Hypothetical - One day AI could gain much more trust and be used for policy guidance

Let's say it's the year 2035. AI has become ubiquitous in our lives and has gained a lot of trust from consumers and the general public.

At some point, someone is going to have the audacity to start asking major questions and running simulations on what the most effective trade policies, tax policies, or even structure of government is. It's unlikely that the results would mirror our current systems/modes of handling things.

If/when powerful AIs start telling us that what we're doing is counter productive do we act on those conclusions or leave things how they are?

In theory this could apply across all political domains. Crime prevention, drug rehabilitation, public assistance, infrastructure spending, national defense etc.