Major customers like UnitedHealthcare, Cisco, and NYC Health+Hospitals have talked about how they were able to find hidden insights in a matter of minutes using Aible and how their experiences led to the creation of ChatAible. The customer advisory board for ChatAible includes three Fortune 100 companies.
Leveraging Aible’s augmented data engineering and analytics capabilities, ChatAible uses an AI-first approach to ask millions of questions and prioritizes the best answers.
Generative AI on its own lacks enterprise context such as exactly what drives sales, churn, etc. at your company. ChatAible leverages prompt augmentation to pass curated metadata to the Generative AI and generates insights based on the enterprise context. Aible does not send raw data to the Generative AI, and because Aible has already asked all of the possible questions of the raw data, it knows exactly the right amount of context to provide to the Generative AI given the constraints of prompt length. Aible also restricts the Generative AI to speak business language by default.
Generative AI ‘hallucinates,’ or makes up facts, over 20% of the time. People can’t possibly manually double check an AI. ChatAible automatically double-checks the responses of the Generative AI using an explainable AI that confirms the generated summary against the raw data and metadata. ChatAible can even auto generate dashboards in popular analytics applications and embed the ChatAible experience in them.
ChatAible also makes it easier for users to triple-check the system by simply tooltipping over the generated text and seeing the corresponding charts based on the underlying data.
In the amount of time, and thus for the cost, of a user manually asking a handful of questions, Aible can ask a million questions and prioritize the answers. As featured in this Google Cloud Blog, a joint Fortune 500 customer with Google Cloud evaluated 75 datasets with 100 million+ rows, prioritizing and saving patterns for 150 million+ variable combinations, for only $80 in total compute cost. At the Gartner Data & Analytics Summit, Cisco shared that they analyzed a 164 million row, 120 column dataset across 4.6 Million variable combinations (questions if you will) with just one click in under 5 minutes and for a total compute cost of $15. The cost of the Generative AI would be in addition to that.
Aible never transmits your raw data to Generative AI - only anonymized metadata. With strict limitations on the length of the questions you can send to Generative AI, the curated metadata approach allows ChatAible to efficiently provide valuable enterprise context to the generative AI. From a privacy perspective, ChatAible enforces k-anonymity so that an individual's data never leaves the enterprise. From a data residency perspective, analysis is conducted where the data already is, in the customer’s own Virtual Private Cloud.