We generally strap the probe to the side of the fermentation vessel and then put bubble wrap and duct tape over it to insulate it from ambient a bit. It is not as direct of a beer temperature as a thermowell would provide, but works well enough for us.
Controlling ambient temperature only is problematic because the beer will take much longer to reach the desired temperature if the temperature differential is small, and changes in beer temp (for instance from heat produced by the yeast) will not be easily detectable as changes in ambient temp. One good thing about slower temperature changes though is they would eliminate or minimize temperature overshoots.
Probably the best approach would be to measure both temps and control the differential. This allows you to more easily predict where the beer will end up, but also allow you to crank up the heating/cooling for fast response. I believe BrewPi uses a temperature differential approach where they measure both the air and the beer to get tighter control. Our system is currently not quite that smart, but we would love to look into supporting something like that, and welcome code contributions if anyone wanted to take up the challenge.