Smart Response is a first-of-its-kind generative AI solution integrated into the CENTRL platform that transforms how you handle incoming assessments. This feature will significantly reduce the time spent on assessment responses as it automatically extracts answers from your existing documents (policies, reports, etc.) and previously completed assessments.
With Smart Response, we back each answer with a rationale and referenced resources, empowering you to review and refine responses effortlessly. With unprecedented accuracy and speed, this innovative AI solution is designed to optimize your assessment completion like never before.
This article covers frequently asked Questions & Answers about our Smart Response feature for users. Please see our other help articles on Smart Response for your reference below:
This article covers the following topics:
General
How do we get started with Smart Response, and what data do we need?
- Getting started is as easy as selecting the assessment you want to answer and clicking a button to initiate Smart Response. This streamlined approach ensures a quick and effortless start without extensive data preparation or time investment.
Will the issuer of the assessment see that I used Smart Response? What will they see?
- No, when the completed assessment is submitted back to the issuer, they will not see any difference between an assessment that is answered manually versus an assessment that uses Smart Response to generate answers.
How accurate are the results? We have heard that these generative AI models are prone to hallucinations. Have you done any analysis?
- Our results are highly accurate, and we've made intentional decisions to avoid hallucinations. Hallucinations often result from writing open-ended queries or asking the model to generate large amounts of text. Instead, we ask the model a very specific question with a list of relevant passages from your sources to ensure focused and accurate responses.
- We've also integrated hallucination detection mechanisms into our system.
Are there any constraints to the types of documents I can utilize for Smart Response?
- We currently support any Word or PDF document. It is on our roadmap to extend the list of supported documents to include Excel and PowerPoint in the future.
- Additionally, we do not currently analyze content in images within documents. This is another enhancement we will be looking at in the future.
LLM Model and Training
What AI tool are you using?
- We use OpenAI, the most prominent Large Language Model (LLM). We continue to update to the latest model offered by OpenAI and testing is done before every release.
How are you training the model?
- The Large Language Model is pre-trained on tens of gigabytes of data. We leverage this existing model to generate responses to your assessments. Additionally, we employ a method known as Retrieval-Augmented Generation (RAG). RAG integrates a retrieval mechanism with the model, enhancing its ability to generate responses by retrieving relevant passages from your provided sources and using them to inform the generated text. This approach enriches the model's understanding and enhances the quality of its responses based on retrieved information.
Do you plan to add or use other open-source models?
- The AI market is rapidly changing, and we are evaluating whether to use an open-source model in the future.
Security and Privacy
Is this secure? What about data privacy?
- Yes, we use a version of OpenAI developed for enterprise users with enterprise security and data privacy. We have configured our license with zero retention, so OpenAI does not store any data from our requests. We do not use the free ChatGPT service. For further information about general data privacy, please click here: API Data Privacy.
How is my data being used? Does CENTRL take data from my environment to train?
- We do not use your data to train the LLM; we only use it to query it and generate a response.
Roadmap and Feedback
What is the roadmap for Smart Response?
- We're thrilled to further enhance Smart Response, making it even more powerful. Our immediate roadmap includes broadening its capabilities to analyze additional document types, such as Excel and PowerPoint.
- If you want to harness the power of Smart Response for all your RFPs and assessments, even those received outside of the CENTRL platform, we're thrilled to introduce Response360—a comprehensive solution designed just for this purpose. Response360 streamlines the response process, enabling responders to significantly cut down on time spent while maintaining high accuracy and consistency in their answers. This end-to-end solution automates the digitization of offline questionnaires and responses, while continually updating a central repository of your response content. To learn more about Response360, please see our full article, here or contact our Support Team.
How can I suggest a feature or improvement and provide feedback?
- In-App feedback: you can provide feedback directly in the CENTRL platform as you review the AI generated answers by giving the answer a thumbs up or down. This information is used to help improve the accuracy of responses over time.
- If you see gaps in our product or have suggestions to improve the existing features, please reach out to our Support Team. We would love to hear from you!