A new study from the National Bureau of Economic Research paints a sobering picture of enterprise AI adoption. After three years of widespread AI deployment, 89% of executives report no productivity gains from AI implementation at their own firms, challenging the prevailing narrative of AI-driven efficiency improvements.
The Findings
The NBER study surveyed executives about their organizations’ AI adoption experiences over the past three years. The results reveal a stark disconnect between the hype surrounding AI technology and measurable business outcomes:
- 89% reported no productivity gains at their own firms
- More than 90% reported no effect on employment from AI adoption
- The survey covered multiple industries and company sizes
These findings stand in contrast to data from platforms like Linear, which shows AI agents tripling pull request volumes — yet development time has actually risen, not fallen.
Contextualizing the Results
The NBER study’s results paint a nuanced picture when combined with other productivity data. While individual developers using AI tools may see efficiency gains at the task level, these improvements don’t necessarily translate to organizational-level productivity metrics. Several factors may explain this:
Implementation challenges: Many companies struggle to integrate AI effectively into existing workflows. The technology itself may work, but organizational friction prevents realization of potential benefits.
Measurement difficulties: Productivity gains from AI can be difficult to isolate and measure, especially in knowledge work where outputs are not easily quantifiable.
Overhyped expectations: The AI industry has promoted aggressive claims about productivity transformation that may not align with realistic adoption timelines.
Industry Implications
The study raises important questions about the ROI of enterprise AI investments. Companies have invested billions in AI infrastructure, talent, and tools — but the NBER data suggests these investments are not yet yielding the promised productivity improvements.
This comes at a time when AI agent deployments are accelerating. Companies are deploying more autonomous AI systems, yet the productivity case remains unproven at the executive level.
The findings suggest the AI industry may need to recalibrate expectations around enterprise adoption timelines and focus on more measurable, incremental improvements rather than transformative promises.