Dr. Kristina McElheran stood before a group of academics and industry watchers at Wharton’s annual AI and the Future of Work Conference and delivered a sharp insight that might even be called disruptive.
AI, the great productivity hope, the digital wunderkind of industry, initially harms productivity.
That unsettling idea, backed by data from the real operations of American manufacturing firms challenges an industry that has bought into the notion that AI seamlessly accelerates productivity from day one.
McElheran's work sketches out a J-curve showing productivity dips hard when companies first embrace AI, only recovering (and potentially exceeding) former levels after substantial disruption.
McElheran, working closely with colleagues MJ Yang, Zack Croft, and Eric Reolson, spent years peeling back layers of operational data from thousands of manufacturing firms in collaboration with the U.S. Census Bureau. Their discoveries, particularly in how AI interacts with organizational structure and management practices, reveals that manufacturing companies adopting AI frequently dismantle (or unintentionally sabotage) the very management practices that once made them efficient.
The research revealed that AI adoption leads many established firms, particularly those rooted in decades-old management processes, to temporarily abandon these proven management frameworks.
The research team contends that while chasing the promise of digital efficiency, organizations inadvertently disrupt routines that historically boosted productivity and supported operational excellence.
The surprise isn’t just the depth of this disruption, it’s the extent to which the AI-driven reconfiguration cuts across core business functions, toppling what was once considered stable management terrain.
It’s worth pausing here to unpack exactly why AI, intended as a management enhancer, initially acts like a management saboteur.
The explanation, as McElheran’s findings hint but do not deeply probe, lies in the radically different organizational demands posed by AI, which presents a radical shift, demanding dramatic and immediate changes in how tasks and processes are conceived and executed.
Older firms, McElheran says, face uniquely severe struggles. Companies with longstanding operational legacies (ironically those that should benefit immensely from AI-enhanced efficiency) see deeper productivity losses because their processes are rigidly embedded.
The older the firm, the greater the friction. Data-driven approaches replace stable workflows, historical KPIs become irrelevant, team routines splinter under uncertainty.
Older companies are forced into a full-scale organizational revolution, one few are prepared to manage gracefully.
And yet, not everyone suffers equally under this revolution. Younger, digitally native firms appear to sidestep much of this disruption. They don’t have decades of rigidly enforced workflows to dismantle.
According to McElheran’s findings, startups and younger enterprises recover productivity faster and more decisively. Firms built around agile principles and flexible digital architectures find the leap into AI integration manageable, even advantageous.
Moreover, startups pursuing aggressive, growth-oriented strategies adapt to AI far better than peers committed to conservative cost-optimization or strict quality-centric approaches.
This sharp divergence between old and young, between agile and entrenched, opens an important conversation: What can established, older firms realistically do to protect and preserve good management practices during the AI transition?
The takeaway? Firms need to consciously guard the integrity of their management practices, keeping their management teams robust and proactive throughout the upheaval.
McElheran leaves open critical narrative space here, space worth thoughtfully exploring.
Might companies explicitly build transitional management teams dedicated not just to implementing AI, but to preserving proven methods amid disruption?
Could targeted training in managing complex change help firms preserve Lean and Kaizen methodologies as complementary rather than disposable frameworks?
Another subtlety arises from McElheran’s data on multi-unit firms. Large, geographically diverse manufacturers show intriguing internal spillover effects, where successful AI adoption at one site boosts overall productivity across the firm.
Could lessons learned from these internal spillovers inform broader strategies to mitigate disruption across different management functions, perhaps by piloting AI in smaller organizational units first, reducing immediate shocks to the larger firm?
The point is that AI demands far more than technological investment. It calls for sophisticated management preservation strategies, many of which are not well understood or widely practiced today.
Companies run the risk of too quickly discarding core management practices as they chase digital transformation, believing new tools will inherently streamline operations. But the truth, hidden in plain sight within McElheran’s dataset, is that organizations thrive not by dismantling proven management principles, but by carefully integrating new technology around them.