AI governance for manufacturing
For manufacturers, the AI Act arrives on two fronts with different deadlines. AI systems embedded in products follow the Annex I timeline, with a 2028 horizon, while systems managing people on the factory floor follow Annex III.
The regulatory framework
- AI Act, Annex I: AI systems as safety components of products subject to harmonisation legislation, with obligations applying from August 2028
- AI Act, Annex III: systems for worker management, from recruitment and task allocation to monitoring
- Machinery Regulation (EU) 2023/1230, which explicitly addresses systems with evolving behaviour
- Cyber Resilience Act for products with digital elements
- Workplace health and safety legislation
Use cases and their risk level
| Use case | Classification | Note |
|---|---|---|
| AI in machinery safety components | High risk, Annex I | Obligations from August 2028, with product conformity assessment |
| Collaborative robots with adaptive functions | High risk | The AI Act and Machinery Regulation both apply |
| Worker productivity monitoring | High risk, Annex III | Worker management, also subject to employment law constraints |
| Visual quality control | Minimal risk when not safety-related | Its classification changes when the control performs a safety function |
| Predictive maintenance | Minimal risk | It should still be included in the inventory |
Where to start
- 1Separate systems embedded in a product sold to customers from those that remain in the plant, because the first group makes you a provider under the regulation
- 2Determine now, not in 2028, which product lines will incorporate AI, because conformity assessment must be designed alongside the product
- 3Review systems that measure people, often introduced as efficiency tools without a risk assessment
What we do for the sector
The path is always the same and the content changes: it starts from the use case map, assesses impact before investing, validates with a prototype, and only then reaches production. Rapid prototyping runs through protot.ai, our validation unit.
Use case definition
We separate systems destined to be embedded in a product sold to customers from those that stay in the plant, because the first group makes the company a provider under the regulation. The resulting portfolio has two separate tracks, with different deadlines.
Impact assessment
For plant systems we measure impact on productivity, quality and safety. For product-bound systems the assessment has to happen now even though obligations arrive in 2028, because conformity is designed alongside the product rather than added at the end.
Prototyping and validation
With protot.ai we test the hypothesis on one line or one cell before extending it, which is also how to learn whether the available data supports the use case. In predictive maintenance the quality of historical data determines the result more than the algorithm does.
Production and oversight
In production the work covers integration with MES and ERP, traceability, and for product-bound systems the preparation of conformity documentation. For systems that measure people, employment law constraints apply on top.