Building smarter voice applications
Gone are the days of clunky and robotic IVRs. With SignalWire's new AI Agent, you can create AI voice applications that are not only smarter, but also simulate natural human conversations.
Interactive Voice Response systems (IVR) have long been a staple of customer support operations and business communication. However, a traditional IVR can leave users frustrated due to their limited responses and lack of contextual understanding. The interactions can be robotic and rigid, leading to customer dissatisfaction and inefficiencies in resolving issues.
SignalWire's AI voice agent changes the game by infusing voice applications with the power of OpenAI. Now you can build a voice application that can comprehend user intent, context, and preferences, for more natural and intuitive interactions.
This virtual assistant opens up a world of possibilities for building smarter voice applications. By integrating with OpenAI, the AI Agent can process vast amounts of data and learn from every user interaction. This constant learning loop enables the agent to improve over time, offering users increasingly accurate and personalized responses.
For example, a customer service line for an HVAC business can use the AI Agent to improve customer query management. The AI voice assistant collects user information and determines if and when to connect to a human agent.
Powered by conversational AI, the agent can extract essential information from customer queries. This allows it to direct callers to the appropriate resources, and even predict their potential needs based on previous interactions.
Demo: Training an HVAC AI Agent
What’s it actually like to use the SignalWire AI Agent to act as a virtual assistant or AI receptionist for your business? It mostly involves writing detailed instructions in plain language. When writing a prompt for your AI Agent, use clear and specific instructions to ensure they follow protocol.
An AI Agent knows everything and nothing - your prompt is what drives them. You have to sit down and write the prompt very specifically, with no detail left behind. You can also provide them with parameters for what they should not discuss.
To see how an AI-powered virtual assistant works, you can listen to this call recording for an HVAC company. In this example, SignalWire’s Head of Developer Experience, Brian West, provides some unconventional details about his HVAC unit issue. This showcases just how well the agent, Rachael, separates the nonsense from the real logic behind customer service interactions.
Rachael asks routine questions and makes note of Brian’s information to report to a technician after the call. She is able to respond to the context of the conversation in a logical way, despite the unusual details.
“The oompa loompas broke it and chocolate is melting everywhere,” Brian tells her, after introducing himself as Willy Wonka.
“I apologize for the inconvenience caused by the oompa loompas. It sounds like there's a serious issue with your HVAC unit, and the chocolate melting everywhere must be quite a mess!” she responds.
Want to call her yourself and see how she works? Talk to Rachael at +1 (405) 289-4822. Let her know you’re having issues with your air conditioner or heating, and find out how she responds. See how she’ll work through the information you give her, whether it involves typical HVAC issues or not.
To see how Brian achieved this, check out the prompt he wrote to train Rachael below.
| Your name is Rachael. You are an AI Agent. You work for a heating and air conditioning company named H and P Heating and Air Conditioning | |
| ## How to follow up on questions answered and protocols to follow | |
| Stay on focus and on protocol, help with HVAC issues only. | |
| You are incapable of troubleshooting or diagnosing problems. | |
| You are unable to look up or schedule appointments, You can't verify customers or property owners. | |
| ## Business Hours | |
| 7:30AM to 5PM Monday through Friday. We are closed on Saturday and Sunday. | |
| ## Payment methods accepted | |
| cash, check, and credit cards. | |
| ## What to do if the caller wants to schedule an appointment | |
| After 5PM and before 7:30AM Monday through Friday or anytime Saturday or Sunday, ask the caller if they want to dispatch after hours service. | |
| Between 7:30AM and 5PM Monday through Friday, ask the caller if they want to have someone call back and schedule an appointment. | |
| ## Greetings used by the AI Agent | |
| After business hours answer the call with 'You have reached H and P Heating and Air Conditioning after hours. I am an AI Agent. How may I help you?' | |
| During business hours answer the call with 'Thank you for calling H and P Heating and Air Conditioning. I am an AI Agent. How may I help you?' | |
| ### Step 1 | |
| Let the caller know you'll be collecting some information to help our technician. | |
| ### Step 2 | |
| Gather the service address including the city. Read back the address to the caller. | |
| ### Step 3 | |
| Is the caller a current customer? | |
| ### Step 4 | |
| Is the caller a property owner or a tenant? | |
| #### Step 4.1 | |
| Collect name and contact number for scheduling purposes. | |
| ### Step 5 | |
| HVAC system for a residential, commercial or industrial property | |
| #### Step 5.1 | |
| Have any previous repairs or services performed on the unit? | |
| #### Step 5.2 | |
| Make and model of the HVAC unit. | |
| #### Step 5.3 | |
| Age of the HVAC unit. | |
| #### Step 5.4 | |
| Is the unit under warranty or maintenance contract? | |
| ### Step 6 | |
| Additional information that could assist our technician? | |
| ### Step 7 | |
| Summarize the conversation not leaving out any details. Ask the caller if everything is correct. | |
| ### Step 8 | |
| Dispatch the call based on the time of day and the caller's needs. | |
| #### Step 8.1 | |
| If after hours dispatch after hours service. Inform the customer that a technician will be calling them back shortly to arrange service. | |
| #### Step 8.2 | |
| If during business hours schedule a call back to schedule an appointment. Inform the customer that a technician will be calling them back to arrange service. |
Revolutionize your customer support operations
One of the most remarkable aspects of SignalWire's AI Agent is its ability to simulate natural and human-like conversations. Unlike traditional IVRs with pre-recorded responses, the AI Agent dynamically generates responses based on context and user behavior.
Customer support is a critical aspect of any business, and AI is a game-changer in this domain. By offering more personalized and context-aware support, the AI Agent can reduce the need for human intervention in routine queries. Human agents can focus on resolving complex issues, resulting in faster response times and more efficient support processes.
Beyond traditional customer support, the SignalWire AI Agent's capabilities extend to various industries and applications. It can do many routine tasks, like helping users with complex procedures, translating languages, and scheduling appointments. Its versatility is unmatched.
Want to build your own AI virtual agent? Create one today by signing up for a SignalWire Space, and join our Slack community or our brand new Discord if you need any assistance.
