Case Study: How Duolingo Built AI Into Its Product

Duolingo, the language-learning app, is one of the best-known examples of a consumer company putting AI at the center of its product. It launched AI-powered features for learners, used AI to help build new courses and then published an "AI-first" memo that sparked a public backlash. The story shows both what AI can do for a product and how quickly the way you talk about it can affect trust.

Duolingo is a popular app for learning languages through short, game-like lessons. Its business depends on keeping learners engaged day after day, which means it constantly needs new content, new exercises and new ways to help people practice. That makes it a natural place to test what AI can add to a product.

This case study sticks to widely reported facts and does not attempt to measure the financial effect of Duolingo's decisions. The goal is not to judge the company, but to draw practical lessons for small businesses thinking about AI in their own products, services and teams. Treat any details about the company's current strategy as subject to change and check recent news.

Background: why AI fit Duolingo's product

Language learning has always had a gap that apps struggle to fill. Exercises can tell you whether an answer is right or wrong, but explaining why, or letting you practice a real conversation, has traditionally needed a human teacher. Teachers are expensive and not available at 11 p.m. when a learner finally has ten minutes to practice.

Generative AI changed what an app could reasonably attempt. A system that can produce natural language can, in principle, explain a grammar point in context or play the other side of a simple conversation. For a company built on scale, that was an obvious opportunity: features that once required a tutor could now be offered inside the app.

Content is the other side of the story. Every new language course takes a great deal of work to write, check and organize. If AI could help draft and structure that material, a company could expand its catalog faster than it could by writing everything by hand.

What Duolingo did

2023: Duolingo Max and learner-facing AI

In 2023, Duolingo launched Duolingo Max, a higher subscription tier with AI features. Two of the best-known features were:

• Explain My Answer. Learners could get an explanation of why their answer was right or wrong, instead of just a red or green mark.
• Roleplay. Learners could practice conversations with AI characters in everyday scenarios.

Both features aimed at the gap described above: explanation and conversation practice. They also showed a common way companies package AI, as a premium feature rather than something baked into every free lesson.

2025: AI-built courses

In 2025, Duolingo announced a large batch of new courses built with AI help. This is a different use of AI from the learner features. Here AI was part of the production process, helping the company create content faster, rather than a feature customers interact with directly.

What happened next: the AI-first memo and the backlash

In April 2025, CEO Luis von Ahn shared an "AI-first" memo describing how the company intended to work. Among other things, it included reducing the use of contractors for work that AI can handle.

The memo drew public criticism. Much of the reaction focused on what it seemed to mean for people: contractors whose work could be replaced, and learners who worried that content made with AI would be lower in quality or less human. For a brand that had built a warm, playful relationship with its users, the reaction showed how sensitive the topic of AI and jobs had become.

It is worth separating two questions. One is whether using AI to build features and content was a sound business decision. The other is how the decision was communicated. Even a reasonable strategy can damage trust if customers and staff hear it first as a story about replacing people.

Results: what the Duolingo story shows

Without inventing figures, the public record still shows a clear pattern with both benefits and costs.

The first lesson is that AI worked best where it filled a real gap for customers. Explaining answers and practicing conversations are things learners wanted and that were hard to offer at scale before. The second lesson is that internal decisions do not stay internal. A memo written for employees became a public conversation about the company's values.

Lessons for a small business

You are not running a global app, but the same questions come up when you add AI to a product, a service or your team's work.

1. Start with a customer gap, not a technology

Duolingo's learner features targeted clear needs. Ask what your customers want that you currently cannot offer quickly or affordably: faster answers, personalized recommendations, explanations, after-hours help. Build AI around that need.

2. Keep quality checks on AI-made content

If you use AI to produce content customers rely on, such as guides, courses, menus, instructions or product information, keep a human review step. Customers notice errors, and "made with AI" becomes the explanation they reach for when something feels off.

3. Decide how you will talk about AI before you announce it

Think about how a customer, an employee and a freelancer would each read your announcement. Explain what changes for customers, what stays the same and how quality is protected.

4. Be fair to the people affected

If AI changes the work of staff, freelancers or contractors, talk to them early, honestly and directly. Offer notice, retraining or new roles where you can. Small businesses rely heavily on reputation and relationships, and a clumsy change can damage both. Employment and contract rules vary, so check the official guidance or a qualified adviser before changing anyone's work. This is general information, not legal advice.

5. Consider whether AI is a premium feature or a standard one

Duolingo placed its first learner-facing AI features in a higher tier. For your business, decide whether AI-powered extras justify a higher price, or whether they simply make your standard service better. Either can work, but be clear and honest about what customers get.

How to plan an AI feature for your own business

1. List customer pain points. Use support messages, reviews and conversations to find where customers get stuck.
2. Pick one gap AI could help with. Choose something low risk where mistakes are easy to catch.
3. Prototype quietly. Test with general assistants such as ChatGPT, Claude or Gemini, or a no-code custom assistant, before building anything.
4. Add human review. Decide who checks AI output and how often.
5. Check data and privacy. Use business plans with suitable data controls and never feed customer data into personal accounts.
6. Plan the message. Write how you will explain the change to customers, staff and freelancers.
7. Launch small and listen. Watch feedback closely and be ready to adjust.

I run a [type of business]. We're considering an AI feature that
[describe the feature]. Write three short announcements:
1. for customers (what improves for them, how quality is checked)
2. for staff (what changes in their work, how we will support them)
3. for freelancers we work with (honest, respectful, clear next steps)
Then list 5 questions or concerns each group is likely to raise.

Frequently asked questions

The bottom line

Duolingo shows how AI can add real value to a product, from explaining answers and practicing conversations to building more courses faster. It also shows that how you describe AI to customers and staff can shape the response as much as the technology itself. For a small business, start with a real customer need, keep humans checking quality and be fair and open with the people your changes affect. Company strategies and AI tools change, so check recent sources before drawing conclusions.