After every run completes, the agent synthesises three plain-English findings (cached for 7 days). Each carries a short headline, a longer explanation, and a list of recommended actions you can approve into the Inbox.
Why three
Three is the sweet spot between “too few” (one finding can’t cover the dimensions of citation movement: prompts gained, prompts lost, technical issues) and “too many” (twelve findings is a wall of text — nothing gets done). Three forces the agent to prioritise.
What a finding looks like
Example structure:
- Headline — “Perplexity now cites you for 4 of your 10 highest-intent prompts.”
- Explanation — 2-3 sentences of context and what changed since the previous run.
- Recommended actions — concrete next steps: “Draft an article for prompt X (3 cr)”, “Install schema on /pricing (free)”, “Draft a Reddit reply on thread Z (2 cr)”.
One-click approval
The action list has an Approve all button. Clicking it creates Inbox cards for every action — articles get queued for generation, social drafts get queued for drafting, schema fixes get marked ready-to-paste. You can also approve actions individually if you don’t want to spend on all of them at once.
Cache + regeneration
Findings stay cached for 7 days. Re-visiting the Run Detail page after a week triggers regeneration on next view (if the run completed long enough ago to warrant a fresh take).