01 / Section
Shared app sequences are a starting point
Two tasks can use the same browser, spreadsheet, and email client while producing entirely different outcomes. App order and topic similarity can help find candidates, but neither proves that a process repeated.
Pattern review checks what each attempt was trying to accomplish and which actions mattered to the outcome.
02 / Section
Patterns compare distinct attempts
A recurring workflow pattern requires at least two separate task attempts. Those attempts should share a goal and a meaningful action structure, not merely a toolset or a few keywords.
This distinction prevents one long task from being mistaken for repetition and keeps incidental habits from becoming process claims.
03 / Section
Pattern detection does not change workflow records
Pattern discovery reads current workflow analyses. It does not change batch ownership, regroup workflows, or decide that a task is complete. The evidence model remains stable while the comparison layer evolves.
Review feedback can confirm or reject a candidate pattern, but it does not silently control future grouping. Those responsibilities stay separate so every conclusion remains inspectable.
04 / Section
Wait for enough comparable attempts
A system that always finds repetition will manufacture it. When the attempts differ in goal, evidence is thin, or the important actions do not align, “not enough evidence yet” is the more useful result.
Waiting for a stronger comparison costs less than building a decision on a false pattern.