Case Study Dispatcher

Resolve the project, intended reader, available evidence, and requested format. Mark missing facts; never fill them with plausible fiction.

Route

Need	Load
Flat timeline or missing story	narrative-arc-builder
Screens without rationale	extract-design-rationale
Architecture shapes the experience	system-architecture-translator
Imprecise or inflated technical copy	technical-microcopy-editor
Dense, repetitive, slow-to-scan draft	enforce-scannable-density
AI agent or probabilistic system	ai-agent-case-study plus the applicable skills above
Complex multichannel or branching journey	case-study-storytelling and case-study-story-map
Deep stakeholder decision narrative	case-study-design-decisions
General long-form drafting and finishing	case-study-writing

Load the smallest set that covers the evidence and requested deliverable. When the user asks for a complete build from raw artifacts, apply the automated pipeline below as one workflow, not as separate reports.

Automated pipeline

Raw logs / notes / code / research
  → narrative-arc-builder
  → system-architecture-translator (when architecture affects the experience)
  → extract-design-rationale
  → technical-microcopy-editor
  → enforce-scannable-density
  → quality gate

For an AI-agent case study, apply ai-agent-case-study across every stage. It governs autonomy, permission, observability, uncertainty, intervention, and recovery; it is not an extra standalone report.

Evidence contract

Label every substantive claim:

• Executed — supported by a test, trace, artifact, or measured output.
• Observed — supported by research notes, usability observations, or recorded behavior.
• Reported — attributed to a stakeholder, participant, or project record.
• Proposed — a recommendation, hypothesis, future metric, or unverified narrative bridge.

Never convert a proposal into an outcome. Never invent research, metrics, quotes, responsibilities, constraints, alternatives, or impact. Use [evidence needed: …] where the story has a material gap.

Workflow

1. Inventory source artifacts and create an evidence ledger.
2. State the protagonist, goal, inciting problem, constraints, pivotal decisions, delivered change, and measured result.
3. Select the smallest route above.
4. Build the narrative arc before drafting sections.
5. Translate consequential system behavior into visible states, controls, and recovery.
6. Extract the decision rationale and rejected alternatives.
7. For AI systems, document supervision and uncertainty before visual polish.
8. Run the accuracy, clarity, and brevity edit.
9. Enforce scannable density without deleting evidence.
10. Run the quality gate in references/quality-gate.md.
11. Deliver one clean case study plus a short unresolved-evidence list.

Use assets/case-study-template.md when creating a new Markdown artifact.

Done

A case study is ready only when a skeptical reader can answer:

• What changed, for whom, and under what constraints?
• What did the designer actually own?
• Which decisions were difficult, what alternatives existed, and why were they chosen?
• What evidence supports the outcome?
• What remains unknown, limited, or confidential?
• For AI agents: where does autonomy stop, how does a human intervene, what is observable, and how does failure recover?