The Hardest Moment in Any Innovation Programme
The hardest moment in any innovation programme is not generating the ideas. Most organisations are surprisingly good at that part. The hard part is the morning after — when the workshop is over, the sticky notes are photographed, and everyone returns to a job whose priorities have nothing to do with what was just discussed.
Ideas that cannot survive contact with organisational reality are not innovations. They are aspirations with a shelf life of about a fortnight.
In 2022 I led a team supporting UNSW's Arts, Design and Architecture Innovation Hub that set out to solve exactly this problem. An intensive design marathon had produced five compelling project teams, each with an idea worth pursuing and limited industry traction to pursue it. My team was brought in to bridge that gap. What we learned applies well beyond social innovation — and in 2025, I think it applies most directly to the challenge of AI adoption.
The pitch is not the idea
The first thing we discovered was that every project team had a sophisticated understanding of their problem and a genuinely novel response to it. What none of them had — and this is almost universal in my experience — was the ability to communicate that response in a way that moved an external audience from interest to commitment.
The problem is not intelligence or intent. It is that people who live inside a complex idea lose the ability to see it from outside. The language they use is precise but opaque. The context they provide is thorough but overwhelming. The ask — when there is one — is often unclear.
We ran a series of workshops focused on what we called a Minimum Viable Movement: the smallest, clearest, most compelling version of each project that could attract genuine external commitment. Not a full business case. Not a prototype. A movement — something with enough clarity and conviction that an external stakeholder could decide, within ten minutes, whether to back it.
That discipline — stripping an idea back to its essential proposition and its most honest ask — is one of the most useful things I do with senior teams, and it is as relevant to an AI transformation programme as it is to a social innovation project.
Events are midpoints, not endpoints
The UNSW team wanted to host an industry engagement event to build support for the five projects. The default approach in these situations is to design the event as the destination: invite people, present the work, hope for the best.
We designed it as the midpoint instead.
In the weeks before the event, we did systematic pre-engagement with industry leaders — not to tell them about the projects, but to involve them in shaping the asks. By the time they walked in the room, they were not arriving as an audience. They were arriving as participants in something they had already begun to invest in.
Each team left with specific commitments — 37 pledges across the five projects, from industry partners offering resources, connections and expertise. The post-event strategy was designed before the event ran. Which is the only way any post-event strategy actually works.
Commitment beats enthusiasm every time
The most important design decision in the event itself was the pledge mechanism: industry leaders were asked not to express support, but to make a specific, named commitment to a specific initiative, in a social setting where others could see them do it.
This is not manipulation. It is behavioural design. The distance between "I found this interesting" and "I will do this specific thing by this date" is enormous, and most engagement events never close it. Creating the conditions for public commitment — specific enough to be real, visible enough to be sustained — is one of the least used and most effective tools in the change management toolkit.
What this has to do with AI
The methodology behind all of this is called life-centred design — starting from the question of what a solution needs to do for a real person, rather than from the question of what a technology or system is capable of. It is the opposite of the way most AI programmes are currently being designed.
Most AI adoption programmes I see begin with capability: here is what the model can do, now let us find somewhere to deploy it. The result is technically impressive tools that are practically ignored, because they were designed for organisational efficiency rather than human experience.
The UNSW project started from the other end: here is a human problem of genuine consequence, now let us design the minimum viable version of a solution and build the coalition to deliver it. The technology was an instrument, not the driver.
In 2025, that sequencing — need first, technology second, honest conversation about both throughout — is the single most important discipline I bring to organisations navigating AI adoption. It is harder than it sounds. Most organisations have spent two years running it in reverse.
