Delivering a Beginner-Level AI-Generated Language Course in 4 Months

The Client: A Language Learning App

The client, a language learning app, serves millions of learners across its digital products. During this engagement, the client was rebuilding its learning experience from the ground up, with AI-generated content as a foundation for more adaptive, personalized learning. The environment was fast-paced and launch-driven, with senior leadership closely involved and tight release cadences.

‍

The Mission: Principal Product Manager — AI Content Creation

The client recognized an opportunity to accelerate a new AI content initiative with principal-level product leadership across four areas: content creation, content enhancement, content storage, and editor tooling. The mission was to enable a major app redesign by delivering a fully AI-generated beginner-level Spanish course across text, audio, and images on a short deadline, while establishing reliable AI generation, automated quality-improvement loops, and scalable content-enhancement pipelines. Key stakeholders included senior product, AI, and learning content leaders.

🟢 Content enhancement is the step that turns raw generated text into usable lessons by adding audio and images before content is reviewed and stored.

‍

Our Main Quest: Deliver a Beginner-Level AI-Generated Spanish Course

Launch: Onboarded Very Fast

We aligned teams quickly around what “good” generated beginner-level content should look like by mapping exercise types to the agreed priority list. We kept decisions moving against the deadline by establishing clear ways of working across content generation, content enhancement, and review, including team dailies and targeted workshops for solution discovery and design. We created shared expectations across technology, learning content, and design stakeholders by actively managing tradeoffs, dependencies, and delivery risks.

Explore and Conquer: Quality-Controlled Generation at Scale

Problem: The client set an ambitious target: deliver a fully AI-generated beginner-level Spanish course, including text, audio, and images, on a short deadline to enable a redesign of the current app. The team recognized an opportunity to generate content at scale while holding every output to a rigorous quality bar across text, audio, and images.

What we did (tied to results):

• Enabled beginner-level Spanish course delivery by generating lessons and exercises through the AI generation system, mapped to the agreed exercise-type priority list.
• Completed the course media package by producing audio and images through AI models and third-party tools, while aligning requirements with stakeholders.
• Made human review repeatable by building a new content review tool and establishing evaluation criteria for generated content.
• Raised the quality threshold on deadline by creating a fast iteration cycle, testing generated outputs, and pivoting the approach where needed.
• Kept output reliable as volume scaled by running continuous quality control and handling bugs across the generation pipeline.
• Protected delivery momentum by structuring backlogs and logging decisions in shared product documentation.
• Restored product clarity across creation, enhancement, storage, and editor tooling by reconnecting the four teams around shared goals and delivery decisions.
• Set the next phase up to scale beyond the initial beginner-level course by defining the product vision, goals, and cost forecast for broader curriculum generation.

‍

Mini-Missions: Our Side Quests

• Image generation discovery: Improved generated image quality by running desk research and prompt testing with the design team, then locking a reusable format standard for the course.
• AI tooling unblocked: Enabled safe use of AI models and third-party generation tools by driving legal clearance and documenting guardrails for the generation pipeline.

‍

Mission Achievements: Delivered Outcomes

✅ Enabled a major app redesign by delivering a fully AI-generated beginner-level Spanish course across text, audio, and images.

✅ Protected content quality at scale by establishing evaluation criteria and human-in-the-loop review across text, audio, and images.

✅ Accelerated review of AI-generated content by building a new content review tool that operationalized the human review process.

‍