The fission of one single atom releases an incomprehensibly tiny amount of energy. It's probably not always exactly the same amount, because there's more than one way a Uranium atom can split, but it probably all fits in to a narrow range.
Anyway, power is energy per unit time, so to control the power output, you need to control how many fission events happen in each second.
Control rods in a nuclear reactor have a somewhat delayed effect on the overall effective neutron multiplication factor of the fuel mass. That is, the average number of additional fission events that are caused by each fission event. If the factor (a.k.a., "$k$") is greater than 1, the power will increase (and keep increasing until some BAD THING happens), if $k<1$ then the power level will decrease, and if $k=1$ the power output should, in theory, stay constant.
It's tricky because the age of the fuel, and the temperature of the coolant, the presence (or not) of steam bubbles, and many other things also affect the $k$ value. So, keeping a contstant power level is kind of like balancing on a unicycle. You have to keep adjusting it all the time: can't stop an rest.