Innovation Diamond

Innovation Diamond

Capstone presentation integrating all course concepts into a complete ML-powered innovation pipeline.

Downloads

Learning Outcomes

By reviewing this capstone, you will:

• Understand the Innovation Diamond framework (divergence -> convergence)
• See how all 14 ML topics integrate into a unified workflow
• Apply ML to transform 1 challenge into 5000 possibilities, then converge to 5 strategic solutions
• Connect design thinking with machine learning methods

Visual Guides

Innovation Diamond Overview

Idea Generation Phase

Final Strategic Solutions

Prerequisites

• All 14 course topics recommended
• Understanding of ML paradigms (supervised, unsupervised, generative)
• Familiarity with clustering and classification
• Knowledge of NLP and sentiment analysis

The Innovation Diamond Framework

Phase 1: Divergence (1 -> 5000)

Expand possibilities using generative AI:

1. Start with 1 ESG challenge
2. Use LLMs to generate 100 initial ideas
3. Apply feature extraction and embeddings
4. Expand to 5000 solution variants

Phase 2: Peak (Maximum Exploration)

The widest point of the diamond:

• NLP and sentiment analysis for idea evaluation
• Topic modeling to identify themes
• Peak pool of 5000 diverse possibilities

Phase 3: Convergence (5000 -> 5)

ML-driven filtering and selection:

• Clustering to group similar ideas
• Classification to tier by feasibility
• Validation metrics to score quality
• Final 5 strategic solutions

Topic Integration Map

Key Takeaways

• ML enables systematic innovation at scale
• Design thinking + ML = Innovation Diamond
• Divergent thinking requires generative models
• Convergent thinking requires discriminative models
• The balance between exploration and exploitation is crucial

Common Pitfalls

• Converging too early (missing diverse solutions)
• Generating without evaluating (idea overload)
• Ignoring edge cases from clustering
• Over-relying on single ML technique
• Forgetting human oversight in final selection

(c) Joerg Osterrieder 2025