The Innovation Advantage GenAI Can’t Give You

For most of modern business times, competitive advantage belonged to whoever had the best ideas. Better ideas meant better products, which meant more customers, which meant more revenue and profit. The entire innovation industry — consultancies, design firms, brainstorming retreats fueled by sticky notes and gallons of La Croix — was built on this premise: If you could generate more and better ideas than your competitors, you would win.

That advantage has been vaporized by AI.

Generative AI has turned ideation into a full-blown utility. Today, anyone with a $20 subscription to a GenAI tool can instantly generate 100 product concepts. That has rendered the raw material of innovation — ideas — as abundant, accessible, and cheap as electricity. And here’s the thing about electricity: Nobody competes on it. You compete on what you build with it. Which means the competitive advantage has shifted upstream, from the solution to the problem — specifically, to how you identify and frame the problem in the first place.

This is something I’ve taught for years — to executives, MBA students, and others — going back to my time as a designer at IDEO. It is called Question Zero: the question before the question. Before you ask, “How do we solve this?” you need to ask, “Are we even looking at the right problem?” The quality of innovation has always been determined by the quality of problem framing. But until recently, most organizations could get away with mediocre problem framing. Why? Because ideas were scarce enough to be valuable on their own.

That’s no longer the case. When everyone has access to the same idea-generation engine, the remaining edge is the insight that tells you where to point your business. GenAI won’t give you this insight, though it can surface data and patterns that help you see it. Let’s examine why businesses continue to frame the wrong problem, examples of startups and established businesses reframing successfully, and how to get started.

Why Most Organizations Frame the Wrong Problem

If problem framing is so important, why is everyone so bad at it?

It’s because the “best” problems — the ones that lead to the most valuable, genuinely differentiated solutions — are almost always hidden. And they’re hidden for a specific, annoying reason: The people who experience them can’t tell you about them.

This is something my colleague Loran Nordgren and I discuss extensively in our book, The Human Element. Users experience friction with your product, your service, your entire category — but they can’t explain it. They know how they feel but not why they feel it. The friction is real. The self-awareness is nonexistent.

Ask a customer why they abandoned your app and they’ll likely say, “I got busy.” The real answer — the one hidden in the emotional recesses of their brain — might be that your onboarding flow made them feel like they’d accidentally wandered into an advanced calculus class. They’re not going to tell you that, because they don’t even know that’s what happened. They just know they stopped opening the app.

This means that the standard problem-identification toolkit — surveys, focus groups, net promoter scores, quarterly voice-of-customer decks — captures only what people can and will articulate. The bad news is that what people can and will articulate is, at best, the surface problem. Understanding the surface problem leads to incremental solutions, which, by definition, are undifferentiated. You end up competing on features, then price, then “vibes.” This is not a strategy; it’s a slow descent into commodified oblivion.

The deeper problem — the reframed one, the one worth solving — lives in the gap between what people say and what they do. Finding that gap has always required the kind of deep, patient observation and investigative interviewing that most organizations can’t afford or feel that they don’t have time for; it’s something that doesn’t lend itself easily to a slick 2×2 framework in a PowerPoint deck. So most companies just skip it and go straight to brainstorming, which they consider the fun part.

AI changes this equation. Not because it replaces human insight — AI has no insight; it has pattern recognition and a Stuart Smalley tone of relentless encouragement — but because it can surface the behavioral patterns that lead to human insight at a scale and speed no human team can match.

Ultimately, then, AI is not the insight but the high-powered telescope that makes the insight visible.

The Startups That Won by Reframing

The clearest proof that problem reframing drives differentiation comes from startups that have broken through in a big way in the past two years — not by having better technology but by asking Question Zero about problems everyone else had framed in less original ways.

Take Cursor, an AI-powered code editor that hit $1 billion in annualized revenue and a $29 billion valuation in 2025. Every other company in the space framed the problem the same way: “How do we help developers write code faster?” GitHub Copilot was already solving that, and solving it well. But Cursor’s founders — four MIT graduates barely old enough to rent a car without extra fees — saw something different. Developers weren’t actually spending most of their time writing code. They were spending it reading code: navigating unfamiliar code bases and trying to understand what someone else built three years ago at 2 a.m. The bottleneck wasn’t composition. It was comprehension.

That reframe — from “write faster” to “understand better” — produced an entirely different product, an entirely different company, and an entirely different, much-higher-value outcome. Same market. Same underlying technology. Very different problem solved.

Meanwhile, Speak, a language-learning app that raised $78 million and reached a $1 billion valuation in late 2024, tells the same story in a different domain. The obvious framing in the sector was “How do we teach grammar and vocabulary more effectively?” Every competitor was running that race, and Duolingo was winning by several laps. Speak’s founders reframed the challenge: “Why are people who study a language for years still terrified to open their mouths and speak it?” The answer isn’t that there’s a knowledge gap. It’s a confidence gap — the fear of sounding foolish in front of others. But nobody describes their problem that way. No language learner walks into a class and says, “I’m here because of shame.” They say they need more practice.

So Speak built an AI conversation partner that lets learners mangle a subjunctive without anyone grimacing at them and then provides a gentle correction. The technology is impressive. But what really made it work was the reframe. The real problem was never learning. It was the emotional friction around learning.

In the productivity industry, Fireflies.ai reframed a common meeting problem. When everyone was asking, “How do we make meetings shorter?” Fireflies asked, “What if the real waste isn’t the meeting itself but everything that happens after it?” That includes the hours spent writing summaries nobody reads, chasing action items nobody remembers, and gently reminding Kevin that he did, in fact, agree to that deadline last Tuesday. The meeting wasn’t the problem; it was the evaporation of the meeting’s output. That reframe produced a product the “shorter meetings” crowd couldn’t compete with, because even though they might have been building a truly better mousetrap, they were in the wrong room from the start.

In each case, these startups didn’t out-ideate the competition. They out-framed them. They saw the same market and found a different problem within it — one that led to a solution nobody else was creating because nobody else had seen the problem the way they had. Ideas were never the bottleneck; the originality of the problem framing was.

How Established Companies Use AI to Surface the Reframe

The startups mentioned above achieved innovative reframing through intuition and proximity. Established organizations can deliver the same through AI-powered behavioral observation at scale. There are multiple examples of this among some of the best-known companies. The pattern is remarkably consistent: The AI agent doesn’t generate the reframe; it surfaces the behavioral data and patterns that make the reframe possible. The human still has to have the insight, but the AI makes sure there’s something to see.

For example, Netflix spent years framing its core challenge as a genre problem: “What genres does this subscriber prefer?” The AI’s job was to match users to categories — perfectly reasonable but also, it turns out, a pedestrian framing of the problem. By using AI to observe behavior at scale, Netflix discovered something no focus group sessions could have surfaced: People weren’t browsing by genre. They were browsing by mood.

The difference between a Friday night with friends and a Sunday alone after a bad week isn’t an action-vs.-comedy distinction — it’s an emotional vibe. Nobody ever submitted a feature request that said, “Let me search by how I feel.” But the behavioral data was unmistakable. To capitalize on this observation, in 2025 Netflix began testing an AI-powered search that lets users describe what they’re in the mood for rather than what category they want. The reframe — from genre preference to emotional need — didn’t emerge from a product road map. It emerged from paying attention to what people actually did, at scale.

Another example is Duolingo’s AI system, Birdbrain, which surfaced a reframe that no curriculum designer had considered. By analyzing billions of data points across dozens of language pairs (a learner’s native language and the language being learned), Birdbrain discovered that certain combinations had dramatically higher dropout rates, but in patterns nobody had expected. Spanish speakers learning Portuguese, for instance, were more likely to stop using the app when working on lessons where the two languages were almost identical rather than where they differed: Similarity breeds overconfidence.

Specifically, learners cruised through lessons feeling great, acing quizzes, collecting little digital trophies — right up until they quietly stopped opening the app altogether. All that reinforcement made them feel like they had mastered the new language when in fact they would have struggled to use it in the real world. No survey would have caught this. People don’t report confidence as a problem — they report it as a virtue.

The old frame: “How do we make lessons more engaging?” The reframe: “Where is false confidence silently killing retention?” That second problem can lead to a fundamentally different — and better — solution, such as more subtle tests of mastery for more similar language pairs.

In a different consumer-focused domain, Procter & Gamble’s AI crawled parenting forums and social media and surfaced a behavioral signal no product team would have thought to look for: Parents were using adult skin-care products on their babies. It wasn’t because they were fans of CeraVe’s minimalist branding but because they had given up on baby-specific products entirely: They’d decided that the whole category was either ineffective or filled with chemicals they didn’t trust.

The old frame: “How do we make a better baby lotion?” The reframe: “Why have parents stopped believing us?” That’s not a product problem. It’s a trust problem. And the reframe changes everything: the product, the messaging, the entire go-to-market strategy. You can’t “new and improved” your way out of a credibility crisis. P&G harnessed that framing to engage with and educate parents better through tactics such as product-level personalization and real-time quality and innovation feedback loops.

Then there’s the most meta example of all. Anthropic, the company behind the AI model Claude, built a tool called Clio — Claude Insights and Observations — that uses AI to observe how millions of people use AI. (Yes, it built an AI to watch people talk to their AI.)

Clio clusters millions of conversations and surfaces behavioral patterns invisible at the individual level. It discovered, for example, that Japanese users disproportionately discuss eldercare — a cultural trend and signal observable only at scale. Additionally, it found that users in crisis arrive through specific conversational pathways that single-message safety filters miss entirely. Subsequently, in a particularly humbling discovery, it revealed that Claude’s own safety systems were simultaneously refusing harmless requests (“kill a process” on a computer) while passing over some genuinely concerning ones that could have placed people at risk in the real world.

Anthropic’s original frame: “How do we make our safety filters more accurate?” The reframe: “We’re measuring safety at the wrong unit of analysis entirely.” The insight and reframing didn’t just improve the product. It changed the company’s understanding of what the problem was.

Three Steps to Get Started

As the examples suggest, the reframing chain works like this: Better behavioral data leads to better problem reframing; better reframing leads to more novel solutions; and more novel solutions lead to more differentiated products, services, and businesses. And that is the only thing that matters when AI has turned raw ideation into something anyone can do in their pajamas. Here are three ways to start the cycle at your organization.

1. Surface the gap between what people say and what they do. Point your AI tools at customer support logs, forum posts, social media mentions, and review data. Look specifically for workarounds — hacks, improvised fixes, ways people use your product that you never intended and would likely even find mildly insulting. Developers spending 70% of their time reading other people’s code is a workaround. Parents using CeraVe on their babies is a workaround. Language learners who ace every quiz but won’t order coffee in the language they’ve been studying for three years is a workaround. Every workaround is a reframe waiting to happen.

2. Audit your problem frames before you generate solutions. Get your team in a room and write down the problem you’re currently solving — the one driving your road map, your next sprint, your big second-quarter initiative. Then ask, “When was the last time we tested whether this is actually the right problem? What might a competitor see that we haven’t been able to? What if the opposite of our core assumption is true?” If the problem frame hasn’t been challenged in the past 12 months, you’re not innovating; you’re redecorating.

3. Use AI to reframe, not just to ideate. Most people prompt AI with “Give me 10 ideas for X.” That’s fine if you want 10 mediocre ideas delivered with confidence. Instead, feed your AI the behavioral data, the workarounds, and the surprising signals and ask it to generate alternative framings of the problem itself. What if the problem isn’t retention but overconfidence? What if the problem isn’t product quality but category trust? What if the problem isn’t the meeting but the aftermath?

Remember: The AI won’t reframe the problem for you. But if you give it the right inputs, it’ll help you generate framings you wouldn’t have reached alone.

Ideas used to be the scarce resource. Now the scarce resource — the thing that actually drives differentiation — is the insight that reframes the problem. Working this way requires a proactive shift from solving the obvious thing to solving the right thing. AI, for all its generative power, turns out to be most valuable not when it produces answers but when it helps you see a problem you didn’t know you had.

The companies that figure this out won’t just build better products. They’ll build products that nobody else thought to build.