DataBot is an AI reporting tool that allows you to use natural language to pull reports on the data in your group using Basejump. Basejump is connected on a per-group basis, and is entirely opt-in. Group administrators will have to reach out to Action Network to enable this feature specifically, and setup may take a few days. This feature is only available to partners who are part of Boost. Click here to learn more and join Boost.
To use DataBot, you will need to be an administrator in the group with reporting permissions. Each user in the group will have a different chat history. This means you cannot see the reports another admin ran using DataBot.
To get started, click the chat icon in the upper right corner that says 'DataBot: Chat with your data'. In the pop up, you'll be able to chat with an AI agent to generate reports on your data. DataBot has access to all of your group's data and can report on it for you, even things the interface can't query. When you make a request, you'll see the agent thinking. If there is anything that is unclear or not possible, you can look here to see why that may be. Your results will be available as a chart, data table, or you can export it. You'll also be able to click Audit to check the query it created and make sure it's accurate. We strongly recommend using the thumbs up or down to let us know if the result was accurate. A real human is reviewing these to improve the quality of results over time.
There are some limitations to all AI powered tools that you should be aware of. First, while this tool can produce data based on prompts, it can't do things like analyze tone, sentiment, etc.
Tips for using DataBot
Use Action Network terms
DataBot uses the same terms our interface does. For example, say "activists" to describe people on your list, not "users" or "supporters."
DataBot uses SQL
It has access to all data in your group (and child groups, if you are in a network). See our SQL mirror docs for available tables and fields. While DataBot can use data from the mirror to produce easily digestible data, it cannot do things like analyze sentiment of submissions.
Be specific
Longer and more detailed prompts are usually better. Name your group, action, petition, and date range instead of "this," "that," or "us."
Define your terms
Instead of, "Get me a count of one month actives," say "Get me a count of activists who opened or clicked an email, or took at least one action in the last 30 days."
Sanity-check results
If numbers look off, be skeptical and use the thumbs up/down buttons -- humans review that feedback and can make adjustments for the future.
Example prompts to start off
How many activists have donated in the last year, grouped per month?
How many activists are subscribed to the email list in each state?
Show me the open rate of the top 5 activist email domains over the last three months, grouped by week.