The answer is simple: yes, they will.
Don’t worry about that. Goblins will be sneaky and cowardly, zombies will be dumb, the guy at the top of the page will be burny. It would be strange if they weren’t.
I’m not entirely sure how this will work in detail, so I can’t tell you for sure yet. However, I can say that I won’t be happy with AI being the same for everything. Getting the character of the different races in there is important. Essential, in fact. For me, being able to add more character into the faction is the main benefit of an AI deck. Might as well turn that one up to 11 😉
Now there are a number of approaches I can try for this. Some involve more stats, most involve being clever with cards. (Well, the latter could be replaced with tables in the book, though I think that’s a second best option.)
At the core is likely to be a deck of cards to turn one/some of when it is the AI’s go. This will define its reaction(s). At least, this will be the core piece of working out its reaction(s). The clever bit is in how the different types of model interact with the different cards, and this is where the important design choices come in. It’s tempting to get complex with this, and so I’ll be resisting that urge and searching for a more elegant way to get varied behaviours into a simple process. Having just done zombie and Martian AIs for Deadzone and Mars Attacks I have an idea of what I want to do here. The DS AI is rather more complex than either of those systems in what it needs to simulate as they were both monocultural systems. Still, they were useful practice.
In order to decide which if the options fits best I’ll try several on the table. There’s no substitute for playing these things through. And, being AI, I can always be sure of an opponent when I need to test something 😉