Duolingo · AI product · Entry · 45 to 60 minutes

How should Duolingo add AI help after mistakes?

Duolingo is a mobile app that helps people learn languages through short lessons and practice exercises. Answer this product management case and get it graded line by line against a weighted rubric.

problem framing
ai ux
prioritization
metrics
experimentation

The situation

You are an entry-level PM on Duolingo's learning experience team. The team is exploring how recent large language models could improve paid learning features, and your director wants a recommendation for one narrow launch: what should happen right after a learner gets an exercise wrong and wants help understanding why. Today, Duolingo already tells learners whether an answer is right or wrong, but repeated grammar mistakes can still feel frustrating, especially when a learner needs the concept explained in a different way or with more examples. Some teammates want to add a free-form AI chat box so learners can ask anything. Others worry that open-ended chat will feel intimidating, require prompt skill, create safety risks, and pull learners away from the lesson flow. Assume this feature would first launch for language learners inside the existing app, and the team can only ship a small first version. Your job is to propose the product approach, explain why it solves a real learner problem, and show how you would measure whether it works.

What to produce

Write a 1 page product memo, 500 to 700 words.

Your answer must

Define the target user problem and when in the lesson flow the AI help should appear.
Recommend the core interaction design for the first version, including how much freedom learners should have to type versus using guided options.
Explain how your design balances learning value, ease of use, and AI safety or reliability risks.
List 3 success metrics, including at least 1 learning or behavior metric and at least 1 risk metric.
Propose a simple MVP experiment or rollout plan that could validate the idea before wider launch.

Constraints

Keep the feature inside Duolingo's existing lesson experience, not a separate standalone chatbot product.
Assume the team can launch only one small AI powered help flow in the next quarter.
Do not rely on human tutors or manual review for every interaction.
The first version must work for beginner language learners who may not know how to write strong prompts.

Format

Use short headings and bullets. Prioritize clear product choices over long background.

How Duolingo actually played it

A user level teardown of Duolingo Max, covering Explain My Answer and Roleplay. It shows how preset options and suggestions avoid the blank chat box.

Offer suggested replies
Put AI where users get stuck
Add small surprises

Built from How Duolingo cleverly integrated AI into their app by Fawzi Ammache.