Skip to main content
An AI table enriches every row of a source with columns you describe in plain language. Point it at your support conversations, sales calls, a research table, or another AI table, say what you want to know about each row, and Rippit fills those columns for existing rows and for every new row that arrives. Use a chat answer for a question you have right now. Use an AI table when you want the same judgment applied to every row, all the time: a topic on each conversation, a churn-risk score on each call, a summary of each web page a research table finds. To create one, open Tables, choose Add table, then Create AI Table. Or describe what you want in any chat, and Rippit builds it in that conversation.

​
What you can enrich

The source needs a date or time column so Rippit knows which rows are new. For a conversation, each column reads the full conversation, every message in order, not just a subject line or summary.

​
Describe the columns

Each column has a name, instructions, and an output type. Write the instructions as a task for one row: “Decide the main reason the customer contacted us. Pick unclear when the conversation doesn’t say.” Choice and number columns chart cleanly, so prefer them over text when you plan to count or compare results. A column can depend on other columns. It then runs after them and sees their results, so you can, for example, pick a topic first and then a subtopic that depends on it.

​
Choose which rows to enrich

  • Row filter. Only rows that match are added to the table. “Only conversations tagged billing,” for example.
  • History window. How many days of existing rows to include when the table starts. After that, new rows are enriched as they arrive.
  • Source fields. By default every column sees every field from the source. You can limit them.
  • Reprocessing. A conversation that keeps changing can be enriched more than once. Set the maximum runs per record and the cooldown between runs.

​
Validate, then start

  1. Save the configuration.
  2. Validate (optional). Rippit tests the table on a sample of rows and reviews each column. It shows what it found, examples from the sample, and suggested changes you can accept or decline. Validating uses AI credits and doesn’t start the table.
  3. Start the table. Rippit enriches the history window first, then keeps up with new rows. The estimated cost is shown before you start, as a starting cost and then a daily cost.
You can pause and resume a table at any time. Once it starts, the source and the set of columns are fixed, and only column instructions can change. To change the source or add a column, create a new AI table.

​
Use the results

Open the table to see:
  • Table. Every enriched row, with a link to the original conversation.
  • Canvas. Charts that build themselves once the first rows land. Click a bar to see the conversations behind it.
  • Validation. The latest check and any open suggestions.
You can also ask about the results in chat, use the table as the source for an Agent App that reports on it every week, or build another AI table on top of it.

​
Model

Every column in a table runs on the same model, in tests and once started. Pick it in the table’s processing settings, or leave the default.