Innovation Instability

In the AI Story

The instability operates on multiple timescales. Short-term: models are updated weekly or monthly, with each update introducing capabilities that change how the tool should be used. Medium-term: major model generations (GPT-3 to GPT-4 to GPT-5) represent qualitative shifts in capability. Long-term: the entire paradigm of how AI is deployed — chat interfaces, agents, embedded assistants, autonomous systems — is transforming at a pace Rogers's framework did not anticipate.

This instability has analytical consequences. Warren Schirtzinger, building on Rogers's framework, has observed that "you can still be an early adopter, twenty years later" — the curve resets with each capability leap, and the categories that describe static adoption curves may need reconception as responses to a continuously transforming trajectory.

The practical consequences are also substantial. Organizations that successfully adopt one generation of AI tools may find themselves effectively starting over with the next generation. Workers who developed deep expertise with earlier models may find that expertise devalued when new capabilities obsolete their mastered workflows. The institutional investments required to produce genuine integration must be repeated for each significant capability leap.

Rogers's framework can be extended to handle this partially. The five-stage innovation-decision process can be understood as iterating with each capability leap — awareness, persuasion, decision, implementation, and confirmation repeat for each major update. But the cumulative effect is a diffusion dynamic structurally different from the single-innovation-through-time model Rogers empirically validated.

Origin

Innovation instability is not a concept Rogers developed. It emerges from applying his framework to AI and discovering that the framework's assumption of innovation stability does not hold.

The theoretical analysis draws on work by Warren Schirtzinger, Geoffrey Moore, and others who have grappled with continuous-update technologies.

Key Ideas

Rogers assumed stability. The framework treats innovation as a fixed object diffusing through time.

AI is trajectory, not object. Each capability leap changes what the innovation is.

Curve resets with leaps. Adopter categories may require reconception as responses to continuous transformation.

Repeated investment required. Institutional adaptation must occur with each generation, not once.