Unless they start with your actual situation.

  1. If you are a medium-sized company and just starting out:
    → Start with an AI inventory. What you don't know, you can't control.
  2. If you are already using AI but have no documentation:
    → Start by classifying the application. High-risk or not – that changes everything.
  3. If you are certified to ISO 42001, but ignore the AI Regulation:
    → Start by mapping roles. Your AI system Owner ≠ Provider or Operator.
  4. If you are a software provider and are embedding AI:
    → Start with supply chain clarity. Are you a provider or an operator? Wrong answer = wrong obligations.
  5. If you work in HR, Finance, or customer-facing roles:
    → Start with prohibited practices. Some AI applications are now simply illegal.
  6. If you are already compliant, but scaling AI usage:
    Start planning your conformity assessment. High-risk systems require auditable documentation.

The inconvenient truth:
There is no universal starting point.

Generic compliance checklists are a waste of time.
Your entry point depends on where you stand – not on where consultants want you to be.

I've seen organisations build governance frameworks for low-risk systems while ignoring prohibited practices. I've seen others chasing ISO certifications while completely missing out on AI regulation classification.

Different situations, different priorities, different consequences.

The question is not: „What should we do first?“
The question is: „Where are we exposed first?“

 

 

Author: Achim Korten, February 2026