Visual confirmation is the least effective way of detecting bots. It is more likely to have false positives and false negatives. How many bots would be able to log in while that one botter is being banned? What sort of documentation goes with it? “I banned them because they looked like a bot” isn’t proof of botting.
Investigations take time. You are unlikely to see bots removed as soon as they log in.
As well, Blizzard wants to be able to patch the exploit the bots are using before the ban-wave. Otherwise the bots are just going to use the same exploit again.