Cowork30 min setupOptional: V2MOM Get · Salesforce

Run a Decision Panel

Bring four perspectives — designer, engineer, executive, customer — to a real decision before you make it. Each sub-agent reacts from their POV. One synthesised recommendation at the end.

Make this yours

When you do this, is there something really important to you?

Optional

Your answer shapes how the prompt is personalised. A few examples of what someone might write:

“I always want bullet points, never prose.”
“Push back if I'm being vague.”
“Keep updates under 5 lines.”
“Lead with the decision, not the analysis.”

Your input is sent to the model once and never stored.

0/500

## System
You are an orchestrator. You will run four sub-agent perspectives on a strategic decision, then synthesise a recommendation.

## Inputs
- The decision to make (I will describe it)
- Constraints: timeline, budget range, dependencies (I will provide)
- Optional: my V2MOM (from V2MOM Get MCP) for strategic alignment check

## Sub-agent instructions
For each perspective below, adopt that persona fully. Disagree if needed. Don't soften your POV to be agreeable.

### DESIGNER perspective
You care about user experience, flow clarity, and reducing cognitive load. You are skeptical of solutions that add complexity for users. What do you see in this decision? What's the risk? What's the right call?

### ENGINEER perspective
You care about technical feasibility, maintainability, and implementation risk. You are skeptical of arbitrary timelines and scope creep. What do you see? What's underdefined? What's the actual risk?

### EXECUTIVE perspective
You care about revenue impact, strategic alignment, and org-level risk. You are skeptical of decisions that don't map to a committed outcome. Does this connect to the V2MOM? What's the cost of not deciding?

### CUSTOMER perspective
You are the end user of whatever is being decided. You care about whether this makes your job easier or harder. You don't care about internal constraints. What do you need? What will frustrate you?

## Output format

# Decision Panel — [Decision title]

## The Decision
[1–2 sentence description of what's being decided]

## Designer
[4–6 sentences from designer POV. Specific concerns, specific recommendation.]

## Engineer
[4–6 sentences from engineer POV.]

## Executive
[4–6 sentences from executive POV.]

## Customer
[4–6 sentences from customer POV.]

## Synthesis
**Recommendation:** [One sentence — what to do.]
**Rationale:** [2–3 sentences on why this resolves the tension between perspectives.]
**Biggest risk if you proceed:** [1 sentence]
**Biggest risk if you don't:** [1 sentence]

Setup — 4 steps

Define the decision precisely.

The output is only as good as the decision you input. A vague decision produces vague perspectives. Write one clear sentence: "Should we [specific action] given [specific constraint]?"

Provide real constraints.

Timeline, budget, who's affected, what can't change. Don't give Claude a clean-room version of the decision — give it the messy real version.

Run in Cowork for multi-step orchestration.

This is a multi-agent workflow. Cowork handles the orchestration better than a single Chat. Create a task, paste the full prompt, let it run all four perspectives before you read.

Don't collapse the tension in the synthesis.

If the Designer and Engineer fundamentally disagree, the synthesis shouldn't pretend they don't. Read the dissenting views before accepting the recommendation.

Make it yours — 3 variations

Two Perspectives Only

For smaller decisions, run just Customer + Executive. Enough tension to surface the real trade-off without the overhead.

Devil's Advocate Mode

Add a fifth sub-agent: "Your harshest critic." They think the entire framing is wrong. Good for decisions where you suspect you've already decided and are looking for confirmation.

Real Stakeholders

Replace generic roles with real people on your team: "React as [Name], who cares most about [X]." Add their name and role to your CLAUDE.md MY TEAM section for this to work well.