Block 5. AI in Learning

AI plugs into a learning product in seven jobs. See which are mature vs risky in 2026, the review gate each needs, and how every one is tracked.

An AI tutor answers from your course, not its memory. See how RAG grounding works, what it costs to run, where it breaks, and when to build vs buy.

Synthetic avatar video can cut course production cost by 80%. See where it works, what it costs, the engagement trade-off, and the new disclosure law.

Automatic captions are a draft, not a finished caption. See the 99% accuracy bar, the caption laws, the cost math, and how to caption courses at scale.

Machine translation makes courses multilingual fast, but raw output is a draft. See the quality gate, the cost math, and one-course-many-languages design.

AI drafts quiz questions from a transcript in seconds, but raw output is a draft. See distractor quality, the review gate, and how to track each item.

AI turns a lecture transcript into summaries, key points, chapters, and flashcards. See extractive vs abstractive, the review gate, and how to track each.

Personalized learning AI explained: real adaptive paths vs the branching illusion, the learner model, the data it needs, and when adaptivity pays off.

Build vs buy learning AI: the three ways to add any AI feature, the four drivers, cost per learner with 2026 numbers, and the privacy and lock-in tests.