It's common knowledge amongst this topic that the AI has a strong preference for delivering fainting blows when it can. This includes using priority moves on 1HP targets, but I'm specifically referring to OHKOs in this case. Aron is a tool for preventing the dangerous layering of attacks from multiple enemies that will prevent an important lead from doing its job. Eviolite Dusclops can survive a Crunch from Tyrantrum4, but not necessarily a Crunch (incurring a defense drop) followed by Exeggutor4's Wood Hammer. Aron mitigates this potential risk, that's all. Of course, because it also can learn Endeavor, plus weather moves, and functions impeccably under Trick Room, it has enough tools to avoid being deadweight beyond protecting its partner(s) from 2HKOs.
Would you even use Aron if it weren't for the ability of Endeavor to make Aron offensively useful? I thought the combo of Sturdy + Endeavor was what made Aron a huge draw in the Maison, but for some reason I'm reading what you wrote as "Aron helps to take KOs but it helps that it can learn Endeavor" which implies that Aron would almost find itself on teams even if it couldn't use Endeavor to its advantage. And given the lack of other Sturdy lv. 1 Pokémon elsewhere on Maison teams, that piques my interest.
Of course, I also do not consider Durant an exploit, but then I'm a firm believer in playing to win (when I'm not dicking around with my randoms, which are generally awful.) The AI is not bugged; it was just programmed that way, and it's up to us to deal with it. Durant and Aron have made it easy to deal with it while also giving us an advantage.
The Maison (and Frontier, Tower, and every other facility of this nature), as a whole, has an inherent advantage over the player. Many, actually.
1) You need to win [x] number of matches in a row, whereas every AI trainer has nothing to lose. This means they can play with very unreliable strats like OHKO moves and 70% accurate ones, moves that rely on status hax to be truly devastating, and items and moves like Quick Claw, Double Team, Lax Incense, without caring about what happens if they lose since each trainer is just another one in an endless line of trainers that are not penalized for losing in the same way as the player.
2) The AI has inherent knowledge of the Pokémon you are leading with. While in most battles careful checking of the trainer is enough to tell you the stats and set upon checking the database, there are some trainers, especially Legendary trainers (!) with whom it is not so cut-and-dried, and there is no way to know the true best move to make on the opposing Pokémon because of this factor. On the other hand, the AI's knowledge of your Pokémon means that they will immediately know the best move to make on your Pokémon (well, "best" as far as the AI programming is concerned).
2.5) They're not susceptible to human error when running these calcs. I lost battle 49 in Multi because I assumed even Scarf Terrakion couldn't be faster than max speed Greninja, and I didn't know which Terrakion it was anyway. Losses like Jumpman's "it was bright and I didn't see which move I had picked previously" or his near-loss in the Frontier because of misclicking a switch-in instead of just checking stats and other such losses are also not something the AI will ever be capable of doing.
3) The AI has an
enormous pool of Pokémon to choose from. You can use 3 (/4/6). These 3 (/4/6) need to be capable of beating as many of the opposing line-ups as possible, regardless of possible bad matchups, whereas a single AI trainer having a bad matchup against
you is hardly relevant since in the next match they may easily bring in the team that is perfectly designed to beat yours.
All of this considered, I don't think there's a single way to play the Maison that can be considered "exploitative" (beyond legitimate cheating). Despite all of this, the AI is still a computer, and doesn't have the insight that human players have with regards to - well -
itself. Leading with a Sturdy, Berry Juice, Endeavor Aron in Doubles or Triples is using a certain aspect of the AI's programming to your advantage. Using a Durant to pass Truant to the opposing Pokémon, and promptly setting up Cloyster, Drapion, or another sweeper, is using a certain aspect of the AI's programming to your advantage. Back in Gen 4, the Trick Scarf strategy was using a certain aspect of the AI's programming to your advantage. As far as I'm concerned, it's all the same, and it's absolutely no different than predicting that a Fire-type opponent will use a Fire attack on your Bellossom and switching to a Kingdra in anticipation. Some strategies may be more effective than others, ultimately, but I believe that since the AI has so many inherent advantages over you, that using the fact that it is a non-self-aware computer to
your advantage is perfectly fine and acceptable. It's true that teams based around things like Durant and Aron are so reliable as to almost be "broken" - but then, so is the goodstuffs team of Jumpman's Kangliscune or VaporeonIce's recent MegaMence/Aegislash/Chansey. It's just that turning AI quirks in your favor tends to turn out more immediate, visible results, for the most part, though I would be very hesitant to call these strategies "exploitative" since it's not like a goodstuffs team like Kangliscune
doesn't use aspects of the AI (like Kangaskhan's weaknesses to Fighting and status, Gliscor's weaknesses to Water and Ice, and the tendency of the AI to not switch out of a boosting Suicune) to its advantage. I suppose the sheer numbers of Durant and Aron teams on the leaderboard is a testament to those strategies' success... but then, MegaKhan and Aegislash in singles and Greninja and Talonflame in doubles/triples are similarly common, but I haven't heard anyone claim one of those are exploitative!!