It is comparing apples to oranges, which may count toward your 5-a-day just fine, but leaves your duck al 'orange tasting really weird. 
If you routinely use cheap or unskilled labour, not something I would ever recommend, then AI will usually outperform them pretty consistently. Especially if they are offshore labour with some language difficulties in understanding instructions comprehensively. And there will still be areas where the AI’s lack of logic make it inferior to a human intelligence that is not particularly intelligent, uneducated to the degree that AI has been, and has a language barrier.
For one of a billion clear examples: https://twitter.com/darth_na/status/1663151926437982211
It’s part of a long discussion I had just the other day (where I attempted to explain exactly how an AI can hallucinate, and from where it gets such completely wrong answers that no human, not even a pretty stupid, barely literate, non-native-speaker would fail.
It happens. Pretending otherwise would be pointless, petty, and setting oneself up to fail hugely (as well as look like a fool). AI is not AGI, and it makes many errors with 100% conviction. It makes them consistently enough that we can pretty much cause it to happen at will by asking things we know it will struggle with, and unlike a human, ChatGPT will not learn from its mistakes, but repeat the same ones over, and over, and over, until an update patch is made.
That’s not to say that AI is without its uses. But there is not one single situation in which ChatGPT produces results better than a skilled and knowledgeable human, the kind you’d actually want to employ if wage costs were no issue. ChatGPT can produce results faster, and certainly cheaper than a human, but you will always then lose at least some of the time you saved in needing to carefully check and verify its output. In worst cases, you may not only need to verify, but on finding flaws, and attempting many prompts, then have to do it by hand with human labour after all, meaning hours wasted.
Does that happen often? Well, with certain topics, yes. Naturally, future generations of AI will improve, but right now, the main advantage of AI over humans is simply the tirelessness and cost saving, and the main disadvantage is inaccuracy and quality.
Anyone that tells you otherwise is selling something dishonestly, or engaging in unwise and unaware wishful thinking.
I’ve been trying to shout this from the rooftops 
