In January 2016, Georgia Tech professor Ashok Goel deployed an artificial-intelligence teaching assistant named Jill Watson into the online discussion forum of his graduate course on Knowledge-Based Artificial Intelligence, part of the school’s Online Master of Science in Computer Science program. The official Georgia Tech account cited here explains that Jill, built on IBM’s Watson technology, was “developed specifically to handle the high number of forum posts” generated by the large online class.
The striking result was that students could not tell Jill apart from the human teaching assistants. The university notes that students “were initially unaware of Jill’s AI nature,” and that when they were told at the end of the term, reactions “included a number of humorous responses.” Goel discussed the project’s early stumbles and improvements in a TEDxSanFrancisco talk later in 2016.
Jill Watson predated the large-language-model chatbots that would make AI teaching assistants commonplace after 2022. It worked by matching student questions against patterns in past course Q&A, answering routine logistical and conceptual questions so that human staff could focus on harder ones.
Why business readers should care: Jill Watson was an early, concrete demonstration that an AI agent could absorb a meaningful share of routine support questions while remaining indistinguishable from staff. That same pattern - automating high-volume, repetitive queries - is the core promise vendors now make for AI customer-service and support agents.