AI Agents in Action: Duolingo’s Lily & Hyper-personalization
What’s the hardest part about learning a new language? For many people, it’s speaking. Speaking, along with writing, are called ‘productive’ language skills, vs listening and reading which are ‘receptive’ skills. According to Cambridge University, everyone finds receptive skills easier to master than productive skills. Practice improves all language skills, but practicing your productive skills is easier said than done. When you practice a receptive skill like listening, you get instant feedback if you’re doing well—if you hear something in a foreign language, you instantly know whether you understand it or not. But if you say something in a foreign language, how do you know whether you’re actually saying anything meaningful? You could ask a friend who speaks that language or a teacher for help, but they aren’t going to be available 24/7. What can you do?
Duolingo, the language learning app, has the answer: use an AI agent specifically built for language learners. Her name is Lily, and you can call her any time you want to practice your speaking skills with a partner who will automatically adjust her speaking style to match your proficiency level.
Lily is an example of an AI agent, which is a computer program designed to perform tasks on its own by understanding its surroundings and context, making decisions, and taking actions to reach specific goals. It learns from data and adjusts its behavior based on what it observes.
Lily understands her context: she knows she is speaking to a language learner and not a native speaker. She makes decisions about what level of complexity she needs to use, depending on the user’s own knowledge. And she takes actions to drive people to improve: according to Duolingo, Lily will interrupt people if they are taking too long to respond, and other small details that make her feel like a real life person. Lily even has a memory, and can read information from past conversations to use in the current conversation. This makes Lily perfect for hyper-personalized learning & development, as she adjusts herself to the user’s comfort level but also pushes them to keep improving.
The remarkable thing about Lily is that she has a personality: a moody teenage girl who will roll her eyes at boring topics and asks follow-up questions based on her own interests. AI agents with distinct personalities can be used to model various stakeholders in a business environment. Along with the agent’s ability to tailor its behavior to the way the user acts, this opens up multiple opportunities for hyper-personalization in a business environment. Consider the following use cases:
An AI agent that plays the role of angry customer, so that customer service representatives can practice handling customer complaints without the risk of upsetting an actual customer
You must give a presentation to a board member, who is notoriously critical and budget-minded. Practice with an AI agent version of him to prepare for questions he might ask. (As a bonus: if the agent can accept multimodal input—it can understand multiple forms of communication, including spoken, written, and visuals—then you can also get feedback on the content of your PowerPoint presentation itself, too!)
A new manager is uncertain about how to have a performance discussion with a poorly performing employee, and uses an AI agent to test different ways of holding the discussion. She can prepare for all the different ways the employee might react to make sure the discussion goes smoothly.
This is just the tip of the AI agents iceberg. Does your organization use AI agents for work? Can you think of any applications for AI agents in your workplace? If not, now might be the perfect time to explore their potential. With AI agents becoming more accessible and versatile, businesses of all sizes can benefit. The future of AI is already here, and those who embrace it will be better equipped to stay ahead in an ever-evolving digital landscape. So, start exploring how AI agents can be the next big step in your organization's journey towards hyper-personalization and innovation.