← All solutions
Product Managers
Decision Making
AI Decision-Making Help for Product Managers
See how Limenbell helps product managers think through a real decision by pulling in cross-domain framing the standard AI answer misses.
🔍 The question
A real question from product managers
I'm a product manager trying to decide between two competing feature ideas and I keep going back and forth. Can you help me think this through?
🔹 The standard AI answer
What a typical AI tool says
The fact that you're going back and forth usually means one of three things: the features aren't actually comparable, the decision is being made without clear criteria, or there's a hidden assumption that needs unpacking.
I can help you think this through. Rather than telling you which feature to build, let me ask you a few questions to structure the decision. Pick the ones that resonate:
**1. What are the two features, in one sentence each?**
**2. If you could only measure one metric for the next 6 months, what would it be?**
(e.g., activation, retention, revenue, referrals, something else)
**3. Who is the target user for each feature?**
Are they the same person, or are you choosing between two different user segments?
**4. What does "success" look like for each feature?**
Can you put a rough number on it? (e.g., "10% of users try it," "reduces churn by 2%," "unlocks enterprise deals")
**5. What happens if you build neither?**
Sometimes this reveals which problem is actually urgent.
**6. Is one of these features a bet on learning something, and the other a bet on executing something you already know works?**
---
A few prompts that often break the tie:
- **Which feature would you still believe in if it had zero impact for 3 months?**
- **Which one would you be embarrassed to explain to a customer if the other one shipped first?**
- **Which one aligns with what your company says it's trying to be in 2 years, not just this quarter?**
Start with whatever feels most alive, and we'll go from there.
✨ Cross-domain ideas from Limenbell
post-scarcity economics
In post-scarcity economics, the binding constraint shifts from resources to attention and meaning—exactly your dilemma when budget is unlimited.
Decide not by ROI but by **which feature creates the richest feedback loop of revealed preference and self-expression**. Build both as simultaneous "sandboxes" where users vote with real participation, not hypothetical surveys. The one that sparks emergent communities, remixing, or identity signaling wins. If tech is infinite, make each feature fork into the other on a single user journey—so choosing is delayed until usage data chooses for you. Scarcity of time and identity remains your real budget, so optimize for lasting narrative pull, not feature count.
Appalachian folk music
App folk music has no single “best” version—the choice depends on the room (dance, dirge, or front porch), just like features depend on context.
Forget picking the winner in your head. Pair the two features like a fiddle and banjo: one carries melody (core user need), the other drives rhythm (engagement or retention). Play them against your hardest user problem. Install one for a sprint, watch if people start tapping their feet — signs of daily repetition, not just likes. If a feature needs a “big stage” to feel valuable, it’s a ballad; skip it for now. Ship the one that survives a kitchen table session.
darts
Feature decisions are like aiming at a dartboard’s treble 20 — precision matters, but reversed rules would make the bullseye the worst place to throw.
If your usual resources or rules are fully inverted, run the two ideas like a “worst throw wins” darts match. Instead of prioritizing features that score with growth or revenue, ask which one deliberately targets the lowest-value, hardest-to-monetize, least-scalable outcome. Whichever idea would fail more efficiently under a normal product lens is the one to reject now — because in an inverted system, it’s the equivalent of hitting the bullseye. After flipping back, pick the other feature. This forces you to stop oscillating: you’re not choosing the “best” feature, you’re eliminating the one that only works when logic is backwards.
Try your own question →