I’ve developed a Zoom Virtual Agent and successfully integrated it with defined intents and a synced knowledge base (via web sync). While the setup is functional, the responses are not consistently accurate or contextually relevant as expected.
The actual ZVA is just a conversational Bot not an Agent though!!
I’d like to understand:
What steps can be taken to improve the accuracy and relevance of the answers provided by the Virtual Agent using the existing KB and intents setup?
Is there a supported way to integrate the Zoom Virtual Agent with external LLMs (e.g., OpenAI, Claude, etc.) to enhance answer quality and handle more complex or dynamic queries?
Any guidance or best practices from the community or Zoom team would be highly appreciated.
Welcome to the community and that is a great question.
There are places where you can upload content for your virtual agent to understand.
You dont need the api to do that.
I am not convinced (not Zoom Employee), that Zoom have all of this connected up yet.
So I would upload content directly to the ‘feature’ you wish to ‘teach’
I would love to be able to do a demonstration webinar and that content feed my bot. But I dont think we are at that level yet. I would have to upload it “by hand”
I would love it if Zoom contridicted me with better answers to use today.
all the best
You are stating the knowledge base pdf upload right. But that is not a permanent solution because the statements we upload in PDF will get expire right ?
All content should be processed into your ‘llm’
I don’t see how it would expire, and I’m still waiting for clarification on how to correct or update the information.
I only installed contact centre this month, so at the start of my journey
