Comment on page
Everything you need to get value from Patterns!
In your first login you'll be prompted to explore a demo Analyst - do that first. This guide assumes you're ready to create an Analyst for use on your own data.
Click on the Analyst name on the top of the screen and select Create New Analyst.
You'll be prompted to connect a database when creating an Analyst. We currently support Postgres, BigQuery, and Snowflake and plan to support every major database in the future.
After having connected your database, you'll be prompted to select Tables to add to your Analysts context as in the below screen. You can select as many as you like, but for optimal performance select only the ones you need.
Now the fun part! Getting answers from data in your database is as simple as asking a question. In the Question interface you can click on tables and documents to explore data that your Analyst has access to and is able to provide answers against.
While it's not required, you should try to ask questions as detailed as possible to get the most reliable results such as
from the orders table what's the total revenue on a monthly basis since 2022.If you didn't now what table to use, you could still ask
hows revenue trendingand you will likely receive the same result.
Beyond asking straightforward questions, you can also:
- Exploring tables
tell me about the orders table
- Brainstorming analyses
what kind of analysis can I do on the orders table
select sum revenue, sum customers from orders join customers
join customer and orders and sum revenue and customers
Just like doing analytics without AI, there will be errors, data will look off, the database you’re executing on is tied up… Atoma is aware of all of this and is designed to self-correct. However, sometimes you need to nudge it to do the right thing. You can:
- Explain results - atoma doesn’t see the data output or charts, so if something is off, you can tell it so it can correct itself
data looks off, can you try again
for some reason all the data in the chart appearing at the top
- Advise the query
instead of using revenue, use gpv_revenue for calculating processed transactions
There are likely many other modes of interaction I have yet explored!
When you ask a question in the chat interface your Analyst will generate SQL and a chart if applicable. You can optionally Edit this SQL and Save so that you're able to add it to a Dashboard and have it part of an automated report in the future.
After saving an analysis, it will appear in the sidebar. Click on the Analysis to view an AI generated explanation of the query and in the future a description of the results!
You can share a link to this analysis anywhere, export it to PDF, and in the future sync it with a Google Slide or Powerpoint.