MIT Study: AI Automation Is a Gradual Shift, Not a Sudden Shock

AI automation is integrating into the professional workplace through a steady, broad-based increase in capability rather than through abrupt, isolated breakthroughs. Research from the MIT Computer Science and Artificial Intelligence Laboratory indicates that the advance of AI is a continuous climb, a pattern that makes the technology’s impact more trackable for businesses and workers.

The findings are detailed in a March 2026 study titled Crashing Waves vs. Rising Tides: Preliminary Findings on AI Automation from Thousands of Worker Evaluations of Labor Market Tasks. The researchers, representing MIT FutureTech, propose that AI automation exists on a continuum between two distinct patterns: crashing waves and rising tides.

The Continuum of AI Automation

The crashing waves model describes a scenario where AI capabilities surge abruptly over small, specific sets of tasks, often after long periods of failure. In contrast, the rising tides model describes an increase in AI capabilities that is more continuous and broad-based, affecting a wide array of tasks simultaneously.

Based on an evaluation of over 3,000 broad-based tasks, the MIT researchers found little evidence of the crashing waves pattern. Instead, they found substantial evidence that rising tides are the primary form of AI automation. This suggests that AI performance is improving rapidly across a wide range of text-based work at once, rather than concentrating in narrow areas.

Methodology and Task Evaluation

To reach these conclusions, the research team analyzed more than 17,000 evaluations performed by workers. The tasks studied were derived from the U.S. Department of Labor O*NET categorization, specifically focusing on text-based work that is addressable by Large Language Models (LLMs). This included tasks involving communication, analysis and writing.

The study tracked how AI performance changed as the complexity of the tasks increased. The researchers observed that performance gains tended to occur across many different tasks at once. This broad lift suggests that AI is effectively raising the baseline of capability across the entire spectrum of text-based professional work.

Quantifying Performance Gains

The MIT study provided specific success rates for AI models performing tasks that typically require three to four hours of human effort. In 2024-Q2, AI models successfully completed these tasks with approximately a 50% success rate.

By 2025-Q3, the success rate for the same set of tasks increased to approximately 65%. This data supports the conclusion that AI capabilities are expanding in a steady, incremental fashion rather than through sudden leaps in proficiency.

Industry Implications

The distinction between crashing waves and rising tides has significant implications for how enterprises prepare for automation. Because the progress is broad and steady, it may be more visible to organizations, allowing them to track the evolution of the technology over time.

The crashing waves narrative was always more intuitive, but less accurate in enterprise settings. What we see instead is simultaneous, broad-based improvement that quietly raises the floor across everything at once.

Ayhan Sebin, Head of Product Incubation at IBM’s Software Innovation Lab

While the progress is gradual, the long-term effects on employment remain significant. Other reports indicate that most text-based tasks are moving toward a state of minimally sufficient automation, suggesting that the disruption to jobs is gradual but inevitable.

The study was authored by a team from MIT FutureTech, including Matthias Mertens, Neil Thompson, Adam Kuzee, Brittany S. Harris, Harry Lyu, Wensu Li, Jonathan Rosenfeld, Meiri Anto, and Martin Fleming.