This kind of thing is often not as obvious as it sounds. I used to work for a company that did a similar trick for ambulances, and it was an amazing amount of work to try to persuade customers that this kind of statistical prediction could be useful.

Also, there are often other considerations; depressingly often there's political pressure, for example. One could easily imagine in this case that there might be pressure to show a stronger police presence in low-risk wealthy suburbs over the hot spots that might tend to be poorer.

From reading the article it seems like it's potentially used to not do that:

“But I’m looking at a map from last week and the whole assumption is that next week is like last week. The computer eliminates the bias that people have.”

I doubt it. It's probably that the software is looking at the map from last week, and the weeks before instead the person deciding that finding the map from last week is too much of a pain in the ass. Not to mention having to coordinate with the other officers so they don't all stake out the same parking lot. I'd bet the software is mostly doing things people would do them selves if they had the energy, patience, and organizational skills to do so.

That oversimplifies it a bit I think. The road gets burgled at the same time every week, so the police then patrol that road and the road doesn't get burgled. Which road is getting burgled instead? That's where it's nice to have already run all the stats.