Cloud technology was the closest thing leaders had to a completed experiment in enterprise-wide technology change. It ran through every function rather than just staying inside IT, and many organizations are nearly on the other side: 84.7% of large European enterprises used paid cloud services in 2025 (Eurostat, 2026).
AI adoption now sits where cloud technology sat a decade ago. US Census Bureau research shows that 18% of American firms used AI in a business function by early 2026, and that those firms account for 32% of employment (US Census Bureau, 2026). The decisions that determined who gained from cloud technology are being taken again now, on a shorter clock.
The Gains Went to the Organizations That Redesigned
The industry’s name for the standard first move to cloud technology tells the story: lift and shift. The organizations that benefited from the transition changed how they worked. Those that moved their systems onto cloud infrastructure but left the original operating model in place gained less from the same technology.
DeStefano, Kneller and Timmis found that cloud adoption raised revenue, employment, and productivity for younger firms, while established firms saw no scale effects and weaker productivity gains (Review of Economics and Statistics, 2025).
The Stanford economist Paul David documented the same move a century earlier. Factory owners first responded to electrification by swapping their steam engines for electric motors and keeping everything else where it stood. Productivity gains took four decades to arrive, counting from the first central power station, because factory owners had to redesign the whole factory around what the new power source made possible; swapping the engine wasn’t enough (David, 1990).
The Returns Dip Before They Climb
The economists Erik Brynjolfsson, Daniel Rock, and Chad Syverson demonstrated why that gap opens. General-purpose technologies demand large complementary investments in process redesign, new business models, and human capital. Firms make those investments long before they harvest them, so measured performance follows a J-curve: it falls first and rises later (Brynjolfsson, Rock and Syverson, 2021).
AI is already tracing the same curve. Census Bureau researchers, studying tens of thousands of American manufacturers, found causal evidence that AI adoption cut productivity and profitability in the short run before longer-term gains arrived (McElheran et al., 2025). Boards that mistake the early dip for failure abandon the initiatives that were about to pay off.
Discipline Decides How Much You Dip
The same Census research shows where those losses concentrated. Older, established firms absorbed the deepest short-run declines, and one behavior explains roughly a third of them. As they adopted, these firms let go of the structured management practices that hold their organizational capital in place, particularly around KPI review and structured targets. The practices that made them productive were the first to slip.
That is the part leaders control. The accumulated know-how in an established firm sits in those routines, and adoption puts pressure on exactly the routines that hold it.
As Kristina McElheran, one of the study’s lead authors, puts it, “AI isn’t plug-and-play”; it demands systemic change, and established firms have to manage the friction that change creates. Firms that paired adoption with a growth-oriented strategy flattened the dip.
That finding cuts against a familiar instinct in transformation: loosen the operating system so the organization can move fast. The evidence says otherwise. Firms that held their management disciplines through the disruption lost less productivity to it.
What This Means for Leaders
The challenge that this raises for leaders is that an organization with thirty years of hard-won judgment about how its work gets done has more to protect and more to unlearn than a younger competitor. Being well run offers less protection than leaders might expect, because the practices that make a company well run are the ones adoption tends to disturb.
Leadership teams need to apply the lessons of the last migration at the start of this one, not after. Leaders should:
The bottom line: cloud adoption proved that access to a technology is universal and advantage from it is earned. AI will prove it faster.
Sources
We use data collected by cookies to analyze traffic on this website. By clicking “Yes,” you agree to our use of cookies as described in our Privacy Policy.