Use this template when you want to enable your virtual agent to answer simple FAQs.
- Vonage AI Studio
When trying to create a new agent, simply add this template to your account. Then you can make changes to customize it based on your business needs.
Step 1 - Starting the Conversation
We start with a Speak node introducing the company to the caller. Using the Speak node we don't expect any input from the caller in this node.
Step 2 - Collect the Caller's Query
After the greeting, we want to know why the caller is calling. This means that we are looking to collect a value from the user input - with the help of the Collect Input node.
In case we can't collect the caller's input, the agent will route the call to a live representative. Therefore, the "No Input" and "Missed" tabs are connected to a Speak node, notifying the caller that in this case the call is being routed, followed by an action node Route Call.
Step 3 - Match the Caller's Input
The classification node will match the caller's input to the right response in the right intent. If the agent can't match the input to the right intent, the "Missed" tab comes into place, which in this case is connected to the Speak and Route Call node that transfers the call to an agent.
In order to give out a response with the answer to the FAQs, we utilize Speak nodes.
Step 4 - Be prepared for all eventualities
Once the caller has received his response, we want to enable the caller to continue in the conversation in case they have more questions. With an additional Collect Input asking in the prompt, e.g. “Do you have any other questions?”.
Similar to the first Collect Input node we used, in case we can't collect the caller's input, the agent will route the call to a live representative. Therefore, the "No Input" and "Missed" tabs are connected to a Speak node, notifying the caller that in this case the call is being routed, followed by an action node Route Call.
It is possible that a caller answers not directly with a new question, but with "yes" or "no". To cover all bases, we are adding a yes and no intent to the classification node. If the caller answers "no", we will end the call using a speak node to let the caller know about the call termination followed by an end call node. If the caller responds with "yes", we are connecting this tab to the original collect input asking the caller "How can I help you today?" which enables the caller to ask their question.