U.S. Productivity Data Shows Mixed Signals With AI Adoption

Two measures of productivity are moving in opposite directions during 2026. Labor productivity, which tracks output per unit of labor, has posted solid gains, while total factor productivity has shown little growth since a post-pandemic increase. The divergence has prompted comparisons to the period when businesses first adopted internet technology.

A research brief published Tuesday by the Federal Reserve Bank of San Francisco examined data from the mid-1990s and concluded that measured productivity gains often lag behind the introduction of new tools.

Studies cited in the brief found that employees using AI completed tasks more quickly in some cases. One analysis by the London School of Economics estimated that certain workers could save up to a full workday per week. However, other research indicated that time saved was frequently redirected to additional tasks rather than reducing total hours worked.

A Harvard Business Review study of 200 employees at a U.S. technology company reported fewer breaks and higher reported burnout among frequent AI users.

The San Francisco Fed researchers noted that similar patterns occurred during the early internet era. Labor productivity accelerated starting around mid-1996, yet broader efficiency gains did not register in aggregate data until several years later.

The brief stated that determining whether a sustained productivity increase has begun is difficult without hindsight. It concluded that current conditions may resemble the initial phase of the 1990s productivity expansion.