Ask Amplitude is a conversational interface for using Amplitude. Intended primarily for Amplitude users with minimal experience using analytics tools, or with limited understanding of the data taxonomy, Ask Amplitude helps you express Amplitude-related concepts and questions in natural language.
With Ask Amplitude, you can:
Ask Amplitude uses a third-party LLM (through the OpenAI API) to understand requests and choose how to respond to questions. OpenAI decides which actions to take (for example, creating a chart versus searching) and synthesizes that information into response messages.
To power Ask Amplitude, Amplitude may send the following data to the OpenAI API:
Amplitude sends the conversation thread history between you and Ask Amplitude to the OpenAI API. However, the OpenAI API doesn't receive your customer data. OpenAI doesn't store or train on any of your Amplitude data.
To create a useful and informative chart, you first need to choose the right events and properties for your analysis. Ask Amplitude helps you do this by looking at the events and properties your organization queries most frequently. It also looks at popular saved charts, to figure out the event and property combinations your organization uses to represent different business concepts.
Sometimes, you may have different versions of the same property, like variations of User_ID
. By looking at patterns of use and existing content, Ask Amplitude can figure out which one is best to use. It'll probably never be perfect, but the more content you save in Amplitude, the smarter Ask Amplitude becomes.
Yes, there are three scenarios where Amplitude may send property values to OpenAI:
plan
property.Filter by Country = United States
as a follow up.Amplitude pays to guarantee OpenAI doesn't store or train on any data that is sent to their API from Ask Amplitude.
While both versions have similar underlying chart creation capabilities, the scope of Ask Amplitude is now more expansive, to better assist with a broader set of tasks. This version also models interactions as a conversation instead of as a single question; this is a better way to model the iterative nature of analytics.
The best way to improve your responses from Ask Amplitude is with good data quality. For example, always label your events events cleanly and clearly, and include descriptions. Use Amplitude's Data Assistant and other data governance tools to help keep your taxonomy clean.
Otherwise, declarative statements of intent that use the specific terminology tend to work best. For example:
Thanks for your feedback!
December 6th, 2024
Need help? Contact Support
Visit Amplitude.com
Have a look at the Amplitude Blog
Learn more at Amplitude Academy
© 2024 Amplitude, Inc. All rights reserved. Amplitude is a registered trademark of Amplitude, Inc.