Heron’s steam engine is described in his Pneumatics 2.11; the best discussion is Landels Engineering in the Ancient World pp.28-31.
‘A ball that spins round on a pivot when a cauldron is boiled’. Pressure builds in the cauldron, steam goes through a pipe into the sphere, and escapes through two bent tubes, causing sphere to revolve. Landels made a working reconstruction. He found that it works best at high speed (e.g. 1,500 rpm), and would have to be highly geared down to be useful. The worst problem is the sleeve joint, where the pipe from the cauldron enters the sphere. If this joint is too loose, steam escapes (inefficient); if it is too tight, there is lots of friction (inefficient). Efficiency could have been so low that for the machine to do the work of one man, it would have required the input of several men. In other words, it would have been a labour-using device rather than a labour-saving one.