A tech vendor helped the city score and peer-review data points, with artificial intelligence ultimately finding ways to save between $19.4 million and $28.4 million. Officials have already taken steps to cut costs.