Insights

Conclusions usually die in a chat log or a slide. Insights keep them in the workspace: one claim — "firms of 250+ people are 55% of the modeled market" — pinned to the chart or spreadsheet that demonstrates it, with its figure and its caveat. The claim and its evidence travel together.

Recording and finding insights

Every chart, dataset and spreadsheet page has an Insights button that opens the drawer: read the claims recorded against that evidence, or add one. The Insights page indexes every claim in the workspace, filterable by status.

The lifecycle

Each insight is open, validated or rejected:

• Validate — you checked the claim against the current data and it holds. The evidence is re-pinned.
• Reject — the claim is wrong or no longer true. It stays visible, dimmed, as a record of what was considered.
• Reopen — put a decided claim back on the table.

Staleness

When the evidence changes — a dataset gets a new version, a transformation re-runs, a chart's query is edited, a model gets a new version — the insight is flagged out of date. That does not mean wrong; it means unverified against what is there now. Re-check and mark it still holds, or reject it.

Faster with an agent — agents record insights as they work: ask for "analyze this model and record your conclusions as insights", and each finding lands with its evidence. Your job becomes reviewing and validating them. The judgement stays yours.