This article is part of a series, "Organizing Against the Machines."
In fall 2025, Portland State University faculty had their first encounter with tangible results of the “PIVOT” initiative, an across-the-board academic program evaluation launched to help close what is now a projected $35 million budget deficit. For many, this began with an email containing a link to a “program vitality report,” and opening it felt a bit like receiving a corporate autopsy. Each section of the report displayed color-coded percentile scores for a dense array of metrics within five domains—student success, market demand, financial performance, mission alignment, and organizational viability—with red, orange, yellow, and green indicating where a program metric fell in the institutional pecking order. A department might find its degree program in the “green” for student retention but “red” for net revenue per student credit hour. The metric receiving the most attention was the “contribution margin,” the percentage of a program’s generated revenue left over after accounting for instructional costs. The faculty quickly grasped that these and other scattered flashes of red and green would be used upstream to sort programs, and possibly entire departments, into buckets: grow, sustain, revitalize, or sunset.
Along with the dashboards came an equally demanding mandate: Every one of the approximately 300 programs had to submit a “Track 1 self-study,” responding in up to 250 words to a battery of standardized prompts across the same five domains. Chairs and program directors scrambled to marshal faculty and staff to draft, revise, and upload these narratives into an online survey system on a tight fall deadline, generating tens of thousands of words of “context” in a matter of weeks. It was hard to escape the question that haunted many of us: In a process moving this fast, how much of that painstaking analysis would anyone with actual decision‑making power have time to read before the classifications—and the cuts—were effectively locked in?
The engine behind this process is Gray Decision Intelligence (Gray DI). This firm’s flagship product, the Program Evaluation System (PES), is marketed as a way to reduce “opinion, rules of thumb, and politics” in academic decision‑making and replace them with ostensibly “data‑informed” clarity. Yet the experience at PSU suggests that what is being sold as decision intelligence is better understood as a new frontier in academic austerity. In this model, the traditional framework of faculty-led governance is increasingly supplemented—and in some cases overshadowed, or even supplanted—by external firms, from global strategy shops like McKinsey & Company to specialized higher education boutiques such as EAB and Gray DI. This shift transforms power dynamics in the academy by applying the “logic of the marketplace” to the core academic mission: Academic departments are reframed as business units, faculty are managed as labor costs, and degrees are evaluated through the lens of return on investment.
The shift at PSU echoes a broader national turn toward consultant‑ and algorithm‑driven governance in higher education. Gray DI is part of an expanding ecosystem of firms that offer administrations standardized restructuring playbooks and portfolio‑management tools. The stated rationale for bringing in these firms is the supposed need for objective, outside expertise that internal constituents allegedly lack because they are parochial or too enmeshed in campus politics. The irony is that institutions facing structural deficits not infrequently devote scarce funds to these systems and consulting engagements up front, in the hope that algorithmically guided cuts and restructurings will eventually produce the promised savings.
Under the hood, PSU’s program vitality reports sit on top of Gray DI’s PES, which assembles local and external data into a scoring framework. PSU’s Office of Institutional Research and Planning was enlisted to provide data on enrollment, completions, student credit hours, and other indicators from PSU’s own systems, which Gray DI combines with labor‑market and competitor data pulled from national sources such as the Integrated Postsecondary Education Data System, job‑posting aggregators, and a benchmarking pool of peer institutions. Those raw inputs are then processed through Gray DI’s proprietary algorithms into composite market‑demand scores, revenue and expense allocations, contribution margins, and percentile rankings for different vitality domains. Much of what faculty see on the page are color‑coded indices rather than the underlying counts. In practice, this means that a program vitality report is less a neutral mirror of institutional and market data than a model shaped by the vendor’s definition of what “healthy” programs look like—one that embeds assumptions about demand, efficiency, and viability before any department chair or faculty committee begins a self‑study.
While it would be misleading to say that academic program reduction at Portland State is being “run by AI,” it would be equally misleading to treat PIVOT as untouched by it. We are told by the Office of Institutional Research and Planning that PSU licensed two primary Gray DI modules—PES Markets and PES Economics—and that at least one of the associated tools, Predict Program Size, relies on machine‑learning techniques to forecast enrollment and program completion. Gray DI describes its PES as a product that layers predictive models and proprietary composite scoring on top of institutional and external data, turning enrollment histories, labor‑market indicators, and financial data into ranked indices that help shape classifications. Even if only a subset of these features is currently enabled at PSU, the structure of the system means that machine‑learning‑based prediction and related AI techniques are already built into how “demand” and “viability” are quantified and displayed—framing what deans, chairs, and faculty committees see before shared governance bodies ever begin their deliberations.
Setting aside the Gray DI tools, there is a quieter, more speculative way in which AI may already be entwined with PIVOT. Track 1 alone required the submission of hundreds of self-studies through an online Qualtrics survey, keyed to multiple questions within each of the five vitality domains, yielding thousands of pages of narrative when multiplied across the institution. Track 2 has asked academic support units, centers, and administrative structures for similarly elaborate accounts of their “vitality,” further swelling the volume of text flowing upward. Nowhere do the publicly available materials explain how this torrent of prose was realistically read, coded, and compared in the span of weeks. In a campus environment where generative AI tools are already commonplace, one pressing, unanswered question is whether overworked administrators are leaning on those tools to summarize, cluster, or otherwise triage the self-studies, even though the university has articulated no norms for how such technologies should or should not be used in high-stakes program evaluation. And, if we’re being honest, administrators are not the only ones feeling this pull. Faculty members crafting self-studies under tight deadlines, faced with dense reports and unfamiliar metrics, may also turn to AI for help in processing, interpreting, and drafting, even as many of us worry about this same technology shaping the terms of academic program evaluation.
Gray DI’s own roadmap makes it clear where this is heading: from static dashboards toward AI-mediated, and eventually agentic, portfolio management. Its current suite of tools already layers several forms of AI onto program evaluation—machine-learning enrollment forecasts in Predict Program Size, large-language-model-generated narrative “AI reports” that synthesize dozens of market metrics into administrator-friendly prose, a conversational economics agent that lets deans query key metrics in plain English, and Program Remix, which uses automated catalog analyses to propose “remixed” academic programs built from existing courses. The direction is toward systems that do not merely answer questions about which programs to grow, shrink, or sink but also increasingly anticipate them—what Gray DI calls a shift from reactive AI to “agentic” decision intelligence, in which the software itself flags low-margin programs, emerging fields, or consolidation opportunities for leaders to act on.
Gray DI is not an outlier here so much as a bellwether for the broader higher education consulting market. Across the tools consultants already license and recommend—from survey and “customer experience management” platforms to program-portfolio tools—generative AI has shifted from optional plug-in to standard feature, handling coding, thematic summarization, clustering, and even draft report writing at scale. If current trends continue, the near future of consultant-led program restructuring is one in which university administrators are presented not just with dashboards but with AI-curated options and scripted rationales for cuts and consolidations—a world where the practical question for faculty governance is not whether AI will be involved but how far institutions will let vendor-designed agents predetermine the range of “reasonable” choices before any campus deliberation begins.
PIVOT, in most respects, has not been a closed, backroom exercise at Portland State. In addition to requiring the self-studies, administrators have gone to some lengths to involve faculty and even union leaders in the process, from summer workshops that trained a select group in using the Gray DI platform (with ongoing access to the dashboards) to campus-wide town halls, college and school meetings with the provost, and regular Q&As in senate and senate committee sessions. Faculty and staff have been kept apprised through repeated PIVOT updates in the campus newsletter, a dedicated web presence for the broader Bridge to the Future 2.0 strategy, and regular “PIVOT Pulse Check” emails from the Office of Academic Affairs, which track milestones and next steps. Yet even against this backdrop of communication and consultation, key parts of the decision-making pipeline remain opaque to the very faculty members whose programs are on the line—most notably, how deans and central administrators actually move from program vitality reports and narrative self-studies to their verdicts. When pressed, administrators have not described any formal rubric or scoring guide they are applying, and if Gray DI has supplied a recommended method for using its metrics and faculty narratives to arrive at these judgments, that logic has not been shared in the PIVOT materials available to the campus.
However extensive PIVOT’s meetings, workshops, and updates may be, consultation is not the same as shared governance. PIVOT is unfolding under the shadow of what university administrators project to be substantial budget shortfalls for the next few years. Although we are told that the university does not need to declare financial exigency, our collective bargaining agreement allows the president to truncate normal governance processes even under conditions short of a formal declaration of exigency (though she cannot simply decree program closures by fiat). That is the situation we now face: Alongside PIVOT, the administration has initiated a formal retrenchment process under the contract, separate on paper from curriculum review but drawing on the same deficit projections and program classifications. In practice, this combination of consultant-guided program evaluation and fiscal urgency has sharply constrained—and in some venues effectively hollowed out—the space in which shared governance can operate. Faculty are invited to review dashboards, complete self-studies, and attend town halls, but the ultimate framing of options and the pacing of decisions are driven by deficit targets and vendor systems that were never themselves the subject of genuine, faculty-driven deliberation.
The structural challenges facing public universities are real, and administrators will continue to turn to data and AI-enabled tools from firms like Gray DI to rationalize program reductions and eliminations. The question is not whether administrations will use consultant-built systems but whether the faculty will allow those systems to operate without clear standards, validation, or limits. We already devote considerable attention to articulating norms for generative AI in classrooms and research; the stakes are at least as high when AI-enabled analytics help decide which of our academic programs survive. Because this era of austerity is increasingly mediated through decision-intelligence platforms, faculty senates and unions must press for concrete safeguards: requirements that models be locally tested against institutional experience, that methods and data sources be open to scrutiny, and that governance bodies—not vendors—define how far algorithmic “advice” can reach into decisions about academic programming and curriculum.
David Kinsella is professor of political science at Portland State University and PSU-AAUP’s vice president for collective bargaining. His email address is [email protected].