The problem is not randomness per se, but the fact that you can not control it in any way. If players are familiar with the game, games are decided randomly and you can not do anything about. For instance, it is usually useful to play green cards in the beginning, as green cards will usually facilitate your progress later on. However, if there are good green tokens available, both players will go for green cards and it is random who will get these tokens first, as there is no really way to facilitate your own progress towards them or prevent your opponents from getting them.
Another example is that getting 3 of the same resource to construct your wonder stage is often a bottleneck - if one player is able to draw golds at that point of the game, that is a huge advantage, which you can not affect in any way. Having green tokens which allow you to draw extra cards (or Engineering) can also help, but remember that it is random who will get them in the beginning, so it is randomness^2.
The worst is, of course, that very often later in the game, all players end up just drawing from the hidden deck, because visible cards are not useful (one reason for that is that military is not strong enough in the game, but often this also happens due to resources nobody needs). In the beginning of the game, this is often alleviated by the fact that many wonders force you take face-up cards, but later in the game there is really no way around it. So it is like playing a Russian roulette.
There are many similar games which are also quite random, but where you are actually able to control randomness. Happy City is an example. There you can go for cards, which give you less points right now, but allow you to stress less about which cards are available later on. Similarly, many specific buildings require taking lesser cards now to stress less about which cards are available later. The most important thing is that there are actually sensible and balanced trade-offs in a sense that a player who went for a specific special building and a player who went for direct scoring are likely to have a close game. Also in regular 7 Wonders (although it is actually not a very similar game), there is clear profiling, so one player can go green, another red etc. In 7 Wonders Architects there is nothing like that: there is one path to victory and it is randomly determined - if you, say, completely randomly missed green cards in the beginning of a game with good green tokens, it is gg: working on your wonder or drawing blues is not going to save you. It would be ok if something like that would only be occasional, but literally every game with 200+ ELO players is something like I described above.
Another example is that getting 3 of the same resource to construct your wonder stage is often a bottleneck - if one player is able to draw golds at that point of the game, that is a huge advantage, which you can not affect in any way. Having green tokens which allow you to draw extra cards (or Engineering) can also help, but remember that it is random who will get them in the beginning, so it is randomness^2.
The worst is, of course, that very often later in the game, all players end up just drawing from the hidden deck, because visible cards are not useful (one reason for that is that military is not strong enough in the game, but often this also happens due to resources nobody needs). In the beginning of the game, this is often alleviated by the fact that many wonders force you take face-up cards, but later in the game there is really no way around it. So it is like playing a Russian roulette.
There are many similar games which are also quite random, but where you are actually able to control randomness. Happy City is an example. There you can go for cards, which give you less points right now, but allow you to stress less about which cards are available later on. Similarly, many specific buildings require taking lesser cards now to stress less about which cards are available later. The most important thing is that there are actually sensible and balanced trade-offs in a sense that a player who went for a specific special building and a player who went for direct scoring are likely to have a close game. Also in regular 7 Wonders (although it is actually not a very similar game), there is clear profiling, so one player can go green, another red etc. In 7 Wonders Architects there is nothing like that: there is one path to victory and it is randomly determined - if you, say, completely randomly missed green cards in the beginning of a game with good green tokens, it is gg: working on your wonder or drawing blues is not going to save you. It would be ok if something like that would only be occasional, but literally every game with 200+ ELO players is something like I described above.