Run your first discovery scan

Once a connector has synced, Discovery clusters the raw signals into candidate processes — recurring patterns of work with real evidence behind them.

Run it

Open Discovery and press Run discovery. The engine analyzes recent signals for:

• Recurring tasks — the same kind of work appearing on a rhythm
• Approval chains — request → wait → approve/reject sequences
• Duplicated work — near-identical effort in two places
• Handoffs — work moving between people or systems

Reading the results

Each candidate shows a confidence score and an evidence summary such as “214 Gmail threads · 38 Slack approvals over 30 days.” Click through to see the actual signal counts behind the claim — nothing is asserted without evidence you can inspect.

Confirm or dismiss

• Confirm promotes a candidate to the Process Graph, where you can edit steps, owners and durations.
• Ignore teaches the engine not to suggest that pattern again (you can undo this in settings).
• Duplicates can be merged — pick the primary and the evidence combines.

When to re-run

Discovery runs automatically after each significant sync, but a manual run is useful right after connecting a second source: cross-tool correlation (email + chat) is where bottlenecks become obvious.