When you are working in a Smart Table, understanding the difference between Filters and AI Columns determines how quickly you get from raw data to actionable intelligence. Both tools work together, but they serve different purposes.
The distinction is straightforward:
Filters decide which rows you see.
AI Columns decide what information you see in those rows.
Filters: Narrow your universe
Filters act as an eligibility gate for your data. Their only job is to control which rows appear in your table by including or excluding them based on criteria you set. You are not creating new information; you are selecting a subset of existing rows.
Primary job: Select which rows are included in your view.
How it works: You apply rules to existing fields, such as including items from a list, setting a date range, or matching specific text.
The result: The number of rows in your table changes, leaving you with a more focused dataset.
AI Columns: Enrich your data
If filters build your guest list, AI Columns create detailed name tags for each guest. They enrich the rows you have selected by performing research to find and structure new information. AI Columns never remove rows; they only add new columns of data.
Primary job: Populate new fields (columns) with researched findings for each row.
How it works: You instruct the AI, either with a natural language prompt or a pre-built Blueprint, to find specific data points. The agent populates new columns with that information, complete with evidence, sources, and confidence scores.
The result: Your table has new columns filled with structured, evidence-backed insights.
Best practice: filter first, then enrich
For the fastest and most efficient results, always apply your tools in this order:
Filter tightly: Use Filters to narrow your universe to only the most relevant rows. This ensures you are not running AI research on items you do not need.
Enrich deeply: Add AI Columns to research and extract the specific data points you need for that filtered set.
This two-step process is the quickest path from broad data to specific, evidence-ready insights.
Quick comparison
Feature | Filters | AI Columns |
What it does | Chooses which rows are included. | Populates what fields are shown per row. |
Where to find it | In the Filter menu. | In the AI Column menu. |
What you provide | Selections from lists, date ranges, text operators. | A natural language goal or a pre-built Blueprint. |
What you get | A smaller table with fewer rows. | New columns populated with cited findings. |
Evidence and sources? | No, because no research is performed. | Yes, every cell includes source links and evidence. |
How to reuse it | Filters are typically set per search. | Save your setup as a Blueprint to reuse anytime. |
Frequently asked questions
Q: I added an AI Column, but my number of rows did not change. Why?
A: That is the correct behavior. AI Columns enrich rows, they do not remove them. To reduce the number of rows, use Filters.
Q: Why did I get zero results when I selected "Double Masking" and "Triple Masking" in the same filter?
A: Inside a single list filter, selecting multiple items creates an AND condition. A trial cannot be both Double and Triple masked. Create two separate Masking filters (one for "Double" and one for "Triple"), group them together, and switch the logic from AND to OR.
Q: My AI Column values updated after I changed my filters. Is that normal?
A: Yes. AI Columns run on the rows currently visible in your table. When you change the filters, the set of rows changes, so the AI re-runs the research for the new set.
Q: How do I pull specific values, like OS/PFS data, into my table?
A: Use AI Columns. Filters can narrow your search based on OS/PFS values (for example, OS > 24 months), but only AI Columns can extract and display those values in a new column.
Q: How can I reuse a set of AI Columns I created?
A: Save your column setup as a Blueprint. Give it a name, and you can apply the same research configuration to other tables instantly.
What you can do next
Create AI Columns to start enriching your table with researched data.
Use filters to narrow your search to the most relevant rows.
Create report blueprints to standardize narrative analyses from your filtered data.