For me, the current (beta) scheduling options are an improvement, and works well for athletes with time limits for training, but still somewhat complex and iterative to use for crafting a routine. Choosing time-limit options does not always lead to the results I would expect, so I have to go back and try again - or just move workouts around each week. Moving workouts around is a workable fallback, but I wanted to throw a couple of idea out there for discussion.
One of my favorite calendar / scheduling implementations is TrainerRoad - for each phase of training (base, build, specialty) it gives you a weekly template of workouts (duration, intensity, sport by workout by day for a week) and then the user can drag and drop to create a good schedule. This schedule then applies throughout that phase. For athletica, maybe that could be implemented for blocks as opposed to phases (I haven’t been with Athletica long enough to know if the blocks lend themselves to this, but my plan is just a few different blocks).
This works for me because I then know the pattern of my weeks, which makes planning much easier. And I don’t have to go in and fiddle with each week, only the weeks that are somehow exceptional.
Another option I’ve seen that works well is choosing the day of the week for long workouts (it can be just a check box or drop down).
Or - if ai is really the way to go - maybe the ai could learn from the user rearranging workouts during the week? For example, if I edit the first few weeks of my schedule in a consistent way. the ai would then “learn” my schedule?