I suppose one reason for that would be that "doesn't work" might be implying 100% failure rates, which isn't really what I mean. What I do mean is that dowsing is no better, overall, than random chance. People seem to use lots of methods that are known (scientifically) to be no better than chance, but people swear by them anyway. Chance can run in your favour, so that a method appears to be very successful, but every time it goes against you, you tend to forget it -- which makes it appear yet more successful.
So in that sense, that water companies use dowsing is hardly surprising. No amount of scientific evidence can persuade some people that ancient techniques aren't worth continuing.