It sounds like you are doing what I do sometimes - focusing on the finding a weakness in the defences of the antagonist rather than asking where the protagonist is in any way better or different.
Greek Tragedy might be of help here. They had a concept called Hamartia - a fatal flaw leading to the downfall of a tragic hero or heroine. Which sounds like what you are looking for. Often this flaw was because of rather than in spite of the strength.
In your case, the computer is unfeeling towards humanity and therefore lacks empathy. As long as the human units act in their own best interests, the AI wins. But, what if the humans and cyborgs act against their interests, or act randomly, or are utterly illogical?
What if the protagonists act from the emotions that the computer does not understand? Surely then it would be at the least surprised.
At heart, your story sounds like an exploration of the human condition, albeit in a really interesting sci-fi that I would love to read. Here your cyborg gang's best attribute is that they do have empathy, they do feel pain and they are not always logical.
Additionally, the AI is designed to manage the odds and minimise risk. The emotional human types would, on the other hand, be willing to gamble on unpredictable outcomes. They might be willing to try hopeless, million-to-one, shots.
Could, for example, this lead to an act of self-sacrifice that brings a faction, unexpectedly (for the cynical god-like AI) to side with the protagonists? Does it inspire self-sacrifice in others? Does the AI have any defense against such actions?
Can the protagonists randomise things so that neither side has any idea what will happen? Or at the very least leave the outcome somewhat open to chance?
The AI is built around understanding gain but it would have a limited understanding of the human ability to willingly lose, give up in the face of apparent victory, or stubborn refusal to take a path of apparent gain.
In fact, there was a Star Trek episode where Data won a game by refusing to take any path of advancement until his opponent rage quit. He points out "I did not win, he quit".
Could the emotional humans provoke an emotional act of frustration in the AI?
Could the cyborgs create so much chaos that the AI has to devote increasing amounts of runtime to calculating what will happen whereas the human protagonists simply go with the flow and hope for the best? Could they, perhaps, do this so much that the AI is locked up with calculations and no longer paying attention somewhere?
There is lots you could do with this, I think. I hope I have given you a few ideas.
TL;DR: The "skills" required here are simply the ability to think in new ways about the differences between the protagonist and the antagonist and make that the "battlefield". Especially, any area where a character's strength might also be a weakness.