Corporate innovation is genuinely hard. The organisations investing in it are not, for the most part, doing it carelessly. There is a structural pattern, though, that recurs often enough to be worth examining: the innovation programme that produces activity without adoption, and the question of whether the structure itself is part of the reason.
The argument here is not that labs never work. It is that the structure deserves more scrutiny than it usually gets.
The pattern
A senior leader gets a mandate to drive innovation. A team is assembled. A lab is built, physically or organisationally separate from the core business. Partnerships are announced. Hackathons are run. Pilots are launched.
Some of those pilots are genuinely good. Some of the ideas are worth pursuing. And yet, eighteen months in, adoption is lower than expected. The business units are less engaged than hoped. The lab is measuring itself on activities because outcome measurement requires the business to have used something, and the business has a day job.
It is worth being precise about what is actually failing here. Some projects fail because the idea was poor and the governance process let it through anyway. That is an idea quality problem and a different conversation. What this piece is about is the structural pattern that makes even good ideas hard to land.
The parallel structure problem
An innovation lab is a parallel structure. It sits alongside the business rather than inside it. The people running it report into a different hierarchy, operate on different timelines, and are measured on different things. From the perspective of a business unit leader with quarterly targets, it is something happening in a different building that occasionally asks for their time.
This creates a predictable dynamic. The lab does work. The business stays largely unbothered. Leadership can point to the programme as evidence that the organisation takes innovation seriously. The connection between what the lab produces and what the business adopts is weaker than it needs to be, because the structure never required the business to have genuine ownership of the problem.
There is also a harder version of this. Some organisations, without necessarily intending to, use the lab to avoid a more uncomfortable question: whether their core governance, their decision-making processes, their appetite for uncertainty, need to change. A parallel structure can become a pressure valve that allows business as usual to continue unchallenged.
Where the hub model does work
It is worth being honest about when the separate structure is the right answer. A hub that succeeds at scale often ends up becoming something different from what it started as: effectively a product or technology function, running at pace, delivering into the business with real adoption behind it. At that point it is not really an innovation lab any more. It is a delivery arm, and should be governed as one.
The question worth asking before standing up a separate structure is whether you need the separation to get there, or whether you could build that capability inside the business from the start. Sometimes the answer is genuinely yes, particularly where the core business is so operationally intensive that innovation work would simply be crowded out. That should be an active decision, arrived at deliberately, rather than a default.
An alternative worth considering
The model that gets less attention is embedding innovation work into existing structures rather than building alongside them. Using the governance frameworks and project pipelines the business already runs. Making participation as frictionless as possible for business unit teams, because the moment you ask people to engage with a parallel process on top of their existing responsibilities, most of them will deprioritise it.
This is less visible as a leadership signal. A lab launch has a ribbon-cutting moment, an away-day, a communications plan. The embedded model has none of that. What it has, done well, is innovation work the business actually owns, because it was never somewhere else to begin with.
The objection is usually that existing governance is too slow and too risk-averse to accommodate genuine innovation. That objection has merit. The answer is to address the governance directly rather than work around it indefinitely.
The discipline that makes it work
Whichever model you choose, the discipline that most innovation programmes never fully develop is upfront clarity on what success and failure look like.
Before any sprint begins, the parameters should be agreed in writing with the business owner in the room: what problem are we solving, what would tell us it is working, what would tell us it is not, and at what point do we stop. That conversation filters out a significant proportion of projects that were not ready to start. It also separates the question of whether an idea is good from whether the conditions for success exist.
Pre-agreed parameters change the nature of stopping. Without them, stopping a failing project is a visible admission of failure, which is why organisations rarely do it decisively. With them, stopping is a governance decision. The data said stop. You stopped. The organisation moves on.
What the embedded model leaves behind
There is a compounding benefit to the embedded model that is rarely part of the conversation. When innovation work happens inside existing structures, with existing teams, those teams get better at thinking about uncertainty as a byproduct of doing it. Through practice, on real problems, inside the structures they already work in, rather than through a workshop or a certification programme.
A well-run hub can deliver the same thing deliberately, through structured knowledge transfer back into the business. It requires that to be a design goal from the start, and it often is not.
The most durable innovation programmes, in my experience, are the ones where the business eventually cannot point to a single team or structure and say “that is where our innovation happens.” It is just how they work.