Literature Review

The FYR literature review, built as a problematization (Sandberg and Alvesson 2011): it surfaces and challenges an assumption a set of literatures shares, rather than finding a gap one of them has left. The shared assumption is evaluative continuity, and the four canonical lenses hold it. PSF breaks it.

The review opens with how the literature was gathered: a structured Scopus search across four blocks (organizational ambidexterity, institutional logics, sociomateriality, and AI and organization) produced the working corpus, tracked, tagged, and annotated in a Zotero library.

The continuity assumption across four literatures

The shared assumption is evaluative continuity: the organization engaging with AI persists as the same evaluating subject post-engagement, so that pre-engagement criteria remain valid for post-engagement assessment. Each of the four lenses names part of the phenomenon and holds the assumption that keeps it from naming the rest.

• Sociomateriality (Faulkner and Runde 2019, Leonardi 2011, Orlikowski 2007) treats technology and organization as mutually constituted through practice and establishes that a tool's function is not fixed in its form. It assumes practitioners can evaluate how the technology has changed their practice using criteria the technology has not already changed.
• Institutional logics (Thornton, Ocasio, and Lounsbury 2012) shows how market logic pushes attention toward efficiency and throughput while professional logic pushes toward domain quality. It does not ask whether the technology itself constitutes the confirming evidence that tips the organization toward one logic.
• Dynamic capabilities (Teece 2007, with ambidexterity and absorptive capacity) holds that firms sense, seize, and reconfigure, yet assumes the sensing apparatus survives the reconfiguration.
• Epistemic regimes (Sergeeva, Leonardi, and Faraj 2026) argues that intelligent technologies produce knowledge claims grounded in logics unlike those underpinning professional expertise, but presupposes that actors experience the new regime as a contest they retain the capacity to navigate.

The resources PSF integrates across levels

PSF assembles its account across levels, taking each resource for the specific work no other resource does. This follows phenomenon-based theorizing (Fisher, Mayer, and Morris 2021): the phenomenon does not sit inside any one literature, so the framework integrates across them rather than extending one.

• Individual level · transformative experience (Paul 2014): why a practitioner cannot evaluate the engagement before entering it. Some experiences change both what you know and what you value, so the criteria you would judge the engagement by are themselves changed by it.
• Tool-organization interface · form and function (Faulkner and Runde 2019): why the same tool produces different gaps in different organizations. Form supplies capacities; function comes from how a community positions the tool. PSF names the property of technologies whose form admits an unusually wide range of positionings fertile form.
• Organizational level · institutional logics as extended: the different visibility of market and professional metrics under AI engagement drives the displacement.
• Field level · Barnesian performativity: field-level discourse performs specific proxies into legitimacy before organizations test whether those proxies track their criteria.

Moving the individual account up to the organization needs an argument, not an assertion, since Paul writes about individual agents. The move holds because Paul's structure fits any agent with evaluative capacity, not only a person.

What the evidence shows, and what it cannot yet settle

The section closes by tracing how evaluative capacity erodes and drawing the evidence strategically rather than cataloguing it. The signature case is METR (2025b), a pre-registered randomized trial in which experienced open-source developers, working on their own code with every reason to judge accurately, expected AI to speed them up and were measurably slower. The dissenting and complicating findings are engaged, not set aside: Humlum and Vestergaard (2025) find small labor-market effects in Denmark despite widespread engagement, and the Yale Budget Lab (Gimbel et al. 2025) finds no significant displacement in high-exposure occupations. A subset of the philosophical apparatus (extensions beyond Paul) is carried forward as adjacent and exploratory, not as settled ground.

The review thus establishes the assumption PSF challenges, the four literatures that hold it, the resources PSF integrates to move past it, and the evidence that the pattern holds across levels.