Dataset Context Updates
Dataset context updates help Eric understand what a dataset is for, which columns matter, and how your team uses it.
Where context updates appear
Section titled “Where context updates appear”The AI context prompt can appear after saving:
- A base dataset
- A Modified Dataset
- A Joined Dataset
Requirements
Section titled “Requirements”All of the following must be true before the prompt appears:
- The company has completed AI onboarding
- The user has AI Context Updates permission
- The save action completed successfully
If onboarding is not complete, the prompt is skipped.
What the flow looks like
Section titled “What the flow looks like”Step 1: Post-save prompt
Section titled “Step 1: Post-save prompt”After a supported item is saved, Eric can ask whether you want to improve its AI context.
This is a lightweight confirmation step. If you decline, nothing is saved.
Step 2: Draft generation
Section titled “Step 2: Draft generation”If you continue, Eric generates a draft context package for that item.
The draft includes:
- A narrative summary
- Business purpose
- Key columns
- Usage notes
Step 3: Review wizard
Section titled “Step 3: Review wizard”The wizard shows the drafted context and lets you review changes. You can:
- Compare the draft against the existing context in a diff view
- Edit fields directly in a manual editor
- Confirm and save, or discard the draft
For key columns, the wizard can show a selectable list of available columns when the app already knows them.
Step 4: Save
Section titled “Step 4: Save”When you save, Eric stores:
- The final narrative context
- Structured fields such as
business_purpose,key_columns, andusage_notes
Future Eric conversations can then use that richer context when reasoning about the item.
What this is good for
Section titled “What this is good for”Use dataset context updates when:
- A dataset name is not enough to explain its purpose
- A join exists for a very specific reporting need
- A Modified Dataset encodes business logic that is easy for humans to forget
- Certain columns matter far more than the rest
Best practices
Section titled “Best practices”- Explain the business goal, not just the technical schema
- Call out exceptions, exclusions, and edge cases
- Highlight the columns people refer to most often
- Update context after major schema or meaning changes