Automated tools are good at narrowing a large amount of roadway activity down to the events worth looking at. They are not the right thing to hand a decision to.
Replacing a camera network to start a safety program is rarely necessary and rarely affordable. What matters is what the existing infrastructure can actually support.
Detecting that something happened and establishing what it was are different problems. Collapsing them is how programs end up defending single observations.
A single mile or kilometre of road can involve a municipality, a transit organization, a police service and a contractor. A program has to fit all of them at once.