A hundred years of receipts · 1926 → 2026

Two speeds.
One growing gap.

Technology keeps accelerating. Humans change at human speed.
The gap between those two rates decides what every rollout pays.

LAG is the gap between a technology arriving and the work around it reorganizing. It measures how long people and institutions take to catch up, not how fast the technology moved.

The two rates · years for each technology to reach mass adoption
Electricity
~46 yrs
Telephone
~35 yrs
Radio
~31 yrs
Television
~26 yrs
Mobile phones
~13 yrs
The web
~7 yrs
Generative AI
~2 months
Meanwhile, the human rate has not moved
One new habit
~2–5 months
Real skill
years
Culture change
multi-year

Generative AI is the first technology to spread faster than a single human habit can form. Every era below shows what happens while work waits to catch up.

LAG: how long work took to reorganize after the technology arrived · Adoption spans: US diffusion data via HBR, 2013 · Habit figures: Singh et al., 2024 · GenAI: UBS, 2023

The receipts · era by era
1926

Electricity finally pays out

Lag ≈ 40 years
  • The techElectric motors, commercially available from the 1880s
  • The stall1900: electricity ran under 5 percent of factory machinery. 1919: barely half. Manufacturing productivity growth crawled below 1 percent a year for two decades
  • The blockerFactories were built around one central steam shaft. Swapping the power source was cheap. Rebuilding the floor plan, the job roles, and the management logic was not
  • The turnOne motor per machine. Machines arranged by the flow of work, which made the assembly line possible. Productivity growth jumped above 5 percent a year in the 1920s
  • The cost40 years of near-zero payoff, and a generation of factory capital torn down and rebuilt
Paul David, The Dynamo and the Computer, AER, 1990 · Kendrick productivity series
1954–1987

The productivity paradox

Lag: 33 years and counting
  • The techMainframes in the 1950s. Minicomputers in the 1970s. PCs in the 1980s
  • The stallUS productivity growth fell from 2.8 percent a year before 1973 to about 1.5 percent after, exactly as computing investment exploded. US service firms spent an estimated 750 billion dollars on IT in the 1980s
  • The blockerComputers automated the old paper process step for step. Same hierarchies, same approval chains, same silos. Faster paperwork is not new work
  • The turnNot yet. Hardware doubled in power every two years. The org chart did not move once
  • The costA 20-year productivity slowdown that computing was supposed to end
Robert Solow, NYT, 1987 · US BLS productivity data · Stephen Roach, Morgan Stanley Economics
1995–2004

The gap closes, again

Broken by redesign
  • The techNetworked PCs, ERP systems, and the commercial internet hitting critical mass inside firms
  • The stallIdentical technology, wildly different returns. IT spending alone predicted almost nothing about which firms benefited
  • The blockerIn lagging firms, decisions stayed centralized, pay stayed tied to old processes, and retraining was optional
  • The turnWinners moved decision rights down, rebuilt processes, and retrained at scale. US productivity growth doubled, from about 1.5 to 3 percent a year
  • The costUp to 9 dollars of organizational investment riding on every 1 dollar of hardware, by market-value estimates
Brynjolfsson & Hitt, 2000 · Brynjolfsson, Hitt & Yang, 2002
2011–2019

Same wall, faster clock

Lag: program cycles
  • The techCloud, mobile, SaaS. Deployment became nearly instant
  • The stallFewer than 30 percent of digital transformations met their goals. Only 16 percent produced improvement that lasted
  • The blockerTransformation ran as an IT project: new tools, old workflows. Incentives and role definitions never changed
  • The turnThe successful minority rewired roles, incentives, and operating rhythms alongside the stack
  • The costOf 1.3 trillion dollars spent on transformation in 2018, roughly 900 billion missed the mark
McKinsey Global Survey, 2018 · Tabrizi et al., Harvard Business Review, 2019
2022–2026

AI, the fastest case yet

Lag: quarters. Same wall.
  • The techGenerative AI. 100 million users in two months, the fastest-spreading consumer application on record at the time
  • The stall95 percent of enterprise pilots show no measurable P&L impact. 42 percent of companies abandoned most of their AI initiatives before they ever reached production, up from 17 percent the year before
  • The blockerPilots bolt onto existing workflows with no role redesign. Learning happens on top of full workloads. 90 percent of workers use AI personally while about 40 percent of firms provide it, so real usage stays in the shadows
  • The turnStill being written. The failure is organizational, not technical. The winners are redesigning work, not buying better models
  • The costAn estimated 644 billion dollars of generative AI spending in 2025, against a 95 percent no-impact rate
S&P Global Market Intelligence, Voice of the Enterprise: AI & Machine Learning, Use Cases 2025
Back to the framework Back to the Biblioteca