I do not think the most interesting AI shift in project management is automation. Not writing tickets faster. Not moving issues around. Not keeping the board cleaner with less human effort. The more interesting shift is reconciliation. Most issue trackers give you a declared structure for work. Parent issue. Sub-issue. Project. Milestone. Status. Label. Humans mostly operate inside that structure because the UI makes it legible. But it is also incomplete. A parent-child hierarchy is a tree. That is the clean version of the story. Real work does not stay there. The moment you add labels, dependencies, shared blockers, docs, PRs, comments, and chat threads, a single issue starts participating in multiple structures at once. It belongs to one parent. But also to a theme. A blocker cluster. A rollout sequence. A conversation thread. A code path. That is not just a tree anymore. It is a graph. I think humans underuse this because most trackers still train us to think in trees. We look at the parent issue. We look at the board. We trust the visible hierarchy. AI does not have to stop there. To AI, an issue is not mainly a card on a board. It is a packet of context. Text. Metadata. Relationships. More context from comments, docs, code, and chat. That does not automatically create truth. A lot of it is noisy. But it does create a richer inferred structure of how the work is actually moving. That is where drift shows up. Drift is the gap between the declared structure and the inferred one. The tracker says the parent issue is still coherent. The inferred structure says the work has split in two. The status says active. The broader context says the real dependency has not moved in two weeks. The hierarchy says one team owns it. The actual movement is happening somewhere else. That mismatch is drift. And I think it is one of the default failure modes of issue tracking. Not because teams are lazy. Not because the tool is bad. Because the official structure stays relatively still while the real work keeps changing shape. That is why I think the interesting use of AI here is not issue automation. It is state reconciliation. Take the declared state: - status - owner - hierarchy - labels - project Compare it against the observed state: - comments - child issue movement - dependency activity - code changes - docs - chat Then ask the real question: Where do these stop matching? That is the product I would want. Not an AI assistant that helps me maintain the tree a little faster. A system that compares the tree against the richer structure underneath it and tells me when the board has stopped being true. Maybe that is the real future of project management. Less grooming. More reconciliation. Maybe that was always the real problem. AI just makes it easier to see. March 29, 2026 — Mill Valley. Late night