New Research Shows AI Is Rapidly Closing the Gap With Human Performance
MIT researchers say AI capabilities are improving steadily across workplace tasks, with many text-based activities already achievable at high success rates.
New research from MIT FutureTech suggests that artificial intelligence is becoming increasingly capable across a broad range of workplace tasks, but the progress is happening more gradually than some predictions of sudden mass automation have suggested. Researchers analyzed more than 60,000 evaluations of AI responses across over 6,000 text-based workplace tasks to examine how quickly AI performance is improving.
The research found that AI systems can already complete roughly 50% to 75% of text-based tasks without requiring human edits, depending on the task and quality standard. Instead of showing sudden jumps in capability, the researchers found a relatively steady improvement across many types of work. This suggests that AI adoption may expand progressively as models become more reliable and businesses discover which tasks can be automated effectively.
Researchers estimate that by 2030, many text-based workplace tasks could reach success rates of approximately 88% to 97% at a minimally acceptable quality level. However, reaching near-perfect performance is expected to take considerably longer. The distinction is important because businesses may adopt AI well before systems become capable of performing every part of a job independently.
The findings also challenge the idea that AI will necessarily cause an immediate collapse in employment. Recent reporting similarly suggests that the large-scale job destruction predicted by some AI leaders has not yet appeared across the economy at the expected scale. Instead, AI is currently changing individual tasks within jobs, allowing workers to automate portions of their responsibilities while continuing to perform other activities themselves.
The research ultimately suggests that the AI transition could look more like a rising tide than a sudden wave. As models steadily improve, workers, companies and governments may have time to adapt through training, redesigned workflows and new forms of human-AI collaboration. The biggest impact may therefore come not from AI suddenly replacing entire professions, but from increasingly capable systems gradually taking over more individual tasks within them.
