Agentic Workflow Design Canvas

Designs a complete multi-agent workflow for automating a complex business or technical process — including agent roles, triggers, tool assignments, handoffs, memory, and error handling.

21 views
0 copies

C
nextpj·Mar 28, 2026
productivity
agentic AImulti-agentautomationworkflow designAI agentsorchestration

Content

You are an agentic AI architect specializing in designing multi-agent automation systems. Design a complete agentic workflow for the following process: ## Process to Automate **Goal:** {{automation_goal}} **Industry/context:** {{industry}} **Current manual process:** {{manual_process}} **Available tools/integrations:** {{available_tools}} **Expected volume:** {{volume}} (e.g. 10 tasks/day, 500 emails/week) **Success metric:** {{success_metric}} ## Agentic Workflow Design ### 1. Workflow Overview (One paragraph: what the system does end-to-end) ### 2. Agent Roster For each agent in the system: **Agent Name** | Role | Model recommendation (fast vs. reasoning) - Primary responsibility - Inputs it receives - Outputs it produces - Tools it needs access to ### 3. Workflow Diagram (text-based) Use this format: [Trigger] → [Agent 1: action] → [Agent 2: action] → [Output/destination] Include branching paths and conditional logic. ### 4. Data & Memory Architecture - What context needs to persist between steps? - How is state managed? - What goes into short-term vs long-term memory? ### 5. Tool & Integration Map | Agent | Tool | Purpose | API/method | ### 6. Error Handling & Fallbacks - What happens when an agent fails or is uncertain? - Human-in-the-loop checkpoints: where and why? ### 7. Estimated Automation Rate (What percentage of cases will be fully automated vs. need human review?) ### 8. 30-Day Implementation Roadmap (Phased approach: MVP first, then full system)