In Harvard Business Review, J.P. Eggers, Sarah Ryan and Asha Dinesh argue that AI has made building easy and that choosing what to build is now the real challenge, a piece Conor Grennan has called the single most practical HBR article of the year.
I agree with almost all of it. Where I’d push further though, is on what has to come before problem definition, and what has to surround it, if innovation is going to drive growth rather than simply more activity.
There’s a moment in many of my conversations with executive teams when someone says, with visible relief, “So the answer is better problem definition.” It’s an understandable reaction. Problem definition sounds teachable, contained and fixable. Run a workshop, introduce a framework, and move on.
The HBR article makes a compelling case for why problem definition matters more than ever.
The authors gave 33 NYU Stern students six hours to tackle some of New York City’s toughest challenges, from grocery affordability to bike lane safety. Teams were randomly assigned, most participants had no engineering background, and none had used the tools before. Every team still produced a functional prototype. What separated the winners wasn’t technical execution. It was how they framed the problem.
Conor Grennan’s summary captured the heart of it: “AI can’t do the hard part. It can’t determine what’s valuable – only humans can do that.”
I completely agree. But for CEOs and executive teams reading this through the lens of growth, I want to add two things. First, better defining a problem is not the same as defining a better problem. Second, problem definition is not a technique that survives on its own. It’s a behaviour, and behaviours only flourish in the conditions leaders create.
What the buildathon gets right
Let me start with what the authors get right, because it’s a lot. Their central observation, that AI has commoditised execution and moved the bottleneck upstream, matches what I see in boardrooms and leadership teams across sectors. When a product manager can build a working prototype without writing code, the scarce resource is no longer the capability to execute. It’s clarity about what is worth executing.
Their reframing of prototypes is equally sharp. When prototypes were expensive, we built them late to show the answer. Now they’re cheap, so we can build them early to “make disagreements visible while they’re still cheap to resolve.” That is one of the most useful lines I’ve read on innovation this year.
And their recommendation to organise teams by perspectives rather than skills echoes something I’ve argued for a long time. Human-centred approaches to unpacking and understanding problems faster sit at the core of how I help leaders learn to lead at pace.
Diverse perspectives see gaps that homogeneous expertise misses.
It’s worth noting the study’s scale: one day, eight teams, 33 students, observed rather than controlled. The authors are careful not to overclaim, and neither should we. But the direction of travel is clear, and it’s the right one.
When anyone can build anything, the scarce resource is no longer the capability to execute. It’s clarity about what is worth executing.
The work the organisers did before the students arrived
Look closely at the design of the buildathon and you’ll notice something important. The students didn’t start with a blank page. The organisers had already chosen the problem domains: grocery affordability, transit equity, bike lane safety, childcare access. Each is a recognised civic priority for the city.
The organisers had also set the judging criteria, explicitly prioritising problem definition and solution design over technical execution.
In other words, before a single student sat down, someone had already decided which problems mattered and what good looked like.
That isn’t a criticism of the study. It’s exactly what a well-designed experiment should do. But it means the buildathon tested problem definition inside a problem space that had already been strategically curated.
Inside most organisations, nobody does that curation. Or worse, everyone does it separately.
What I see far too often is organisations encouraging teams to “innovate with AI” and ending up with dozens of pilots, many of them thoughtfully framed and impressively built, solving problems that are entirely real but are not the problems standing between the organisation and where it has said it wants to go. Each one makes sense on its own. Together, they add up to motion rather than progress.
Better defining random problems isn’t the answer. Better defining strategically aligned problems is.
Strategy is a bet. Problems are how you place it.
Every strategy is, at its core, a bet about the future: where value will be created, which customers will matter most, what capabilities will differentiate, and where the organisation has chosen to play and not to play. Innovation-led growth happens when the problems an organisation chooses to solve are the visible barriers and opportunities that sit between today and that future.
This is why I encourage organisations to pose problems rather than ask for ideas.
Open ideas campaigns, without strategic framing, tend to produce a long list of internal frustrations rather than growth.
The more effective route starts with the strategic issues and opportunities tied directly to corporate strategy, framed as clear problem spaces and opportunity spaces, including those in the market and with customers, not only inside the organisation. Only then do you invite people to bring their creativity, and increasingly their AI tools, to bear.
For most established organisations, the richest territory sits in what I call the differentiated middle ground: problems and opportunities significant enough to need more than continuous improvement, but which don’t require reinventing the business model, or trying to be radical.
It’s where repeatable innovation capability pays for itself, and where strategic alignment in problem selection makes the biggest difference.
AI raises the stakes considerably. When building a prototype took months and a team, the cost of building acted as a crude filter. Projects had to justify themselves before anyone invested in them. But, that filter has gone. An organisation can now produce brilliantly framed, fully functional solutions to the wrong problems at extraordinary speed. AI is an amplifier. It makes louder whatever you point it at, including your lack of direction.
The research I cite most often suggests that approximately 84% of executives regard innovation as critical to growth, yet only around 6% are satisfied with their innovation performance. In my experience, a significant part of that gap is explained not by a shortage of good ideas or well-built solutions, but by effort pointed at problems that were never connected to the bet the organisation had made.
AI has removed the cost of building. It hasn’t removed the cost of building the wrong thing. It has simply made it faster to get there.
Who does the alignment work
If problem selection has to come first, the obvious question is whose job it is. My answer is that it’s shared, but not equally.
Executives own the innovation agenda. Their job is to make the strategic bet explicit, translate it into a small number of priority problem and opportunity spaces, and make those spaces visible across the organisation. Middle managers drive it. They turn those spaces into the specific problems their teams work on, allocate the time, and protect the space for exploration. The wider workforce contributes, and that’s where many of the perspectives the HBR authors rightly prize actually live: the person who has experienced the customer’s problem first-hand, the colleague who understands the regulatory landscape, the team that sees the operational reality every day.
Notice what this does not say. It doesn’t say executives should become better problem-solvers.
Problems get solved and opportunities get captured in the middle of the organisation. The executive role is to point the organisation at the right problems and to build the environment in which they get solved well.
It also suggests adding one perspective the buildathon didn’t need, because the organisers had supplied it: the strategic one. Every team working on a meaningful problem benefits from someone who can ask, “If this works, does it move us towards where we’ve said we’re going?”
Problem definition is a behaviour, not a technique
Now look at what actually made the strongest buildathon teams succeed. They pushed back on each other’s assumptions. They tolerated debates that felt circular. Someone was willing to slow everyone down to ask, “Wait, what problem are we actually solving?” They expected the first prototype to be wrong. They drew on lived experience of the problem, not just data about it.
None of those are techniques. They’re behaviours. And they depend on a set of conditions that most organisations claim to value and far fewer consistently live:
psychological safety, so people can challenge without fear;
human-centricity and empathy, so teams understand users deeply rather than abstractly; curiosity, so questions are welcomed before answers are demanded;
a growth mindset, so a wrong first answer is treated as progress rather than failure;
a genuine appetite for experimentation, so learning fast matters more than looking right.
Here’s the uncomfortable part. The buildathon was close to an ideal environment for those behaviours. Participants had no shared history, no hierarchy, no political capital at stake, a clear clock and an explicit instruction that framing mattered most. Organisations are rarely like that.
In many organisations, the person who slows everything down to question the brief isn’t celebrated. They’re seen as difficult. The circular debate gets cut short by the most senior voice in the room. The first prototype that turns out to be wrong becomes a mark against its sponsor. Teams are assembled for diversity of perspective, and then only the safest perspectives get voiced.
This is what I call ‘the innovation culture double bind’ in action: executives demanding the behaviours innovation needs while failing to demonstrate them, and often unintentionally punishing them.
Diversity of perspective without psychological safety is just a more varied group of people agreeing with the boss.
Leading FOR innovation, every day
So the executive agenda that flows from the HBR piece isn’t “teach problem definition.” It’s learning to lead FOR innovation: deliberately building the conditions in which strategically aligned problem definition and fast, safe experimentation happen as a matter of course.
That isn’t done at an offsite. Offsites have real value as thinking time, which is where futures get imagined and designed. But the conditions for innovation are built or eroded daily, in small moments that people watch closely:
How a leader responds to bad news.
Whether they ask questions before offering answers.
Whether they thank the person who challenged the brief.
Whether they protect exploration time when quarterly pressure bites.
Whether customer empathy is a slide in the strategy deck or a habit in the leadership team.
Leaders get what they role-model, tolerate and reward, which is why I keep returning to the play on “Culture eats strategy for breakfast” that I’m well known for i.e. “Strategy, leadership and culture need to eat breakfast together.”
Strategy defines the bet and the problem spaces worth solving. Leadership translates them and models the behaviours that make solving them possible. Culture determines whether people bring their best thinking or their safest. Remove any one of the three and problem definition quickly becomes another form of innovation theatre, just with better prototypes.
In the work I do with executive teams, the most effective approach I’ve found is to make real, costly, strategically important business problems the environment in which leaders learn to lead for innovation, involving managers and employees from the start. The problems get solved while the leadership capability and cultural conditions are being built, which means the return is visible rather than theoretical.
You can teach people to define problems in a workshop. Whether they do it on Monday morning depends on how their leaders behaved last Friday.
Building on the three recommendations
The authors offer managers three clear recommendations. Each is sound. Each becomes considerably more powerful with one addition.
Overindex on problem definition, yes, but first overindex on problem selection. Before a team debates how to frame a problem, the top table should be able to explain why that problem space matters to the strategy.
Use prototypes to sharpen the problem, yes, and use them to test strategic fit. A cheap, early prototype should surface “who is the actual user?” and also “if this succeeds, does it move us towards the future we’ve bet on?”
Organise by perspectives rather than skills, yes, and make it safe for every perspective to be heard. Include the strategic perspective and the customer’s perspective, and then make sure the leadership behaviour in the room allows the least senior voice to change the direction of the conversation.
Three questions for your top table
If the HBR piece has prompted conversations in your organisation, as I hope it has, let it also prompt these three questions.
- Could your executive team name the handful of strategic problem and opportunity spaces that stand between today and where your strategy says you’re going, and would your middle managers name the same ones?
- Where are your AI pilots and innovation investment actually pointed? At those spaces, or at whatever problems teams found interesting, visible or easy to build for?
- When someone in your organisation slows a project down to ask, “What problem are we really solving, and why does it matter to our strategy?”, are they thanked, tolerated or sidelined? And how confident are you that you would know?
The authors close by saying that the teams that figure this out first are “going to build things that matter.” I’d simply ask: matter to whom, and towards what?
For organisations pursuing innovation-led growth, the things that matter are the things that move the organisation towards the future it has chosen, and that customers recognise as genuinely valuable. Getting there takes excellent problem definition. It also takes strategic alignment, and leadership that brings the conditions and culture for innovation alive every single day. Problem definition is essential. On its own, it isn’t the panacea.
