Chat icon
By Tim Warren on July 17, 2023

How Ambit AI Manages Quality Assurance in GPT Responses

Ambit has designed the Ambit Generative AI functionality to extend the power of the existing Ambit Platform. We have architected the platform to ensure that customer conversations are ring-fenced to specific information (for example, information you supply from your website, PDF documents or other controlled content, etc).

In addition, conversations that require steps that have been vetted by a security / compliance team can continue to use a structured process flow.

As part of our release to customers, we also carry out extensive testing.

It is important to note that because of the “generative” nature of GPT and LLMs, there is always a slight risk of hallucinations. Hallucinations occur when a LLM generates content that is misleading or incorrect. 

Ambit has several tools available to further mitigate this risk, including prompting the LLM to return a pre-approved text response, when certain questions or topics are asked.

- Can bot responses provided by GPT be filtered to use for quality assurance purposes?

Yes - using the Ambit platform you can intercept responses and filter or modify them where required. We also log conversations in the Ambit Platform so you can review them and use them to improve your conversation model.

Published by Tim Warren July 17, 2023