AI Adoption Looks Widespread Until You Measure It

By Dr. Gleb Tsipursky 

A Monday morning earnings call ends, and the CFO opens a chat window to draft the board update in minutes. Down the hall, a frontline finance team still works the old way because access, training, and incentives never arrived. That split reality sits at the heart of the new internationally representative firm survey of almost 6,000 CFOs, CEOs, and senior executives across the United States, United Kingdom, Germany, and Australia, published in the National Bureau of Economic Review by Nicholas Bloom from Stanford University and other scholars.

The NBER paper, titled Firm Data on AI, reads like a progress report and a warning. Executives describe fast diffusion, limited realized impact so far, and large expected gains soon. They also forecast a smaller workforce, largely through slower hiring. The real story for leaders sits in the gap between ambition and daily use, plus the widening disconnect between executive expectations and employee beliefs.

Productivity Expectations Will Pressure Every Operating Model

Senior leaders in the survey expect AI to move the productivity needle in a way that dwarfs most operational initiatives. Across the four countries, executives forecast about 1.4% higher productivity over the next three years from AI adoption, with the United States at about 2.3% over the same horizon, a pace that translates to roughly 0.77 percentage points per year. Those results nearly double baseline growth when firms already plan around about 1% trend productivity.

Yet the paper also reports that realized impact over the past three years stayed modest, with an average realized productivity gain around 0.29% across firms. That “quiet period” matters because it explains why many organizations still treat AI as a pilot program rather than an operating system. Executives forecast acceleration because deployment patterns shifted sharply during 2025, including a jump in usage frequency and a drop in the share reporting zero use to about a quarter of respondents. The adoption signal is clear in the paper’s executive use measures, and it sets expectations that teams will soon face new usage standards. For instance, Accenture now tracks how often senior employees utilize artificial intelligence on a weekly basis, according to recent reports. The firm links these adoption metrics to promotion opportunities for veteran staff to ensure they embrace the growing role of technology in the workplace.

For professionals running functions, the best comparison comes from measured deployments rather than hopes. A large field study of a generative assistant in customer support showed about a 14% productivity lift on average, with the biggest gains among newer workers, a pattern documented in generative AI assistance. That result aligns with what many operators already sense: AI often standardizes and raises the floor before it raises the ceiling. Leaders who plan for broad productivity gains should therefore pair targets with workflow redesign, quality metrics, and role-based enablement, since a tool alone rarely changes an operating model.

Hiring Slowdowns Will Be The First Employment Effect

Executives in the survey predict a net employment decline of about 0.7% over the next three years as AI spreads, and the authors note that this implies roughly 2 million fewer jobs when applied to more than 250 million employed people across the four countries. That estimate matches what many companies already signal in practice: hiring plans move before layoffs do because hiring sits inside annual budgeting, headcount approvals, and backfill decisions.

This is where leadership teams can gain an advantage with clarity. If employment effects arrive through reduced hiring, then workforce planning becomes less about crisis management and more about precision: which roles receive augmentation, which roles consolidate, and which roles shift toward higher-value tasks. That approach also aligns with labor-market exposure research that frames AI as task transformation rather than job deletion. The global exposure estimate from the IMF puts nearly 40% of global employment in AI-exposed categories, emphasizing that complementarity and inequality risks travel together. Meanwhile, the ILO’s analysis finds the strongest exposure in clerical work and expects augmentation to dominate overall effects, detailed in GenAI exposure research.

For executives, the key operational move is to convert “reduced hiring” into an intentional design decision. That means defining where AI substitutes for routine throughput, where it improves decision quality, and where it opens capacity for growth. It also means protecting trust. Employees watch hiring freezes and interpret them as a signal about career paths. Leaders who connect hiring decisions to visible upskilling and internal mobility programs preserve engagement while capturing the productivity upside they forecast.

The Adoption Gap Creates Risk And Opportunity At The Same Time

The paper’s most surprising statistic feels mundane: executives report about 1.5 hours per week of AI use on average, and about 25% report zero use. Those numbers sit alongside a headline that around 70% of firms actively use AI, suggesting a two-speed economy inside the same organization. The adoption headline comes from firm AI usage, while the usage intensity points to a deeper truth: adoption without habit formation stays shallow.

This matters because the paper also finds a stark perception gap. Employees surveyed separately predict AI will increase employment by about 0.5% over the next three years, while executives predict a decline. That divergence appears in the paper’s employee expectations and raises a leadership challenge: execution requires shared belief about what work will look like. When employees expect expansion and leaders expect contraction, governance and change management become decisive.

External surveys show that disagreement is common. The OECD’s cross-country work on job quality and AI points to uneven adoption, mixed perceptions, and the need for worker involvement in deployment design, summarized in job quality evidence. At the macro level, many employers forecast churn: the World Economic Forum projects large job creation and displacement through 2030, with a net gain, while also warning that disruption touches a sizable share of roles, detailed in job disruption outlook. Finance-side estimates skew more aggressive on substitution, including a widely cited projection that AI could expose the equivalent of 300 million full-time jobs to automation, described in automation exposure estimate.

Senior leaders can turn this uncertainty into advantage by measuring reality faster than competitors. The winning play combines three disciplines: instrument adoption by role and workflow, link usage to quality and cycle-time outcomes, and convert productivity gains into a transparent talent agenda. When leaders treat AI as a capital allocation decision, they demand unit economics, control risk, and scale what works. When leaders treat AI as a culture project, they build shared capability and reduce fear. The survey suggests both are required, because expectations already run high and the adoption base still has room to grow.

The executive survey offers a clear message: leaders expect meaningful productivity gains and a smaller payroll footprint, even while recent realized impact stays limited. Those expectations will reshape budgets, performance targets, and hiring plans. Professionals who act early can shape the trajectory by moving from slogans to operating discipline, from scattered pilots to workflow ownership, and from headcount anxiety to skill-based mobility. The organizations that close the adoption gap first will capture the gains their leaders already forecast, and they will do it with a workforce that understands where it fits.

About the Author