I'm an infra/admin jack of all trades with a comp sci degree and have been a hobby programmer for 30 years. I have a Lichess rating of 1000 on a good day.
We tried doing a chess bot competition (open book, use AI to program it, pull in opening books, end game tables, whatever, free for all) and I absolutely stomped him, but I've only beat him in real life over the board twice in 20 years.
He will beat 99% of random players in real life, and I will beat maybe 20%.
I'm not sure what I'm trying to say, but it seems to me that maybe domain knowledge isn't everything anymore? Or the domain itself has shifted?
For example, I used to do integrations for sports betting sites. AI is going to help with the basics, like understanding the default puck line is 1.5 in hockey. AI is not going to realize that Bet365 changes their API endpoints for each season, so you need to be ready to fetch the updated ones before the new season starts, whereas most other sportbooks have consistent endpoints that you don't need to keep updating.
How much domain knowledge is actually unavailable to AI is going to vary by domain, as will the value of that. Chess is probably one extreme, where all knowledge is public, whereas something like military R&D might be the other extreme where domain knowledge is tightly guarded.
But OP wasn't talking about solving optimalization problems, but understanding the rules of a business domain.