On Sunday, December 8th, the show 60 Minutes featured a story about how AI was being used in the classroom to help teachers and students learn with the help of AI tutors (via Open AI).  The AI is used with the curriculum to help students move through a series of question-and-answer interactions that help them use critical thinking to solve math and science problems. The system developed by Sal Khan, and Open AI, provides a dashboard for the teachers to use to follow the flow of conversations with the student and their AI tutor. The teacher can focus on students’ identified trouble areas vs. trying to spread their attention and feedback with less data to tailor their feedback for more immediate impact and retention.   

This parallels with Dr. Benjamin Bloom’s 2 Sigma theory on the best way to teach a student a new skill/knowledge and retain it longer.

Similar to the AI Tutor in the classroom, Benjamin Bloom analyzed teacher-to-many, peer-to-peer, and tutor-to-student methods of learning. The results of the tutor-to-student had the highest results. At the time his study was done, there was no scalable and cost-effective way to provide this in the classroom.

AI has allowed these tutors to be deployed across a large population base and respond at lightning speed.

Our ACES simulator does the same thing. We use AI to play the role of the built-in tutor or coach. Managers have access to very detailed analytics, just like the teachers have, to focus on the tasks an employee may be struggling with in the simulator as a roadmap to use as a way to provide the most relevant coaching and feedback. This allows the manager or trainer to be more Proactive than Reactive. 

The report on 60 Minutes also addresses the risks or negatives of using generative AI in the classroom. As with a lot of these systems, there needs to be some form of guardrails in place.

For this reason, our ACES platform uses a combination of rules-based AI and NLP. When using our simulator in a call center, you want to design the simulations based on best practices for any given interaction in that call center. When using generative tools to help you “build” the learning activity, this runs the risk of not having consistent learning for each agent that challenges them to perform the proper tasks following your best practices. Call centers know what type of calls they have and the types of situations employees will be tasked with. Simply pulling examples from the hours of recorded calls can be used to create the scenarios that agents will need to learn how to respond to.

We all know these new generative tools are very attractive and can reduce our workloads tremendously. We still need to be mindful of when we use them and how. I believe we still have a ways to go before we can blindly deploy LLM (large language models) at scale but there is no doubt they are a technology disrupter.