A chatbot for problem solving
ELIZA was successful in part because it worked in a problem domain where lack of knowledge was no impediment. I find that in the lab we have a lot of conversations where one person just needs to talk through a problem while the other person nods sympathetically and prompts them with vague questions. Seemed like a chatbot could perform just as well.
This is stupid simple and has been more useful than it has any right to be.