Glossary · AI engineering and governance
AI system lifecycle
Also known as: AI lifecycle, AI system life cycle
German: Lebenszyklus eines KI-Systems
The AI system lifecycle is the sequence of stages an AI system passes through from inception to retirement: typically design and development (including data acquisition and model training), verification and validation, deployment, operation and monitoring, re-evaluation and retirement. ISO/IEC 22989 describes such a lifecycle model and ISO/IEC 5338 defines the corresponding processes.
- Industrial AI
- AI
- Standards
In one sentence
The AI system lifecycle spans design, data and training, validation, deployment, operation and monitoring, re-evaluation and retirement.
Example
The lifecycle plan for an AI-based sorting system defines retraining triggers, revalidation after each retraining and the steps for decommissioning the model.
How it applies
- Engineering: Unlike conventional software, AI systems can change behavior when data changes. The lifecycle therefore includes continuous monitoring and repeated validation, not only a one-time release.
- Safety: When an AI function is part of a machine or plant, its lifecycle must be aligned with the machine's development and with the Safety lifecycle where safety functions are affected.
- Documentation: Each stage produces evidence: data descriptions, validation reports, deployment records and monitoring results. Plan which documents are maintained over the lifetime and who updates them.
AI system lifecycle vs. model lifecycle
The model lifecycle covers training, versioning and replacement of the model. The AI system lifecycle covers the whole system, including data pipelines, user interface, human oversight and integration into the machine or process.