In traditional agile development, issues often have many workflow statuses, and each status should represent a real handoff point or waiting queue. But as teams grow to hundreds or more, Kanban boards become bloated, and the number of issue statuses can even reach over 20. Configuring workflows and permissions in Jira becomes a total nightmare, creating all kinds of technical debt. Some companies even hire dedicated Jira administrators—I have serious PTSD from this. But now things are changing. The new statuses look like this: Backlog → Refined → AI Generating → AI Output Review → Human Augmentation → AI Test Gen → Hallucination Check → QA → UAT → Released → Done The traditional "In Progress" has been broken down into the cycle of AI Generating → Human Review → Human Augmentation, and this cycle might iterate multiple times. The role of humans has shifted from "executor" to "reviewer + curator," and the bottleneck in the process has moved from "writing code" to "validating the correctness of AI outputs." The bottleneck lies in how the product defines issues and the time spent on final QA.