I'm probably going to sound like an armchair developer here, but this isn't really that impressive. Pretty much all they did here was link AI chat responses via speech recognition to an extraordinarily shitty voice and animation system in pretty much static scenes.
It's cool that this might be a thing in the near future, having actual conversations with NPCs instead of just predetermined options and responses, but this needs a lot of work to look and feel less cringe-inducing, and a lot of testing so that the AI doesn't end up giving misleading or false information to players.