Operationalizing AIMI
Unintended Consequences - AI Checklist
- AI needs to be understood and sponsored at board level
- Education, Training & Awareness of AI is a company wide responsibility
- Upskilling of your workforce is an imperative for future growth
- Ethical AI Guidelines & Guardrails should be chaired by a Multi-Stakeholder Group
- Diversity & Inclusion should form part of every AI initiative
- AI technology should be seamless to use for everyone and not just for the Data Science community
- AI Regulation and Compliance are significant operational risks
- Consumers by law are entitled to request explainable and auditable (human interpretable) decisions made by AI
- AI Algorithms must be robust, reliable, monitored and tested throughout their existence
- Human & Machine Bias (Fairness) go hand in hand – Eradicating bias in protected classes is hard and non-trivial
- Monitoring of AI systems counteracts adversarial attacks on AI models, loss of data and AI business knowhow
- Without regular AI Health checks companies are widen open to regulatory fines and significant operational risk
Rigorously Validate, Update & Monitor AI Models Pre & Post Deployment
Explain
Trace
Vulnerabilities
Testing
Action
Fairness
Imitation
Privacy
Multi-Stakeholder AI Review
Ethics
Framework
Governance
Compliance
Regulatory
Risk
Strategic AI
- Use Case Acceptance
- Regular Health Checks
- Pre & Post Production Reviews
- AI Model Monitoring
- Guidelines / Guardrails
- Approvals / Rejections
- Robust / Safe
- Explainable & Auditable