Multi-Role Collisions — LLM Prompting for PMs, Vol. 2

When it comes to Product management, context swtiching is a b*tch. You’re expected to think like your user, reason like your engineer, anticipate the CEO, and still get the roadmap out on time. 🙄

Normally, you'd Slack teammates, draft decision docs, or talk it out on a whiteboard. But what if you could simulate those debates instantly, before you even open Slack?

That’s what Multi-Role Collisions let you do.

Multi-Role Collision is a technique where you prompt the LLM to adopt multiple conflicting perspectives, then resolve the tension. Think of it like forcing your internal Slack debate into a single AI prompt. Instead of asking AI to generate a decision, you make it simulate the decision-making process itself.

Here’s how to use it in real PM scenarios:

1. “Should We Build This Now or Later?”
   You’re considering a new dashboard feature. Users want it. Dev says it’s heavy. Marketing says it’s critical for conversions. You need a drink to pre-process the debate.

Prompt:
“Play out a conversation between a PM, a software engineer, a marketing lead, and a user about whether we should prioritize this new analytics dashboard. Each persona should bring their unique viewpoint and argue for or against. Conclude with the PM’s final decision based on the input.”

Why it works:
You don’t just get a “yes” or “no.” You get a reasoned, multi-sided conversation where the model surfaces friction you may not have noticed. It prepares you for the actual cross-functional meeting, or helps you skip it entirely, but I won’t advise this. I would prefer you think of this as a tool to cover your bases.

2. “Will Leadership Buy This?”
   You’ve got a bold new idea, but you’re worried your line manager will shoot it down.

Prompt:
“Play the role of a product manager pitching this feature to a skeptical product line manager. Include objections they might raise about timing, ROI, and technical risk. Then respond to each objection as the PM.”

Why it works:
Instead of generating a rosy pitch, you get a combative one that is complete with pushback. This trains your reflexes for real-life exec conversations and helps you strengthen weak points in your argument before you say a word aloud.

3. “My Designer and My Engineer Disagree”
   Your designer wants full custom animation. Your engineer wants off-the-shelf components. You’re stuck in the middle, and it doesn’t help that they’re both making strong arguments for their case when all you want to do is watch Netflix.

Prompt:
“Simulate a debate between a product designer and a front-end engineer about whether we should use custom animation for our onboarding flow. Include arguments around user delight, performance, and maintainability. End with the PM weighing in.”

Why it works:
You don’t need to mediate this fight in real life until you’ve seen how it could play out. The AI gives you a calm, reasoned version of the clash, helping you find compromise positions or even better alternatives.

‣ Why Any of This Matters
Multi-Role Collisions let you turn AI from a passive assistant into an active sparring partner. It doesn’t just do what you ask, it challenges assumptions, surfaces blind spots, and stress-tests your thinking. You get to preview the political, technical, and user backlash before it costs you time and social capital.

‣ Pro Tip: Add Emotional Tuning
Make roles more realistic by adding tone:

• A frustrated engineer

• A bullish CEO

• A cautious legal advisor

• An enthusiastic but naive user

This adds texture and uncovers edge-case reactions that matter in the real world.

—

Next in the series: Recursive Refinement Loops — how to get the AI to improve its own output until it’s good enough to ship (and how to set the right boundaries).