AI and Competent Person Authority

LinkedIn Post 7: AI and Competent Person Authority

Here's a governance principle that separates mature organizations from those running experiments:

Competent person authority must be real, not nominal.

When a competent person approves an AI-recommended inspection deferral, they must:

1. Understand the AI system (not deeply, but genuinely)

2. Evaluate the recommendation against defined criteria

3. Have the authority to override the recommendation

4. Document their reasoning

5. Accept accountability for the decision

If any of these elements is missing, the competent person has become a rubber stamp for the algorithm.

This is not theory. This is how law and regulation work. A competent person who cannot genuinely evaluate a decision is not competent. A competent person without authority is accountable without agency.

Organizations that implement real competent person authority:

• Require training in AI basics

• Create decision templates for approval/rejection

• Establish regular review of past decisions

• Document reasoning for every approval and rejection

• Empower competent persons to override AI recommendations

These organizations operate with lower liability risk. They also make better decisions because they combine human judgment with AI insight.

Is your competent person authority real, or just nominal?

#CompetentPerson #SafetyGovernance #Authority #Accountability #PSSR