How I Built a Conversational AI Agent to Teach Customer Interviewing—In Under 10 Minutes

💡 Teaching with Tech: From Classroom Pain Points to Conversational Agents

As an educator in entrepreneurship, one of the recurring challenges I see among students is mastering the art of the customer interview. Asking the right questions to truly understand user pain points isn’t easy—even for seasoned professionals.

So I asked myself: what if a conversational agent could help guide them through the process?

In this post, I’ll walk you through how I used ElevenLabs’ conversational AI tools, combined with ChatGPT and a curated knowledge base, to create a bilingual support agent that trains students in customer interviewing.

In this video, you can also find a full demonstration and step by step description:

🚀 The Use Case: Helping Students Explore Energy Bill Pain Points

To demonstrate the power of this agent, I started with a realistic scenario: exploring the pain points of people struggling to pay energy bills in Germany. The target group? Individuals living alone in rented apartments.

The conversation with the agent goes like this:

Student: “Hi, can you help me prepare for a customer interview?” Agent: “Of course! Are you trying to validate a specific problem or just explore pain points?” Student: “I want to understand the struggles of people paying energy bills.” Agent: “Great. Start by asking them to describe their daily routines or last experience with paying an energy bill…”

It quickly moves into contextual, actionable advice—including smart question templates like:

• “Can you walk me through your last experience paying an energy bill?”

• “What happens in a typical day in your apartment?”

• “What factors made it difficult to pay?”

This interaction is dynamic, language-aware (German/English), and personalized to a student’s goals. All built using simple tools available to anyone.

🛠️ Behind the Scenes: How I Built It Step-by-Step

Here’s how I pulled it off:

1. Start in ElevenLabs

Once logged into the free version of ElevenLabs’ app, I navigated to the Conversational AI section and selected the “Support Agent” template as a starting point.

2. Customize the Agent Persona via ChatGPT

Rather than using the generic template, I asked ChatGPT:

“Can you create a persona for an agent that helps entrepreneurship students learn to do customer interviews?”

I provided a copy of the default system prompt as context, and ChatGPT returned a detailed, pedagogically sound personality tailored to my use case. I then copied this new prompt into the ElevenLabs system prompt field to override the generic behavior.

3. Make It Multilingual

Next, I set the language settings to support both English and German—perfect for my student cohort.

4. Embed Domain Knowledge

To make the agent even more effective, I added a knowledge base: a PDF generated using ChatGPT’s deep research feature titled “Best Practices for Customer Interviews.” I uploaded it directly into the agent’s backend so it could cite and reference relevant tips during the conversation.

That’s it. No scripts. No APIs. Just a few strategic prompts and uploads.

🔄 Takeaways: Want to Try This Yourself?

If you're in education, entrepreneurship, or consulting, here's how to get started:

• ✅ Use ChatGPT to write your agent's personality prompt based on your learners' needs.

• ✅ Use ElevenLabs to deploy the conversational interface.

• ✅ Feed it a focused knowledge base to ground its responses in best practice.

The process takes under 10 minutes once you know what you're doing.