Templates
Use-case card
| Element | Response |
|---|---|
| Exact problem | |
| User | |
| Affected people | |
| AI task pattern(s) | generation / retrieval / extraction-classification / prediction-recommendation / conversation / multimodal / agentic action |
| Interaction pattern | embedded / batch-offline / copilot / user-facing conversation / background agent |
| Inputs | |
| Expected output | |
| Output modality | text / code / image / audio / video / sensor / structured data / action |
| Source of truth | |
| Knowledge and state | base model / authorized retrieval / live tools / session memory / persistent memory / fine-tuning |
| Deployment | supplier SaaS / direct API / self-hosted or open weight / on-device or edge |
| Operating mode | batch / scheduled / event-driven / real-time |
| Work mode | copilot / bounded automation / strong automation |
| Architecture | one model or assistant / tool-assisted workflow / one business agent / orchestrated agent team |
| Exact action boundary | A0 / A1 / A2 / A3 / A4 |
| Informs, recommends, decides, or acts | |
| User-facing AI | yes / no; disclosure and human channel |
| Possible actions | |
| External effect | none / draft / communication / system write / physical or financial effect |
| Human validation | |
| Reversibility | |
| Baseline | |
| Primary KPI | |
| Severe error | |
| Frequency / volume | |
| Suppliers / systems | |
| Jurisdictions and roles | CH / EU / other; provider / deployer / controller / processor / other |
Separate scoring
- Value (1–5) and rationale:
- Difficulty/risk (1–5) and rationale:
- Risk R0–R3:
- Least complex first experiment:
To print or save as PDF: Ctrl+P (⌘P on Mac).
Source and history · GitHub