Skip to main content

Creative Problem Framing

5 min read

Software Arch

The best architects don't just solve the problem they're given. They ask: 'Is this even the right problem?' AI can't do that.

Data Sci

AI can run the analysis. It can't ask 'what question would actually change the business?' You frame the question.

Tpm

Product says 'we need a dashboard.' You ask: 'What decision will this unlock?' That reframe changes everything. AI gets the reframed question.

Creative Problem Framing

TL;DR

  • AI is great at answering. It's terrible at asking.
  • Research shows users converge too quickly on early "good enough" AI results—a phenomenon called design fixation. Once you accept the first output, creative exploration stops.
  • Critical decisions remain human: "what to build and how to balance difficult tradeoffs." Both Claude Code and Copilot assume you bring the problem to the table.

Einstein (allegedly) said: "If I had an hour to solve a problem, I'd spend 55 minutes thinking about the problem and 5 minutes thinking about solutions." AI does the opposite. It jumps to solutions. You do the thinking. MIT research (2025) confirms: reaching high automation requires humans to own the critical decisions of what to build. AI scaffolds neither divergent thinking (ideation) nor convergent thinking (refinement) well—you're the scaffolding.

Why Framing Matters

Wrong Frame, Wrong Answer

  • "How do we make the checkout faster?" — AI suggests caching, indexing, query optimization. Good. But maybe the real problem is: "Why do users abandon checkout?" — Could be UX, could be trust, could be price. Speed might be a red herring.
  • "How do we reduce support tickets?" — AI suggests better docs, chatbots, FAQs. Maybe the real problem: "Why are we getting these tickets?" — Could be a bug. Could be a confusing product. Fix the cause, not the symptom.

AI Answers What You Ask

  • "Optimize this query." — AI optimizes. Maybe the right question: "Do we need this query at all?" — Different question, different solution.
  • "Build a feature for X." — AI builds. Maybe the right question: "Should we build X or buy it or not do it?" — AI doesn't ask. You do. Current interfaces impose either rigid step-by-step workflows or unguided free-form exploration—neither effectively supports flexible ideation. You're the one who explores high-level alternatives before AI locks you in.

Reframing Creates New Options

  • "We need to migrate to microservices." — Or do we? Maybe the real goal is scalability, and there are other paths. Reframing opens options AI wouldn't suggest.
  • "The system is slow." — Slow for whom? Under what load? Maybe the problem is perception, not throughput. Different frame, different fix.

What AI Can't Do

Challenge the Premise

  • AI assumes the problem is correctly stated. "Build X" → it builds X. It doesn't ask "should we?"
  • You can. "Hold on — is X actually what we need? What if we did Y instead?" The paradigm gap: human-AI co-creation needs new interaction patterns. Balancing guidance with flexibility—while keeping creative momentum—is still unsolved. You're the balancer.

Spot the Hidden Question

  • "We need better reporting." — Hidden question: "What decisions are we trying to support?" Unlock that, and the "better" becomes clear.
  • "We need to reduce costs." — Hidden question: "Which costs? And what are we willing to give up?" AI will optimize. You define the constraints.

Invent New Frames

  • "What if we thought of this as a supply chain problem instead of a software problem?" — Lateral thinking. AI works within frames. You create them. Claude Code thrives with a blueprint; Copilot with precise, specialist prompts. Both need you to bring the framing.

How to Use This as a Moat

  1. Before you prompt AI, ask: "Is this the right question?" Spend 2 minutes. It might change everything. Resist design fixation—don't accept the first "good enough" output before exploring alternatives.
  2. Practice reframing. Take a problem. State it three different ways. Each frame suggests different solutions. That's a muscle. Build it.
  3. Use AI for the framed problem. Once you've got the right question, AI accelerates the answer. But the question is yours.

Quick Check

A PM says: 'We need to make the checkout page load faster.' What's the most valuable thing you can do?

Do This Next

  1. Take one problem from your work. Write it down. Now write 2 alternative framings. "We need X" → "What we're really trying to achieve is Y, and X might be one way." See how it changes the solution space.
  2. Before your next AI prompt, add one reframe. "Not just 'how do we do X'—we're also considering whether Y would achieve the goal better." Make the framing explicit. That's your value. Resist accepting the first output—force at least one alternative before you converge.