Here is the link. It's quite interesting, but what especially caught me eye was this:
With Fallen Enchantress: Legendary Heroes, Wardell noticed he was seeing a lot of complaints about AI player despite the AI playing what he thought was a pretty strong game. What he realized was that, because the AI didn’t cheat and acted rationally based only on what it knew, players were reading far more mistakes into its behavior than it was actually making. It was a critical moment for his understanding about how players regard AI.
“It's having the AI communicate with the player more, having dialogues come up where the AI actually is explaining in a conversational way, ‘I see what you're doing there, and I'm going to go and do this,’” Wardell says. “By communicating more with the player, even if the AI is making a bad strategic decision, the player can at least understand why they're doing it.”