Microsoft AI Researchers Just Discovered Something That’s Going to Make Their Bosses Extremely Mad

AI automation is typically exactly what it sounds like: automating tasks — many of which were previously carried out by humans — in an attempt to boost productivity and efficiency, often in a prelude to laying off workers wholesale.

However, a new yet-to-be-peer-reviewed paper conducted by a group of Microsoft researchers and spotted by IT Pro found that today’s top AI systems remain eyebrow-raisingly weak at real-world workplace tasks. In fact, they often screw them up badly: the team studied frontier models including OpenAI’s GPT 5.4, Anthropic’s Claude Opus 4.6 and Google’s Gemini 3.1 Pro, and found that during complex assignments, those cutting edge bots corrupted an average of 25 percent of the content in documents. (Older models failed even more severely.)

The researchers concluded that, overall, these…