i did some digging, and found out that 5.2 was fine tuned for the midwit office worker. it specifically reasons less, and acts more confident. i tested mistakes 5.2-thinking made with 5.1-thinking, and 5.1 spotted the mistakes immediately. thats the problem with big corporate ai.
Discussion
aren't you assuming too much competence from them? they don't know or can't train stuff specifically for office workers or anything, they don't have that much control