Group Work Exercise

AI Solutions Lab: Designing within the Doughnut

IMC University of Applied Sciences Krems, Austria
Roman Mesicek

SAG Part 3

Workshop Mission

You are sustainability consultants.

Your client needs an AI solution that:

  • Operates within planetary boundaries
  • Strengthens social foundations
  • Makes economic sense

You have 90 minutes to develop a proposal.

Workshop Structure

Phase Duration Activity
Setup 10 min Challenge selection & team formation
Design 30 min Solution development in breakouts
Ethics 15 min Ethical evaluation framework
Present 15 min Rapid-fire presentations
Synthesis 15 min Pattern recognition
Close 5 min Reflection & voting

Your Workspace

Miro Board Structure:

  • Each group has dedicated workspace
  • All templates pre-loaded
  • Reference materials available
  • Shared synthesis area

Phase 1: Challenge Selection

10 Minutes

Choose your mission

Five Real-World Challenges

Each group receives one challenge card:

  1. Manufacturing - Reporting compliance for Austrian Company
  2. Housing - Energy optimization without surveillance
  3. Just Transition - Coal region workforce transformation
  4. Circular Economy - E-waste in electronics retail
  5. Agriculture - Small farm sustainability

Selection Process:

  • Random break-out groups in teams

What You'll Receive

In Your Miro Workspace:

  • Detailed challenge card with constraints
  • Solution design templates
  • Impact estimation tools (simplified)
  • Ethical evaluation framework
  • Presentation prep box

Key Constraints:

  • Limited budget
  • Real stakeholders
  • Austrian/EU context
  • Specific timeline

Preparation

Key Questions for Solutions

  1. Scale Management: How to capture efficiency without enabling overconsumption?
  2. Equity Integration: How to ensure benefits reach those most in need?
  3. Lifecycle Thinking: How to account for full system impacts?
  4. Governance Frameworks: What policies prevent negative outcomes?

Design Principles

  • Absolute targets, not just efficiency metrics
  • Accessibility requirements from inception
  • Transparent impact measurement
  • Democratic control of data and algorithms

Remember: Technology is neutral - deployment determines impact

Phase 2: Solution Design

30 Minutes (Breakout Rooms)

Build your proposal

Design Sprint Structure

Part A: Problem Analysis

  • Current state metrics
  • Root causes
  • Key affected groups
  • One "killer fact" that justifies AI

Part B: AI Solution

  • What it does (simple terms)
  • Data requirements
  • Implementation timeline
  • Technical complexity level

Part C: Impact Estimation

  • Environmental: Better or worse?
  • Social: Who wins, who loses?
  • Economic: Does it pay back?

Solution Design Tips

Keep It Simple:

  • You're not building the AI, you're proposing it
  • "Black box" descriptions are fine
  • Focus on outcomes, not algorithms

Think in Patterns:

  • Prediction → "AI predicts X to prevent Y"
  • Optimization → "AI finds best combination of..."
  • Classification → "AI identifies/sorts..."
  • Automation → "AI handles routine task of..."

Quantify with Ranges:

  • "10-25% improvement"
  • "€50-200k investment"
  • "6-12 month payback"

When stuck: What would a human expert do? Can AI do it faster/better/cheaper?

Impact Estimation: Environmental

🌍 Environmental Impact

Rate with traffic lights:

🔴 Worse = More energy than saves, increases emissions
🟡 Neutral = Efficiency gains offset by AI consumption
🟢 Better = Clear net reduction in energy/emissions

Quick check questions:

  • How much energy will the AI system use?
  • How much energy/resources will it save?
  • What's the net effect?

Reference: One server ≈ 100 laptops ≈ 5,000 kWh/year

Impact Estimation: Social

👥 Social Impact

Rate with traffic lights:

🔴 Negative = Jobs lost > created, increases inequality
🟡 Mixed = Some win, some lose, unclear net effect
🟢 Positive = More opportunities, broader access, fair distribution

Quick check questions:

  • Who benefits from this AI?
  • Who might be excluded or harmed?
  • Does it reduce or increase inequality?

Consider: Jobs, digital access, skills needed, language barriers

Impact Estimation: Economic

💰 Economic Viability

Rate with traffic lights:

🔴 Poor = Payback >3 years or never
🟡 Acceptable = Payback 1-3 years
🟢 Strong = Payback <1 year

Quick check questions:

  • What's the total investment needed?
  • What savings/revenue will it generate?
  • How quickly does it pay back?

Rule of thumb: If payback > 2 years, need strong non-financial justification

Phase 3: Ethical Evaluation

15 Minutes (Breakout Rooms)

Check your solution

Three Ethical Tests

1. Necessity Test

Is AI actually needed?

  • Could this be solved without AI?
  • Is AI the simplest solution?
  • What unique value does AI add?

2. Justice Analysis

Who benefits, who bears costs?

  • Distribution of benefits
  • Distribution of risks
  • Power dynamics
  • Voice in decisions

3. Governance Check

Who controls what?

  • Data ownership
  • Algorithm control
  • Accountability mechanisms

Red Flags to Watch For

⚠️ Warning Signs:

  • ✓ Benefits concentrated → Costs dispersed
  • ✓ Vulnerable groups bearing risks
  • ✓ No opt-out mechanism
  • ✓ Black box decision-making
  • ✓ Data extraction without benefit sharing

If you find red flags:
Don't abandon solution - propose safeguards!

Phase 4: Presentations

15 Minutes (Main Room)

Share your solution

Presentation Format

2 Minutes Per Group - Strict!

30 seconds each for:

1️⃣ Problem: The killer fact + Why AI?

2️⃣ Solution: What the AI does + Key impact number

3️⃣ Sustainability: Environmental/Social/Economic verdict

4️⃣ Ethics: Biggest risk + Mitigation

Then: 1 minute Q&A (choose one challenge from chat)

Phase 5: Synthesis

15 Minutes

Finding patterns

Pattern Recognition

Together we'll identify:

Success Patterns 🟢

What worked across multiple solutions?

Challenge Patterns 🔴

What obstacles appeared repeatedly?

Trade-off Patterns 🟡

Which tensions emerged consistently?

Use sticky notes in shared Miro space.

Connecting to Course Frameworks

Session 1: Planetary Boundaries

  • Which solutions help stay within ecological ceiling?
  • Which might push us further into overshoot?
  • Where do we see rebound effects?

Session 1: Social Foundation (Doughnut)

  • Which solutions strengthen social foundations?
  • Which might create new deprivations?
  • Who is left behind?

Session 2: Business Ethics & Responsibility

  • Are we creating value or extracting it?
  • Who is responsible when AI fails?
  • What would ethical business practice look like here?

Phase 6: Closing

5 Minutes

Reflection

Reflection

Three Quick Questions:

  1. Which AI application would you champion in your organization?

  2. Which AI application would you reject?

  3. What's your key insight from today?

Post responses in Miro reflection space.

Takeaways

What You've Practiced:

✅ Rapid problem analysis
✅ Solution design under constraints
✅ Impact estimation (even with uncertainty)
✅ Ethical evaluation frameworks
✅ Recognizing affected groups
✅ Trade-off recognition

Key Insight:

AI is neither good nor bad for sustainability.
It's a powerful amplifier that requires careful design and governance.