Survey: 80% of Companies Report No AI Productivity Gains
Despite billions invested in artificial intelligence, over 80% of companies report seeing no measurable productivity gains from their AI deployments, according to a new industry survey. The finding highlights a significant disconnect between AI's theoretical potential and its realized business value. Common obstacles cited include integration complexity, workflow misalignment, and a lack of user training.
- The survey by the National Bureau of Economic Research polled almost 6,000 executives across the United States, United Kingdom, Germany, and Australia. While 69% of their businesses use some form of AI, 89% reported no change in productivity as measured by sales volume per employee over the past three years. - A significant gap exists between leadership expectations and the current reality of AI's impact. Executives forecast that AI will increase productivity by 1.4% and decrease employment by 0.7% over the next three years, while employees anticipate a 0.5% increase in jobs. - Common uses for AI in the surveyed companies include text generation with large language models, creating visual content, and using machine learning for data processing. However, even among executives who use AI, the average time spent using these tools is only 1.5 hours per week. - Experts point to a modern "productivity paradox," similar to the introduction of computers in the 1980s, where technology is adopted faster than it can be effectively integrated into workflows. Key obstacles include poor data quality, the complexity of integrating AI with legacy systems, and a shortage of skilled AI talent. - The size of the company appears to be a factor in realizing AI benefits, with medium and large firms reporting stronger productivity gains from AI than smaller companies. This raises concerns about a widening competitive gap. - A separate MIT study highlights the "jagged technological frontier" of AI, where its capabilities have sharp, unpredictable limits. When used for tasks within its capabilities, generative AI can improve a skilled worker's performance by nearly 40%; however, performance drops by 19% when it is used for tasks outside that boundary. - Many companies are struggling to move AI projects from the pilot stage to full production. A 2025 MIT report noted that 95% of generative AI pilots result in no tangible business impact, often because the systems don't learn from feedback or integrate well into daily operations. - Some research suggests that while AI can speed up the initial stages of a task, significant human expertise is still required for refinement and completion. One analysis posits that AI can get a task to about 60% completion quickly, but the final, most valuable stages still depend on human skills and context that AI lacks.
