Lecture Notes on AI and Bureaucracy

Introduction: The Paradox of AI and Bureaucracy

• Companies have aimed to reduce bureaucracy for the past decade.
• AI technologies could paradoxically increase bureaucratic tendencies.
• Bureaucracy favors rules/procedures over human judgment, similar to AI's operation.
• Algocracy is the term for a system where AI makes critical decisions devoid of human control.

The Human-Zero Mindset

• Executives often seek to eliminate human involvement in favor of AI.
• This approach is termed the "human-zero mindset."
• Relying solely on AI can lead to poor outcomes due to lack of human oversight.

Importance of Keeping Humans in the Loop

• AI can behave irrationally without human guidance.
• Examples include:
  • Amazon's algorithm suggesting inappropriate products to customers.
  • AI rejecting university applications based on flawed historical data.
• Consequences: Lack of accountability when AI makes decisions.

The Human Plus AI Approach

• The preferred alternative is the "Human plus AI" model.
• This approach is more challenging:
  • 10% coding algorithms.
  • 20% building technology around algorithms.
  • 70% integrating AI with human processes to maximize outcomes.
• Skipping the 70% can lead to wasted resources or dire consequences (e.g., aircraft crashes).

Case Study: Healthcare Application

• AI was used to identify patients at risk for heart attacks when starting a new drug.
• Initial model flagged 62% of patients at zero risk.
• Collaboration with doctors was vital for assessing the medical logic of predictors.
• Result: Developed a protocol that improved patient safety and drug sales.

Case Study: Fashion Retailer

• AI forecasted sales, outperforming human buyers initially.
• However, human insights were invaluable in refining predictions.
• Outcome: A new model incorporating human input led to a 50% reduction in forecasting errors and significant savings.

Ethical Considerations in AI Deployment

• Humans must define ethical guidelines for AI behavior to prevent manipulation or discrimination.
• Example: Health insurer's AI program that risked contacting clients unaware of their impending hospital visits.

Building Effective AI Teams

• Companies should mobilize diverse teams (tech, AI, ethics) to tackle AI deployment.
• This includes a structured cycle of the 10-20-70% framework.
• Taking bold steps to apply AI effectively can yield significant benefits.

Public Concerns about AI

• Surveys show public fears regarding AI's impact on jobs, privacy, and a dehumanized society.
• Companies face backlash if they push for algocracy.

Conclusion: The Future of AI in Organizations

• Successful organizations will invest in human knowledge alongside AI.
• Key Takeaway: Human knowledge is essential for deriving insights from data, ensuring responsible AI integration.