Generative Power — compounding your organization’s creative capability with AI
When Jon Gertner was researching the history of Bell Labs, he noticed something peculiar about the building. The corridors connecting different departments were long and unavoidable. If you were a physicist and wanted lunch, you had to walk past the engineers. This was not accidental: the most valuable creative work happens at the intersection of people who know different things deeply, and the primary task of organizational design is to make those intersections possible.
AI changes what is possible. For the first time, the behavioral traces of creative knowledge work — what gets solved, whose outputs others build on, who gets called when something genuinely hard needs thinking through — can be analyzed continuously and at scale. The Bell Labs corridor can now span an entire organization. But the same tools that surface and connect human creative capability can, if deployed without deliberate design, quietly erode it.
"The Bell Labs corridor can now span an entire organization. But AI also introduces a risk that most leaders are not tracking: the same tools that connect human creative capability can, without deliberate design, quietly erode it."
This paper introduces Generative Power — the organizational capacity to compound human creative capability through structured interaction with AI — and the Creativity Stack, the six-layer architecture that builds it.
AI raises individual output — and narrows what a population collectively produces
The difference has nothing to do with which tools are used. It is entirely about how the interaction is structured. Without deliberate design, the economic default is substitution: immediate output is maximized when AI does the cognitive work.
Four dimensions of Generative Power — multiplied, not added
As AI commoditizes general human capital — the analytical, technical, and communicative skills any organization can now access through the same tools — the advantage that endures is the creative capability built through the specific interactions, practices, and accumulated experience of a particular organization. Generative Power is not a stock you possess. It is a capacity you develop through deliberate design of how AI-mediated creative interactions are governed. It is the human-side counterpart to the machine-side Learning Power of Paper 04.
The four dimensions multiply rather than add. High Depth and Variety with slow Velocity is potential that does not compound; all three are irrelevant if Trajectory is negative. The system is bounded by its weakest dimension — so developing the weakest produces more return than investing further in the strongest.
Six layers that build Generative Power
Why the default trajectory is negative
Toyota’s kaizen system illustrates what a full Stack looks like: workers detect problems, represent them through the kaizen card, receive genuine recognition, and remain actively engaged in improving the work. When Western manufacturers copied the visible tools — kanban cards and quality circles — without the underlying culture of engagement and recognition, the systems consistently failed. The tools without the governance produced nothing. This is the pattern AI deployment is now repeating at scale.
Organizations adopt AI tools and watch output metrics rise. What they are not measuring is whether the people generating those outputs are becoming more capable or less. The divergence is invisible until it is significant — and by then, the capacity to rebuild the lost expertise has itself declined. Recovery is substantially harder than prevention.
Three design choices that determine Trajectory
- 01Design for engagement, not approval. AI-assisted creative tasks should require humans to genuinely evaluate, question, and restructure AI outputs — not review and approve them. The same AI tool produces prosthetic or substitutive interaction depending on whether the workflow requires thinking or only clicking. This is a workflow design decision, not a policy statement.
- 02Measure Trajectory alongside output. Add periodic assessments of unassisted performance, calibrated to role and level, to track whether human capability is developing or becoming dependent on AI. If unassisted performance is flat while AI-assisted performance rises, the organization is on the depletion trajectory. Report the gap to senior leadership alongside standard performance metrics.
- 03Govern Variety as a managed metric. Track the diversity of your organization’s creative outputs over time. Rising output volume alongside converging style is the homogenization warning — rotate the perspectives used in creative review and deliberately commission work from the edges of the current distribution.
