Do I need to train on Private Data?
Built-in support for private data sources
Requires engineering-time and development to use APIs to train
Do I need AI to handle Personal Identifiable Information (PII) Data
Requires Engineering time and development to use APIs to train
Are hallucinations [retrieving answers from internet corpus rather than customers' data] a concern
Retrieval Augmented Generation, ensures responses are either from customer data or no answer is given
The generic chat model is known for hallucinations. Hense cannot be trusted on private data.
Do I need Human-In-Loop control over AI
Edit AI responses that are used in future interactions
No human-in-loop training
Do I need to keep AI Up-to-date with the latest information?
Automated refreshes keeps AI up-to-date with the latest product and operations information
Requires engineering-time to setup automation
All of this is early will the vendor listen and work with us to improve the product per our needs
Twig pushes weekly updates the product 80% of refinements are from our user feedback
OpenAI has 21M users. Getting your feedback prioritized on their product roadmap is a challenge.
Twig assigns a customer success person