Leila FieldingAI Agent Design Canvas
Design the job before designing the agent.
SHARE THIS TOOL
Use this when: writing an initial brief for an agent's job and boundaries—not detailing a full architecture or implementation.
WHAT YOU'LL ACHIEVE
Create a clear description of an AI agent's purpose, capabilities, boundaries and responsibilities before beginning development. An effective AI agent starts with a defined business outcome, not a list of exciting technologies.
THE TOOL: AI AGENT DESIGN CANVAS
Build your agent brief
Example text is guidance only. Enter your own answers to include them in the copied design.
HOW TO USE THIS TOOL
Define the job, then the limits.
Start with one clearly defined problem. Describe what the agent should achieve, the information it requires and the tools it may use.
Then define its limits. Identify actions that need approval, situations that should be escalated and the evidence required to judge whether the agent is working correctly.
WORKED EXAMPLE
Helping employees find policy information
An HR team wants an agent to help employees find relevant policy information.
The agent can search approved documents and provide answers with references. It cannot alter employee records or make employment decisions. Unclear or sensitive enquiries are directed to HR.
WHAT TO CHECK
Permissions have to do the work.
An agent should have the minimum access needed for its purpose. Its capabilities must be enforced through actual system permissions and controls, not just written instructions.