YESSS. And IIRC, another positive was that if a player had a certain role, it affected their Importance (IMP) rating, which I understand is a quantitative measure of how a team values a player. I don't recall if Madden currently uses that rating now though.
As it stands, if there were two relatively equal 75 OVR QBs, with one being 21 years old and the other being 30 years old, the AI would probably pick-up the younger QB because they're probably cheaper and because I assume the AI has at least been built programmed to prefer younger players.
But if that 30 year old QB had a Mentor role, maybe the Mentor role can boost the IMP rating and be more valued to a team in free agency.
If they returned to that player archetype/badge system, they could really really provide some flavors to how AI coaches and GMs build their roster. I mean, just take QBs for an example. Obviously, all teams would want to have the Franchise QB... but then from there, other teams prefer to only carry 2 QBs (1 franchise QB and 1 serviceable back-up), while others like 3 (1 franchise QB, 1 mentor, 1 QB of the future).
They've kind of got a good thing going on with making scheme-specific teams. You'll see more Zone CBs going to Tampa 2 zone coverage style defenses because of it. The Frank Gores of the world would be less likely to sign with a Run N Gun team than a power running team. But I think the player archetype/badge/weapon role system would be a great way to provide flavor and preferences to AI roster decisions.
In all honesty, the IMP rating and how a team values players shouldn't just be on how well they fit into their system... it should take past production into account. It should take their current ratings and future production into account. It should take the current roster into account. It should take age into account.
Past Production can be expressed through Player Archetypes. If a CB balls out enough to earn a Shutdown Cornerback role, that should increase their value.
Players should progress/regress in an organic way through in-season training and based on who their teammates/coaches are. A player shouldn't get better because he had a good year. It's the opposite--a player has a good year because he got better.
I'm pretty sure that the AI can be programmed to take all of this into consideration.