{
  "uid": "cogitave.learn.operate-from-day-1.flow-based-ways-of-working",
  "kind": "moduleUnit",
  "href": "/modules/operate-from-day-1/flow-based-ways-of-working/",
  "title": "Flow-based ways of working",
  "summary": "",
  "type": null,
  "products": [],
  "roles": [],
  "levels": [],
  "subjects": [],
  "headings": [
    "The choice: flow, plus betting for the big rocks",
    "Why not Scrum",
    "The WIP limit protects human attention, not agent throughput",
    "Feedback loops, and the deviation rule"
  ],
  "source": "The **outer loop** - how work is planned, prioritized, and flowed as a team - is\nthe directive choice in\n[ADR-0025](../../../../standards/docs/decisions/0025-flow-based-ways-of-working.md),\noperationalized in\n[ways-of-working](../../../../standards/docs/standards/ways-of-working.md).\n\n> [!NOTE]\n> Day-0, honest: no squad is running this method yet. This is the spec for the\n> method the internal team - humans and agents - will run, not a report of one\n> already running.\n\n## The choice: flow, plus betting for the big rocks\n\nThe day-to-day method is **WIP-limited continuous flow** (Kanban-style: visualize\nthe workflow, limit WIP, manage flow, make policies explicit, run feedback loops,\nimprove collaboratively) - pull-based, no fixed sprints. **Larger bets** are\nshaped and bet **Shape-Up style**: an **appetite** (the time an idea is worth)\nreplaces the estimate, inside the quarterly OKR frame. The\n[request-lifecycle](../../../../agents/lifecycle/LIFECYCLE.md) Request - the same\nseven-stage Request from the previous module - **is** the work item that flows;\nthis decision does not re-sequence those stages, it defines how items are\nprioritized, WIP-limited, and pulled through them. Cogitave does **not** run\nvanilla Scrum.\n\n## Why not Scrum\n\nFour Day-0 facts decide it. **Small team plus agents as team members** - Scrum's\nceremonies exist to manage a coordination problem this team does not have.\n**Trunk-based continuous delivery** already ships continuously behind flags; a\nsprint boundary would batch work that is otherwise done. **Velocity is a poison\nmetric** - trivially gamed, meaningless as a capacity number once agents are team\nmembers - so flow metrics (cycle time, throughput, WIP, aging) replace it, and\nnever as a stack-rank. **Cert-grade evidence must be a by-product**, not a\nseparate ceremony - PR review, a signed merge, and a blameless post-mortem already\n*are* the evidence an auditor reads.\n\n## The WIP limit protects human attention, not agent throughput\n\nWIP limits apply to the **human-attention budget** - items awaiting review,\nsign-off, or the exception gate - not to raw agent throughput, so the one scarce\nresource (human review) never becomes a silently-growing queue. Agents pull,\nexecute, and finish inside their least-privilege grant at the highest autonomy\nlevel the rails permit, emitting WORM evidence as they go; the **human gate stays\nan exception handler**, not a routine checkpoint on every item - the same\ndoctrine [`AGENTS.md`](../../../../../AGENTS.md) states in rules 6 and 7.\n\n## Feedback loops, and the deviation rule\n\nThe method carries its own loops: code review per Request, a design/architecture\nreview at the lifecycle's document stage, a demo on bet completion, a\nretrospective at the cool-down beat, and a blameless post-mortem per incident -\neach closes into a tracked, owned action, never left as an un-actioned signal.\nFlow is the default; a squad **may** adopt sprints only under a written, specific\nreason - a fixed external delivery date, or a new team that cannot yet self-organize\nflow - re-evaluated each quarter. Absent a written reason, the answer is flow.\n",
  "partOf": "cogitave.learn.operate-from-day-1",
  "durationInMinutes": 7,
  "quiz": null
}