AI in e-learning: practicing with realistic customer conversations

The learning platform, the ETD Academy, is used by employees of stores like MediaMarkt and Expert. They acquire product knowledge there, but also work on skills that you need on the shop floor. “Think, for example, of having a good sales conversation,” says Nick Broers, designer at Freshheads. “That's what we are using AI for now.” 

Experimenting and expanding with AI

OFED itself comes up with many ideas and initiatives for deploying AI. “That's great to see,” says Nick Broers. “We don't have all the answers yet, but we are experimenting and testing fully. The technical choice for n8n helps with that. It gives OFED plenty of freedom in managing the AI models and adjusting workflows. That makes it possible to experiment quickly and add new applications. Step by step, we are expanding the learning platform further and making it even more relevant for all kinds of roles in the electro-technical retail sector."

Skills you don't learn from theory

The reason to start with AI comes from a simple observation: you don't learn many skills from theory. “You need basic knowledge, but you mainly gain experience in practice,” says Nick. And exactly that practice is hard to simulate in a digital learning platform.

Traditional solutions fall short here. Without AI, you quickly end up in fixed scripts and selection menus. “Then you have to think of every possible reaction in advance,” says Nick. “That takes an enormous amount of time and never feels like a real conversation.” OFED was therefore looking for a way to practice realistic conversations. Without a fixed script, but within clear learning objectives. AI turned out to be a logical tool for that.

Learning from realistic shop floor scenarios

The most important application within the platform is an AI-driven conversation simulation. AI is good at interpreting open-ended answers and can respond to what someone actually says. This results in conversations that feel more natural and connect better with practice.

“With AI, someone can react just as they would in the store,” says Nick. “The conversation doesn't follow a fixed script, but within frameworks that fit the learning objective.” This balance is essential: freedom for the user, with guidance from the learning platform.

How the conversation simulator works

Employees have a conversation with a chatbot posing as a customer. Not a generic bot, but a persona with character, age, and behavior. “For example, you speak with Anna, who is 60 years old,” Nick explains. “The bot then also behaves like a 60-year-old woman. That is, of course, different from Floris of 23.”

That makes practicing fundamentally different from traditional learning methods. Instead of just testing knowledge, you practice behavior: listening, asking further questions, and advising. “Where podcasts, videos, and quizzes are fixed formats, an AI roleplay brings something that normally only happens on the shop floor. Employees have conversations that strongly resemble situations from their workday and learn by doing.”

Direct feedback follows after the conversation. AI rates the conversation on pre-set learning objectives, such as customer focus or questioning technique. The participant gets a score and concrete feedback. “For example, you hear that you asked good follow-up questions, but were too direct.” This feedback is immediately applicable. You start the conversation again and adjust your approach. This creates a learning loop in which practicing, reflecting, and improving follow each other immediately.

From typing to talking

A next step was voice interaction. Where users still type now, they will soon also be able to talk to the AI. “That changes the input immediately,” says Nick. “With text, you think longer. When you talk, you respond more spontaneously. That feels much more realistic.”

Speech also adds emotion to the conversation. Tone, tempo, and intonation become part of the interaction. This makes the conversation resemble a situation on the shop floor even more. Nick still sees a lot of potential here. “In the future, we hope to be able to look at how someone says something, not just what someone says.”

AI support for trainers

AI can also mean a lot for trainers. “Due to lack of time among trainers, we saw many training sessions on the platform without a quiz. That means knowledge is not being tested.” That is why we are exploring whether AI can help in creating quizzes. Trainers could, for example, upload content and indicate a learning objective, after which AI automatically generates a quiz. “The trainer then only needs to fine-tune it, and that makes adding quizzes a lot more accessible.”