They also spawn in large numbers, but if you just spam units, it is pretty easy. What you need to do is just found 5-6 more cities, and you will easily win the score.
The big thing we really find out from this scenario is none of the really basic problems with the combat AI, even if fixed, will make it much better. This scenario has barbs as the opponents, so they:
1) Don't have the timidity issue of the AI civs;
2) Don't get distracted and change target every other turn;
3) Don't have settlers to steal;
4) Have upgraded units;
5) Get a proper unit mix from the spawning script; and
6) Get large numbers of units from spawning.
And the combat AI, on deity, is still completely useless. They take maybe one city state in 60 turns, never a AI Civ city (despite that one to the north east often getting to zero health). That means the combat AI is unfixable by modders with the current tools available. It likely means it can't be fixed economically by redoing the AI coding.
Maybe it could be fixed by more direct bonuses (to combat strength and promotions) and making the game easier for the AI to play#. Which everyone will hate. So I am calling it - Civ 6 will end its development cycle with combat AI / overall challenge at brokenly low levels (i.e. less than Civ4 and even Civ5).
# This is what stacks did in Civ4, and the 4 city limit in Civ5 really (because conquest wasn't so important, it mattered less the AI was no good at it).
* The encampment heavy AI Civs in the Viking scenario provide a bit more challenge (effectively game over by turn 30 of 100 on deity, but still it feels more interesting). Being on the defensive against coastal attacks seems to provide it some help. But maybe that is the Panzer General thing - A combat puzzle to attack is somewhat better for 1upt.