Learning Platforms & Capability Transfer

KI-Werkstatt

Knowledge Becomes an Experience.

The difficult idea

Turn rapidly changing AI expertise into a learning world that can be created, reviewed and expanded as the subject evolves.

FTL role: Creative concept, experience design, visual language, technical foundation and production workflow
Domain: Universities & Institutions
Format: Learning environment · Agentic workflow · Platform
Credits: Developed for Munich University of Applied Sciences.

Context

Students, lecturers and staff need practical access to complex AI topics. Static course pages cannot match the speed of new technology, policy and institutional questions.

Creative vision

Make knowledge feel navigable: compact Learning Nuggets combine explanation, interaction, quizzes and conversational moments inside one distinctive digital world.

Challenge & constraints

Information, didactics, visual identity, interaction, technical architecture and institutional responsibility had to function as one system.

What was made

A coherent interactive learning environment with guided experiences, visual characters, scenarios, quizzes and compact Learning Nuggets.

What was built

The Nugget Generator: a reusable, human-reviewed workflow coordinating specialist steps for extraction, structure, design, implementation and validation.

Dual lens

Experience / System

The outcome and the infrastructure remain part of the same project truth.

The illustrated KI-Werkstatt learning environment on a desktop display

What people see and feel

A world to enter

Learners encounter AI topics through a calm, authored environment rather than a stack of disconnected documents.

Compact journeys

Each Learning Nugget combines clear explanations, guided interactions, practical scenarios and checks for understanding.

Human clarity

The interface makes complex material approachable while keeping institutional context and responsibility visible.

System

Approved knowledge. Expert sources and a defined learning objective establish the project knowledge layer.

Coordinated production. Specialized agent and tool roles support extraction, didactic planning, content creation, build and validation.

Human approval. Editors and subject experts accept, correct or return outputs before anything is published.

Human + technical orchestration

Direction remains visible at every stage.

Human responsibility

Define learning objective
Approve sources
Direct the experience
Review accuracy and tone
Authorize publication

Technical responsibility

Extract relevant concepts
Structure candidate content
Assemble interactive output
Run validation checks
Preserve reusable components

Iteration & validation

Factual, structural, visual, technical and accessibility review is separated from generation. Failed checks return the Nugget to the responsible stage.

Output & deployment

Approved Learning Nuggets enter the university learning environment; the workflow remains available for future modules and updates.

Reusable capability

A production environment, component library and review process the institution can continue using as requirements evolve.

Learning

Meaningful AI education emerges when content, interaction, identity, architecture and editorial responsibility are designed together.

Working system

From source to deployed Learning Nugget.

Every handoff, decision and correction path remains observable.

01
Source
Input
Approved expert material
Output
Bounded project knowledge
Human
Chooses sources and learning objective
System
Indexes source content
Decision
Is the evidence approved and sufficient?
Correction
Request missing or corrected source material
Reusable · Knowledge layer
02
Extract
Input
Project knowledge
Output
Relevant evidence and concepts
Human
Confirms relevance
System
Finds and cites candidate material
Decision
Does every claim trace to a source?
Correction
Return unsupported claims
Reusable · Extraction pattern
03
Structure
Input
Approved concepts
Output
Ordered learning sequence
Human
Sets didactic intent
System
Proposes modular structure
Decision
Does the sequence serve the learner?
Correction
Reframe or reorder
Reusable · Nugget schema
04
Design
Input
Learning sequence
Output
Experience specification
Human
Directs character, interaction and tone
System
Maps content to components
Decision
Is the experience clear and coherent?
Correction
Correct visual or interaction logic
Reusable · Design system
05
Build
Input
Experience specification
Output
Interactive Learning Nugget
Human
Reviews authored result
System
Assembles content and interactions
Decision
Does the build match the specification?
Correction
Return implementation defects
Reusable · Component library
06
Validate
Input
Built Nugget
Output
Check results
Human
Interprets severity and context
System
Runs factual, visual, technical and accessibility checks
Decision
Are all release conditions met?
Correction
Route each failure to its source stage
Reusable · Validation suite
07
Review
Input
Nugget and check results
Output
Approved or corrected version
Human
Accepts, edits or rejects
System
Records version and corrections
Decision
Is publication authorized?
Correction
Return with explicit correction
Reusable · Approval record
08
Publish
Input
Authorized Nugget
Output
Deployed learning experience
Human
Owns release
System
Publishes into the environment
Decision
Is the deployed result healthy?
Correction
Withdraw or roll back
Reusable · Deployment path
09
Improve
Input
Observed use and feedback
Output
Next-version decisions
Human
Interprets feedback
System
Surfaces patterns without replacing judgment
Decision
What should change and why?
Correction
Preserve the approved version
Reusable · Learning loop

Evidence & limits

Claims carry their context.

Observed

The platform and generator demonstrate a complete path from expert material to a reviewable interactive learning output.

What becomes possible next

Extend the production system to new subject areas, formats and institutional teams without abandoning human approval.