: We have two locations. Which one do you prefer?
Caller: Whichever has an appointment open!
Agent: What kind of appointment are you looking for?
Caller: Just a regular exam and cleaning.
Agent: Okay! Do you want to come in on Friday the 8th?
Caller: Yes, Friday the 8th.
Agent: I can do a 9:00 a.m., a 10:00 a.m., or a 2:00 p.m.
Caller: Those times work! I will call back to confirm. Thank you so much!
Agent: No problem! Have a great day!
This interaction demonstrated the AI’s ability to improvise, asking follow-up questions that were not defined in the initial guidelines.
Ethical Considerations
As I reflect on these interactions, I ponder the ethical implications of utilizing AI in this manner. Is it acceptable for AI to conduct business transactions without human oversight? As it stands, AI agents efficiently manage appointment bookings, and I observe a trend toward increasing acceptance of such technologies in business operations.
Conclusion
Overall, I am thrilled with how well the AI agent performs. The ability to capture structured data from conversations using Whisper technology is robust. The improvements in voice quality, alongside the cost reductions, make these systems incredibly viable moving forward.
Stay tuned for future updates and experiments involving different AI applications in real-world scenarios.
Keywords
AI Agent, OpenAI, Realtime API, Dental Appointment, Speech Processing, Cost Reduction, Prompt Caching, Ethical Considerations.
FAQ
Q: What is the purpose of the AI agent?
A: The AI agent is designed to call various companies to gather information, specifically to book appointments.
Q: What changes were made in the recent update?
A: The major updates include new expressive voices, cost reductions through prompt caching, and enhanced functionality for speech-to-speech experiences.
Q: How many calls were successful?
A: Out of eight calls made, six were successful in gathering the desired information.
Q: Is it ethical to use AI in customer service interactions?
A: The ethical implications are still being debated, but the trend shows increasing acceptance of AI in business transactions.