We must indeed use rules as a matter of practical necessity, but it's just that: a matter of practical necessity. We can't model the entirety of our future lightcone in sufficient detail so we make generic rules like "do not lie" "do not murder" "don't violate the rights of others" which seem to be more likely to have good consequences than the opposite.

But the good consequences are still the thing we're striving for -- obeying rules is just a means to that end, and therefore can be replaced or overriden in particular contexts where the best consequences are known to be achievable differently...

A consequentialist is perhaps a bit scarier in the sense that you don't know if they'll stupidly break some significant rule by using bad judgment. But a deontologist that follows rules can likewise be scary in blindly obeying a rule which you were hoping them to break.

In the case of super-intelligent agents that shared my values, I'd hope them to be consequentialists. As intelligence of agent decreases, there's assurance in some limited type of deontology... "For the good of the tribe, do not murder even for the good of the tribe..."