The jobs involved simple jobs of which represented less than 6% of the work they do, showed a 70% improvement. So it was a small set of curated tasks given where they found that the use of AI was better with a human in the loop.
In my experience one of the challenges has been that when humans are using AI they may complete a task quicker, however in QA I've found the results to often be incorrect or require more QA and I'm not talking about coding jobs. I am assuming that the tasks being marked completed quicker, were also QA'd and marked as completed appropriately.
phyzix5761•1h ago