AI / Automation / Prototype
Agent Canvas
A lightweight method for describing an AI agent before discussing frameworks or models: why it exists, what it knows, what it may do, when it must stop and how success is measured.
Interactive prototype
Define the operating model.
Mark each area as defined. The goal is not a perfect specification, but to expose missing decisions early.
Agent design starts before the model.
It is easy to begin with implementation questions: Which model? Which orchestration framework? Which tools? Which vector database? Those decisions matter, but they belong later in the process.
The Agent Canvas moves the first discussion toward the operating model. If purpose, permissions and stop conditions are ambiguous, a more capable model usually amplifies that ambiguity instead of resolving it.
A useful design test
Could a new team member understand what the agent may decide without reading the code?
If not, the operating model probably exists only implicitly in prompts, integrations and developer assumptions.
Autonomy is a spectrum.
The same use case can often be designed at different levels: observe, recommend, prepare, execute with approval or execute autonomously. The canvas makes that choice visible so teams can begin with the least autonomy needed to create value.
What I would add next
- Risk tier: Apply stronger controls to higher-impact actions.
- Data scope: Name sensitive sources and retention expectations explicitly.
- Test scenarios: Define normal, ambiguous and adversarial situations before rollout.
- Observability: Define which decisions and actions must remain explainable after execution.