Adoption vs Integration

In the AI Story

Edgerton's use-centered framework is, at its core, an insistence that integration matters more than adoption. The historical record shows repeatedly that adoption curves dramatically overstate the actual transformation that follows. The automobile was "adopted" rapidly in the United States in the early twentieth century, but full integration — the construction of paved roads, fuel distribution networks, mechanic infrastructure, traffic regulation, and the economic and social practices that made the automobile a load-bearing element of daily life — took decades. The personal computer was "adopted" rapidly in the 1980s and 1990s, but integration into the actual practices of most knowledge workers took until the 2000s, and arguably is still ongoing.

Applied to AI, the distinction reveals what the dramatic adoption curves do not. The Berkeley research that Edo Segal cites in The Orange Pill Chapter 11 documents the integration phase, not the adoption phase: workers in a 200-person technology company actually using AI tools in their daily work over eight months. The findings — task seepage, work intensification, the colonization of pauses by AI interactions — describe what integration actually looks like at the ground level, and they do not match the smooth productivity curves the adoption narratives predicted.

The integration phase is where most of the work of technological transition actually happens, and where the use-centered analysis becomes possible. Adoption is dramatic, fast, and easy to measure. Integration is slow, uneven, ambiguous, and only visible to those who do the patient ethnographic work of watching what people actually do with tools after the novelty fades. The marketing manager in Cincinnati who uses Claude to draft a slide deck and forgets she used AI by Thursday is in the integration phase; her experience is invisible to every adoption metric and represents most of what AI is actually doing in the economy.

The framework intersects with persistence of the old and the global deployment gap. Integration is what reveals which old practices persist alongside the new tool, which workflows are restructured and which are not, and which populations have the infrastructure necessary to convert adoption into integration. Adoption metrics treat all users as equivalent. Integration analysis reveals that some users derive sustained value from a tool while others abandon it after a few weeks, and the difference between these populations is the actual story of the technology's impact.

Origin

The distinction is implicit throughout Edgerton's work and is one of the analytical engines behind the use-centered framework. It draws on adjacent work in the sociology of technology, particularly Wiebe Bijker's social construction of technology framework and Susan Leigh Star's ethnography of infrastructure.

Key Ideas

Adoption is downloading; integration is changing how you work. The two metrics measure entirely different phenomena.

Adoption curves dramatically overstate transformation. Across every technology of the past century, the adoption curve has been faster than the integration curve by years or decades.

Integration reveals what adoption hides. The patient observation of how people actually use tools over time exposes patterns invisible to adoption metrics.

The marketing manager in Cincinnati. The median user, doing mundane work with the tool and forgetting she used it, represents most of what integration actually looks like.