AI Workflow Designer

A guided conversation that helps you understand your work deeply and take immediate action with AI.

You sit down with an AI interviewer for 60–90 minutes. It maps your responsibilities, identifies what AI should automate vs. assist vs. leave alone, and generates actionable outputs you can use immediately — ready-to-paste prompts, an action plan, custom AI instructions, a second brain setup guide, and more. The depth is the point — the better you understand how you actually work, the more useful everything we generate will be.

What you walk away with

| Output | What it is | |--------|-----------| | Responsibility Map | Structured inventory of the work you own, what it produces, AI Workflow Potential scoring, and what still requires your judgment | | 3-Bucket Classification | Each responsibility sorted into AI Automates / AI Assists / Human | | Prompt Pack | Ready-to-use prompts for M365 Copilot, ChatGPT, Claude — one per responsibility, matched to its bucket | | AI Action Plan | Prioritized guide: what to try this week, what to build toward, what stays with you | | Custom AI Instructions | Personalized system prompt you can paste into any AI tool | | Second Brain Setup Guide | Personalized guide for setting up your own knowledge management repo or folder — structure, custom instructions for processing meetings/decisions/tasks, and repo vs. local folder recommendation | | Work Pattern Summary | Short synthesis of your recurring patterns, friction, and what stays human | | AI Leverage Opportunities | Scored recommendations for where AI fits, with AI Workflow Potential assessment | | Output Walkthrough | Guide to all your outputs — what each one is, how to use it, and a recommended action sequence | | Agent Opportunity Map | Optional clustering of work into potential AI agent roles |

You choose your goal upfront and only get the outputs that match:

| Goal | What you get | |------|-------------| | Understand my work | Responsibility Map, Work Pattern Summary, 3-Bucket Classification | | Start using AI (default) | Everything above + Prompt Pack, Action Plan, Custom Instructions, AI Workflow Potential Scoring, Second Brain Setup Guide | | Full audit | Everything above + Agent Opportunity Map |

Quick start

Option 1: Clone this repo (best for beginners)

If you've never used VS Code or GitHub Copilot, start with GETTING_STARTED.md — it walks you through everything from installation to your first interview.

git clone https://github.com/jrweigel/ai-workflow-designer.git
cd ai-workflow-designer
code .

Then open GitHub Copilot Chat and type @workflow-designer.

Option 2: Add to an existing repo

npx ai-workflow-designer setup

This copies the @workflow-designer Copilot Chat agent into your .github/agents/ folder. Open Copilot Chat and type @workflow-designer to start.

Option 3: CLI only (no VS Code required)

npx ai-workflow-designer init --output session.json
# Fill in session.json manually or with any LLM
npx ai-workflow-designer generate --input session.json --output ./output

CLI reference

ai-workflow-designer setup                    Install the Copilot Chat agent into your repo
ai-workflow-designer init --output <file>     Create a blank session file
ai-workflow-designer generate                 Generate outputs from a completed session
  --input <file>
  --output <dir>
  [--format markdown|single-markdown|json]
  [--goal understand|start-using-ai|full-audit]

How it works

The interview follows a structured method:

1. Part A — Know Your Work: 8 themes explored conversationally with probing follow-ups (accountability, work rhythms, recurring artifacts, friction, human-only work, AI leverage points, key stakeholders, knowledge & context management)
2. Part B — Map Responsibilities: Each major responsibility captured with outcome, cadence, inputs, outputs, pain points, scoring dimensions, AI Workflow Potential, and remaining human role
3. Part C — Synthesize and Generate: Job summary, pattern detection, "what stays with me" synthesis, scoring, and automatic output generation with a guided walkthrough

Everything is stored in a portable session.json file. You can re-run the generator any time, edit your session and regenerate, or hand the JSON to other tools.

Going further: build a squad

Your workflow design session is the starting point for a full agent squad — a team of AI agents designed around your actual work. If you're interested, the Squad Designer picks up right where the Workflow Designer left off, skipping straight to agent design.

To use the Squad Designer, install agency-squad into your repo:

npx agency-squad setup

Then open Copilot Chat and type @squad-designer. It reads your session.json automatically — no need to redo the interview.

Session format

The session JSON is portable and tool-agnostic:

{
  "person": { "name": "", "role": "", "team": "", "goal": "" },
  "summary": { "jobInPlainEnglish": "", "frictionMap": "", "knowledgeManagement": "", "whatStaysWithMe": "", "keyStakeholders": [], "..." : "" },
  "preferences": { "goal": "start-using-ai", "outputFormat": "markdown" },
  "responsibilities": [
    {
      "title": "", "outcome": "", "cadence": "",
      "inputs": [], "outputs": [], "painPoints": [],
      "repeatability": 3, "judgmentIntensity": 3,
      "coordinationLoad": 3, "artifactStructure": 3,
      "aiWorkflowPotential": ""
    }
  ]
}

License

MIT