What is the best way to verify my customer during a chat with my LLM dialogflow CX ?

[kata1000]

Suppose I have a DataFrame/CSV containing all the data required to verify my customers (such as customer ID, company name, address, email), stored at bucket let say.

How can I securely use LLM (Language Model) to extract and verify customer data in the backend(without the risk of reverse enginner prompt that user can ask about my database data? Additionally, can I use LLM to chat with my customers in the front end natually while extracting those information ?Once the backend model verifies the customer, then trigger actions or webhooks, Is this possible with the current Vertex AI Conversation and Dialogflow CX? 

[xavidop]

Hi,

To extract customer data, I would suggest doing that using intents and entities. I will recommend using LLMs for summarizing content, retrieve infromation from data stores, etc.

You can use LLMs on Dialogflow CX by using generators at any time: https://cloud.google.com/dialogflow/cx/docs/concept/generators?hl=es-419

And also you can use data stores and generators together to "veirfy" your users!

Please check this video where we explained all these features: https://www.youtube.com/watch?v=HK__EJHMSMU

Best

Xavi