Stop Rushing for Answers: Problem Framing Beats Faster AI

From problem solver to question definer

The tell was always workshops — the kind everyone dreads. “When you're sitting in a room with 20 people and everybody gets to talk, you're just sitting there sipping your coffee. I just want to go back to my desk and solve this issue because I know the answer already.” That line is the whole pathology: certainty arriving before the problem is understood. The fix was a new self-definition — not a problem solver but a question definer, someone who “doesn't immediately think they have the solution” and instead “stays in the moment of darkness for a little time” before deciding what's actually being asked.

This isn't a soft-skills aside. A Harvard Business Review survey of 106 C-suite executives across 91 companies in 17 countries found 85% agreed their organisations were bad at problem diagnosis, and 87% agreed the flaw carried significant costs — because managers “switch quickly into solution mode without checking whether they really understand the problem.” The apocryphal line usually pinned on a famous physicist captures it: “If I had an hour to solve a problem, I'd spend 55 minutes understanding the challenge and five minutes coming up with the solution.” Quote Investigator found no evidence he ever said it — the earliest trace is a 1966 manufacturing collection — but the misattribution barely matters. Productivity is set by how well you frame the problem, not how fast you reach for the answer.

Lego, Nike, Apple — three brands, one wrong question

Lego had the right word — creativity — and the wrong interpretation. Executives reasoned that creativity meant freedom: fewer instructions, more improvisation. It sounded unimpeachable. It was wrong. Lego's former CEO admitted it took “quite a few failures” to learn that “80 to 90% of those who build Lego love to build other people's creations” — that the age marking on a box isn't how old you are, but how advanced you are at following instructions well. The better question wasn't how do we give more freedom to explore? — it was what makes the experience rewarding? The proof is the Lego Builder app, which lets a stuck kid rotate the instructions in 3D. Not a freedom feature — a “help more people succeed” feature. In its 2025 full-year results, announced March 2026, Lego reported record revenue up 12% to DKK 83.5 billion, more than double the broader toy market's growth.

Nike is the cautionary version of the same failure, and the telling stays scrupulously fair — “sometimes you lose the focus and these kinds of things can happen even to a world-class brand.” Nike became Nike through advertising, not product innovation — Dream Crazier is the textbook case of a brand with cultural stakes in people's lives. Then, over roughly five years, Nike “chose to prioritise data-driven efficiency and direct consumer sales.” Nothing wrong with that in isolation — but marketing turned tactical, and the brand that made Nike iconic got starved. The result: the single-day stock drop of 28 June 2024, roughly 20% and $28 billion erased, the worst one-day fall since the 1980 IPO. It was the leading edge, not the bottom — FY2025 revenue landed at $46.3 billion, down 10%, with guidance pulled. Nike's own leadership later confirmed the diagnosis, shifting dollars back from performance marketing toward brand building. Two full years on, FY2026 full-year revenue is flat at $46.4 billion, with Nike Direct — the very channel it over-prioritised — still declining in Q4.

Right execution, wrong altitude. Performance marketing wasn't the villain — performance marketing answering the wrong question was.

Apple's case is thinner on evidence but sharper as illustration. The company's mission was always “what makes life better” — the 2007 iPhone reveal is mission expressed as product. “But somehow along the way… they had the wrong question in mind: how do we launch more products?” The 2025 correction was telling: rather than launch more, Apple re-sorted the iPhone line into four legible jobs — a budget 16e, a standard 17, a thin Air, and the Pros — so each device could lean into its strengths again. Not how do we launch more? but what is each product actually for?

The illusion of clarity: what AI actually changed

Here the argument turns urgent. “If we ask questions and we get answers just as fast, we get into the wrong path.” A confidently-worded AI answer feels like a finished thought, and that's the trap: ask how do we create content fast? and AI answers beautifully — but was speed ever the problem, or was it standing out? The feed is already saturated: independent 2026 estimates put genuine TikTok uploads at 23-34 million a day, with Meta's Reels reshares running into the billions daily. Creating more is easy; standing out is the hard part.

A run of well-made 2023 campaigns — Coca-Cola's Masterpiece, VisitDenmark's AI-ttractions, easyJet's Get Out There, Nestlé's The Milkmaid — all use the identical conceit of artwork coming to life. Each is genuinely well made. Watched back to back, they look the same. AI is very good at mimicking style and feel; it's not good at original concepts — it replicates and remixes existing patterns rather than inventing new ones. A landmark Science Advances study on AI-assisted short-story writing found the effect precisely: individual stories got “more creative, better written, and more enjoyable,” while “generative AI-enabled stories are more similar to each other than stories by humans alone.” Writers are individually better off; collectively, a narrower scope of novel content gets produced. The homogenization isn't a law of the technology — it's what happens when many people pull from the same model with the same shallow prompt. A sharper, human-defined brief measurably widens the output again.

“We outsourced the thinking and kept the confusion”

The sharpest practical idea inverts the usual AI productivity story. Type “come up with the brand strategy for this one” and you get seven confident, well-structured pages. The problem is never what AI produced — it's that “you never spent time on defining what you actually needed. You asked AI to do that for you.” AI didn't remove the work; it moved the work to the wrong stage, and now a team spends a week decoding what those confident pages actually mean.

Adoption data shows how widespread this trap has become: Salesforce's State of Marketing 2026 found 87% of marketers now use generative AI in at least one workflow, up from just 51% in 2024. Meanwhile trust is moving the opposite direction — Deloitte's 2025 Connected Consumer Survey of 3,524 US consumers found 74% of people familiar with generative AI say it makes it harder to trust what they see online, and 82% say the technology could be misused. A controlled study by the Nuremberg Institute for Market Decisions found that the identical ad, merely labelled “AI-generated,” was rated less natural and less emotionally resonant, with measurably lower willingness to click through — and the penalty hit hardest on emotional, brand-led work, exactly the Dream Crazier register that made Nike Nike. Adoption is near-universal; trust in the output is eroding. That gap is what an industry gets when it automates answers without improving its questions.

The quality of an answer is always limited by the quality of the question.

The EcomExpo playbook

The argument is not anti-AI. It is a discipline for where to spend the time the machine just gave back:

• Sell question definition as a billable phase — price the discovery step, don't give it away.
• Force a written test before any brief starts: “This brand word means ___ to us, and here is the evidence for the mechanism.”
• Audit winning clients for “right execution, wrong altitude” before the brand asset erodes.
• Treat “create content faster” as a symptom, not a brief — the real ask is almost always help us stand out.
• Use AI to expand the option set, then have humans force the divergence AI can't reach.
• Make the human — not AI — write the problem statement before any prompt is fired.

Make “I need more information” a respectable answer. The trick question every question-definer keeps in their pocket: what is the capital of Georgia? The answer isn't Atlanta, and it isn't Tbilisi either — it's “I need more information.” In a year when every competitor has the same answer machine, the durable edge is sitting in the unknown a little longer than everyone else is willing to.