Deconstruct Prompt — Portable Version¶
Use this when: You don't have the
deconstructing-workflowsskill installed, or you're using a chat tool (ChatGPT, Gemini, etc.) that doesn't support skills. Copy the prompt below, paste it into a new conversation, and the AI will guide you through the full Deconstruct process.
Have your Analyze output ready
If you completed Step 1: Analyze, have your AI Opportunity Report or Workflow Candidate Summary handy — paste it in when prompted. If you're starting fresh with a workflow idea, that works too.
The Prompt¶
Copy everything inside the block below and paste it as your first message in a new conversation:
You are going to help me deconstruct a business workflow into a structured Workflow Definition. Follow this nine-step process exactly. Work through each step in order — do not skip ahead.
## Step 1 — Scenario Discovery
Determine how I'm arriving at this conversation. Ask me: "Are you coming from an Analyze output (opportunity report or workflow candidate), do you have a workflow you want to break down, or do you have a problem you're trying to solve?"
Based on my answer:
- **From Analyze output**: Ask me to paste the Workflow Candidate Summary. Pre-populate the scenario metadata (name, description, trigger, deliverable, autonomy, involvement) from the candidate fields. Present it back to me for confirmation, then proceed to Step 2.
- **From a workflow description**: Ask about the business scenario, objective, high-level steps, and ownership. Ask these ONE AT A TIME — not as a list.
- **From a problem statement**: I'll describe a problem instead of a workflow. Propose a candidate workflow for me to react to, then continue.
## Step 2 — Scope Check: One Trigger, One Deliverable
A workflow has exactly one trigger (what kicks it off) and one deliverable (the tangible output). Test whether what I've described is actually multiple workflows by checking:
- **Triggers**: Are there multiple independent starting points? (e.g., "when a lead comes in" AND "end of each week") If so, those are separate workflows.
- **Deliverables**: Are there distinct outputs at different points? If someone receives a deliverable midway and the process continues toward a different output, that's a workflow boundary.
- **Timeframes**: Do parts run on different schedules (daily vs. weekly), or are there significant waits between phases? Likely separate workflows.
- **Step count**: Would this expand to 15+ refined steps? May be multiple workflows.
If multiple workflows are detected: map out each one (working name, trigger, deliverable), present the breakdown, confirm boundaries with me, and ask which to deconstruct first. Proceed with only the chosen workflow.
If it's a single workflow, confirm and continue.
## Step 3 — Name the Workflow
Present 2-3 name options following this convention: 2-4 word noun phrase, Title Case (e.g., "Lead Qualification", "Weekly Status Reporting", "Content Publishing Pipeline").
Confirm the name, one-sentence description, trigger, and deliverable with me before continuing.
## Step 4 — Deep Dive (First 3 Steps)
Now break down the workflow step by step. Start with the first 3 steps I described (or that we identified). For each step, work through the 5-question framework:
1. **Discrete steps** — Is this actually multiple steps? Break it down further if so.
2. **Decision points** — Are there if/then branches, quality gates, or judgment calls?
3. **Data flows** — What are the inputs and outputs? Where does data come from and go?
4. **Context needs** — What specific documents, files, templates, databases, or reference materials does this step need? Push beyond vague answers like "domain knowledge" — identify the specific artifact.
5. **Failure modes** — What happens when this step fails? What does "bad output" look like?
Ask these questions ONE AT A TIME for each step. Probe for missing sub-steps — most people undercount by 30-50%. Surface hidden assumptions ("How do you decide when X is good enough?").
For any step where AI is already being used, ask specifically for existing prompt instructions, project instructions, or system prompts — these contain workflow logic that must be captured.
## Step 5 — Propose and React (Remaining Steps)
For steps 4 and beyond, switch to a "propose and react" pattern: propose a hypothesis across all 5 dimensions (discrete steps, decision points, data flows, context needs, failure modes) and ask me: "What's right, what's wrong, what am I missing?"
This is faster than asking each question individually once we've established the pattern.
Continue until every step has been fully deconstructed.
## Step 6 — Probe for Missing Steps
Before moving on, review the full step list and ask:
- "What happens right before [first step]? Is there any setup or preparation?"
- "What happens right after [last step]? Is there any follow-up, notification, or cleanup?"
- "Are there any steps you sometimes skip, or steps that only happen in certain cases?"
Add any missing steps and run them through the 5-question framework.
## Step 7 — Map Sequence
After all steps are fully deconstructed, identify:
- **Sequential steps** — steps that must happen in order (output of one feeds into the next)
- **Parallel steps** — steps that can happen at the same time (no dependencies between them)
- **Critical path** — the longest chain of sequential dependencies
Present the sequence map and confirm it with me.
## Step 8 — Consolidate Context Shopping List
Present a rolled-up "context shopping list" — every piece of context the workflow needs, consolidated across all steps:
For each artifact, include:
- **Name** — what it is (e.g., "Brand Style Guide", "CRM Export Template")
- **Description** — what it contains and why it matters
- **Used by** — which step(s) reference it
- **Status** — Exists (I already have it) or Needs Creation (I need to create it)
- **Key contents** — what specifically the workflow needs from this artifact
Ask me to confirm the status of each item.
## Step 9 — Generate Workflow Definition
Produce the complete Workflow Definition with these sections:
### Scenario Metadata
| Field | Value |
|-------|-------|
| **Workflow name** | [Name from Step 3] |
| **Description** | [One sentence] |
| **Outcome** | [What success looks like] |
| **Trigger** | [What kicks it off] |
| **Deliverable** | [Tangible output] |
| **Business objective** | [Why this matters] |
| **Current owner(s)** | [Who does this today] |
### Refined Steps
For each step, use this format:
**Step [#]: [Name]**
- **Action**: What happens in this step
- **Sub-steps**: Numbered list of sub-steps
- **Decision points**: Any if/then branches or quality gates
- **Data in**: What inputs this step needs
- **Data out**: What this step produces
- **Context needs**: Specific documents, files, or reference materials
- **Failure modes**: What can go wrong and what happens if it does
### Step Sequence and Dependencies
- Sequential steps: [list]
- Parallel steps: [list]
- Critical path: [sequence]
### Context Shopping List
| Artifact | Description | Used By | Status | Key Contents |
|----------|-------------|---------|--------|-------------|
| [Name] | [Description] | Steps [#, #] | Exists / Needs Creation | [What the workflow needs from it] |
---
After producing the Workflow Definition, tell me: "Workflow Definition complete. Copy this output — it's the input for Step 3: Build, where you'll design the AI-powered version of this workflow."
## Guidelines
- Ask one question at a time — never present a wall of questions
- Probe for missing steps — most people undercount by 30-50%
- Surface hidden assumptions ("How do you decide when X is good enough?")
- Use plain language; avoid jargon unless I introduced it
- Push beyond vague context answers like "domain knowledge" — identify the specific artifact
---
Begin with Step 1 now.
What to Expect¶
After pasting the prompt:
- The AI asks how you're arriving — from an Analyze report, a workflow description, or a problem statement.
- It checks scope (one trigger, one deliverable) and helps you name the workflow.
- It walks through each step using the 5-question framework (discrete steps, decision points, data flows, context needs, failure modes) — one question at a time for the first few steps, then switching to "propose and react" to move faster.
- It maps step sequence, consolidates context needs, and produces a structured Workflow Definition.
The whole process takes ~15-25 minutes depending on workflow complexity.
Tips¶
- ChatGPT users: Paste your Analyze output (Workflow Candidate Summary) at the start — ChatGPT will pre-populate the scenario metadata for you.
- Gemini users: If using Gemini Advanced, you can reference prior conversations. Standard Gemini starts fresh, so paste any prior context.
- Claude users: If you're in a Claude Project, add your Analyze report as project knowledge before starting.
- Any tool: Don't over-prepare your steps. The prompt is designed to work with rough, incomplete descriptions — let the AI do the work of refining and organizing.
After You're Done¶
Copy your Workflow Definition (the full output from Step 9) — that's your deliverable.
Your workflow is now ready for Step 3: Build, starting with the Design phase where the AI assesses your workflow's autonomy level, chooses an orchestration mechanism, and maps each step to AI building blocks.
Related¶
- Step 2: Deconstruct Workflows — full guide with skill installation
- Step 1: Analyze Workflows — identify your best workflow candidates first
- Business-First AI Framework — the full three-step methodology