Sber • 2026
An AI-assisted design stage for course production
A research and design stage that runs before a course gets built. AI helps go from source materials, the brief and stakeholder interviews to a chain that starts with the behavior gap.
The problem
TODO: what the design stage looked like before, and where time or quality got lost.
Constraints
- TODO: platform, deadlines, regulation, data.
My role
What I did
I proposed a separate research and design stage before assembly and designed how it works, from the inputs to the course outline.
I was an early user of the assembly tool and acted as its methodology expert.
What the team did
The tool that assembles xAPI packages was built by a colleague.
TODO: who else took part and what they did.
Key decisions
Decision 01
Where the AI starts
Why
TODO: what we chose and why.
Decision 02
Where a human review is required
Why
TODO: what we chose and why.
Decision 03
Theory first or practice first
Why
TODO: why practice comes before theory.
The artifact
The stage takes everything we know before production starts, including source materials, the brief, surveys and interviews with stakeholders. AI helps turn it into a chain that goes from the behavior gap to learning outcomes, then to practice, and only then to theory.
Below is the same chain on a made-up example from a different field.
- Source materials & stakeholder interviews
- Behavior gapHuman review
- Learning outcomesHuman review
- Practice
- Theory
- Course buildbuild stage — built by a colleague
- What people do nowSupport agents close tickets after sending a template answer, without checking whether the problem is actually solved.
- What they should doConfirm the problem is solved before closing, and note what fixed it.
- Learning outcomeGiven a customer ticket, the agent verifies the resolution with the customer and documents the fix before closing.
- PracticeThree simulated tickets where the template answer looks right but doesn’t solve the problem. The learner decides whether to close and why.
- Theory (only what practice needs)A short checklist for verifying a resolution.
How we’ll know it worked
The stage is being piloted on the first course.
TODO: what we’ll compare (for example, time to launch, number of iterations with the client, share of practice in the course).
What I learned
TODO: 1–3 sentences.