The AI is actually a bit smarter than that (or at least it tries to be a bit smarter than that). There are special situations that make the AI more likely to chop for instant productivity; such as when they're building a wonder, or when their preparing for total war. And even when they aren't chopping for productivity, the AI still doesn't "randomly chop vs let it stay". Instead, the AI weighs up the yield rates of all possible plot improvements, and picks the one that it wants the most. It also takes into account the health bonus from the forests. In fact, there is nothing explicitly random in the way the AI chooses it's improvements.
However, the AI does have some serious flaws in its worker management. In particular, when there is an improvement that isn't connected, the AI currently just rushes to it with whichever worker it happens to cycle to first - regardless of how far away that worker is, or how close other workers are, or how important the worker's other jobs might be... This can cause some significant inefficiencies in how workers are used.
As I mentioned in a recent post, the worker AI is something that I'd like to improve soon.
In the mean time, I'm pretty much ready to release a minor update. I'm mostly just waiting until I've actually play tested it first. (I'm still pretty busy at the moment).
Okay, so AI chopping is seemingly random instead of totally random, but the issue remains that it's bad decision-making.
As for the worker management, I don't really see the problem. You want to make the game more fun, don't you? And you have. In K-Mod, I'm willing to automate my workers much earlier than in stock BtS, because they're a lot smarter, but there's still a cost in efficiency. And that's fine. Nobody wants a game that plays itself. It just needs to take tedious micromanagement of our hands.
For that same reason I was so disappointed with the last change to the citizen management. You made it smarter in the sense that it produced better results when left completely alone, but that's not what I need. I need the ability to make my cities do what I want them to do without having to micro each citizen.
An example: I have a six city empire with a big, cottaged-up, bureaucracy capital with Oxford University and Globe Theater, and I'm in a teching phase of the game with my science slider way up. I look at one of my secondary cities. It has some food, some mines, some coast, a library and a university. I understand that in this situation a piece of gold is worth way more than a beaker, but the citizen manager doesn't. I want that city to produce gold to finance the research in my capital, where the multipliers are more favorable, so I have it "build wealth". And what does it do? It switches citizens away from mines and onto coast and scientist spots because it sees the +50% beaker modifier, leaving me to micro everything by hand or accept that building Oxford that early was a waste of hammers.
Another example of essentially the same problem: I have all the techs I need and am engaged in or preparing for a major war. I have my slider down because I'll need the money to finance and recover from the war I'm about to start. I look at my cities and what do I see? They're working scientists instead of commerce tiles, because those don't benefit form my libraries at the moment.
Even more potential for improvement I see in the build automation, which currently I never use. Correct me if I'm wrong, but automated cities currently can't whip more than one pop, right? And they can't accumulate whip anger either? If that were to change for both the AI and automated cities, I think that would be a pretty big deal.