Automating versus Informating

In the AI Story

The distinction dissolves the debate between technologists who celebrate capability expansion and critics who document displacement. Both are right: the capability expansion is real, the displacement is real, and they are features of the same technological event. The question is not which dynamic dominates but whether institutions build the structures required to capture the informating potential or whether they extract the automating cost-savings and discard the informating investment as expensive inefficiency. Zuboff's empirical record shows the latter is the default institutional response across industries, decades, and geographies. The paper mills that computerized in the 1980s could have trained floor workers to interpret production data, elevating embodied expertise into analytical expertise. Most did not—they moved experienced workers to monitoring roles, reduced hiring, captured automation's labor savings. The informating potential remained potential.

AI produces both functions at unprecedented scale. The automating side: entire categories of cognitive labor—coding, analysis, drafting, design—approaching zero marginal cost of production. The software death cross, the SaaSpocalypse, the restructuring of knowledge work documented in The Orange Pill—all expressions of automation's comprehensive reach into domains built on the assumption cognitive labor could never be automated. The informating side: equally unprecedented expansion of what humans can understand—patterns in datasets too large for individual cognition, hypotheses generated at scales human researchers cannot approach, integrated cross-domain insights that disciplinary boundaries previously blocked. Segal's thirty-day Napster Station build demonstrates the informating dividend realized: human judgment amplified by tools that eliminated translation friction between vision and artifact.

The institutional choice between the two functions is not made once but continuously. Every organization deploying AI confronts the choice in every budget cycle: convert the twenty-fold productivity gain into headcount reduction (automating) or invest in expanded team capability (informating)? The market rewards the first choice with higher margins. The second choice is a bet that the long-term value of developed human capability exceeds the short-term value of reduced labor costs—a bet that quarterly earnings pressure makes structurally difficult to sustain. Zuboff's framework predicts that without institutional structures operating at market scale (regulation, labor protections, educational standards, collectively bargained norms), individual organizational choices toward informating will be outcompeted by choices toward automating, and the informating dividend will be systematically squandered as it has been in every previous transition.

Origin

The distinction originates in In the Age of the Smart Machine (1988), Introduction and Chapter 1, where Zuboff first observed that information technology was categorically different from industrial machinery: it not only executed work but recorded its own operations, generating data as inherent byproduct. The observation drew on cybernetic theory (Wiener's feedback loops) and organizational learning theory but was distinctively Zuboff's in its insistence that the data-generation function created genuine new knowledge potential—not merely faster operations but qualitatively different understanding—whose realization depended on institutional choice rather than technological capability.

Key Ideas

Not alternatives but co-occurring dynamics. Every smart machine simultaneously displaces labor and generates knowledge—the question is never whether to automate or informate but which function institutions prioritize through investment and organizational design.

Markets select for automating. Cost reduction is immediately measurable and rewarded in quarterly results; human development is long-term investment whose returns operate on timescales competitive markets systematically undervalue.

Informating requires institutional investment. Training workers to engage with new knowledge forms, redesigning organizations to distribute authority to those who possess new knowledge, protecting practice opportunities that productive efficiency would eliminate—none occur without deliberate institutional choice.

Default is squandering the dividend. Zuboff's empirical finding across four decades: the cheaper path prevails absent institutional structures that force internalization of long-term costs—the eight-hour day, weekend, labor protections did not emerge from enlightened management but from collective action.

AI is the largest test case. The informating potential is unprecedented; the automating pressure is correspondingly intense; the institutional adequacy of current responses is demonstrably insufficient to capture the dividend at the scale the transition demands.