Illustrative example: how a representative compliance engagement would run for an industrial manufacturer ahead of the 2 December 2027 obligations
This case study illustrates a systematic approach to AI-system inventory, risk classification and remediation under the EU AI Act. The scenario shows how a representative engagement would run — no client outcome is claimed. Illustrative scenario, not a specific client engagement.
Omvang: Representative: 1,000–20,000-employee industrial manufacturers
Reach a defensible EU AI Act position for AI systems across plants — inventory, risk classification and assessment-ready documentation — before the 2 December 2027 obligations bite.
An indicative portfolio: dozens of AI systems across plants — visual quality inspection, predictive maintenance, production scheduling, worker-safety monitoring — with no central inventory
A worker-safety monitoring system likely lands in Annex III high-risk territory and lacks the required risk-management documentation
Vision inspection models retrained by line engineers with no change-control or logging of the kind Article 12 requires
A predictive-maintenance vendor cannot produce the technical documentation the Act requires of providers
Human-oversight arrangements (Article 14) exist in practice on the line but are documented nowhere
The board wants a defensible answer before the 2 December 2027 enforcement date
A representative engagement would run end-to-end — from system inventory through technical measures, documentation, and ongoing monitoring — sized to an industrial plant portfolio.
The engagement would start with complete AI system discovery and risk classification, implement the required technical measures (bias testing, explainability, human oversight) and compliant documentation for each high-risk system, and establish an AI governance office with clear roles, processes, and audit trails.
Complete AI system inventory across all business units. Risk classification per Annex III. In the modelled portfolio, 12 high-risk systems are prioritised.
3 weeksLogging and change-control for retrained vision models, explainability where the Act requires it, and documented human-oversight workflows for the line.
5 weeksTechnical documentation per Annex IV, an AI risk-management system, and a model registry with automatic documentation.
3 weeksA mock audit with external counsel, and training for the AI governance team on ongoing compliance.
2 weeksIllustrative outcome: a classified inventory, per-system remediation plans, and governance a hostile auditor could follow — owned by the manufacturer's team. No client outcome is claimed.