We had the exact same problem (also around 4 years ago). The large companies which have the volume of tweets to need a service like this don't care too much if there is fancy machine learning behind it, but do care if there's sufficient marketing/brand knowledge to deliver the exact things they want to see in the right format.

The two angles that worked out best were finding larger organisations which wanted to deliver this kind of "cut through the noise" information to consumers (e.g. picking out reviews), and delivering broader "market intelligence" (e.g. showing what's being tweeted about car buying in a certain area for a car dealership, vs. just monitoring that dealership's name/accounts). Good luck.