Interactive PDCA Cycle Simulator

Walk through a complete SOAI-PDCA cycle with a fictional AI system and see how the proposed 33-seat Council workflow could structure review at each phase.

Everything on this page — the system, the figures, and every “council recommendation” — is a scripted illustration written in advance. No council deliberates when you click, and no number here is a measurement.

HealthBot AI

AI-powered medical symptom checker and health advisor

High-Risk

Applicable Frameworks:

EU AI Act
HIPAA
NIST AI RMF

Scale:

500,000 monthly active usersillustrative, not a measurement

PDCA Cycle Progress25% Complete
PLAN
DO
CHECK
ACT

Phase 1: PLAN

Define objectives and establish safety requirements

Scenario

HealthBot AI is a medical symptom checker used by 500,000 patients monthly. Recent user feedback indicates potential bias in diagnostic suggestions for certain demographics.illustrative, not a measurement

Key Activities

  • Document AI system scope and boundaries
  • Conduct comprehensive risk assessment
  • Map applicable regulatory frameworks (EU AI Act, HIPAA)
  • Identify potential bias and fairness issues
  • Define success criteria and KPIs
  • Establish baseline compliance requirements

Designed council recommendations (simulation)

Scripted examples of a proposed 33-seat workflow with a target threshold of 23/33. No live voters operate here; independence and fault tolerance are not demonstrated.

🤖 Designed council (simulation, not a live vote) — analysis: "High-risk medical AI system requires enhanced oversight under EU AI Act Article 6"illustrative, not a measurement

🔍 Bias Detection: "Recommend demographic fairness audit across age, gender, ethnicity, and socioeconomic factors"

📊 Risk Priority: "Focus on diagnostic accuracy disparities (Critical), data privacy (High), and explainability (High)"

✅ Gap note: "HIPAA controls documented. EU AI Act conformity assessment required before deployment in EU"

Identified Risks

  • Diagnostic bias leading to health disparities
  • Privacy violations of sensitive health data
  • Lack of explainability in medical recommendations
  • Regulatory non-compliance in multiple jurisdictions

Phase Outcomes

  • ✓ Risk Assessment Matrix completed (12 risks identified)illustrative, not a measurement
  • ✓ Compliance requirements mapped to 3 frameworksillustrative, not a measurement
  • ✓ Implementation roadmap approved (16-week timeline)illustrative, not a measurement
  • ✓ Stakeholder sign-off obtained