COMPEL Glossary / model
Model
In AI and machine learning, a model is a mathematical representation learned from data that can make predictions or generate outputs.
What this means in practice
A model is the trained artifact -- the learned patterns encoded in numerical parameters -- that an organization deploys to automate decisions or augment human judgment. Models range from simple linear regressions with a handful of parameters to large language models with trillions of parameters. For governance purposes, each model in production should be tracked in a model registry with documented ownership, risk classification, performance metrics, and known limitations. In the COMPEL framework, model governance spans the entire lifecycle from initial development through deployment, monitoring, retraining, and eventual retirement.
Why it matters
AI models are not static assets but living systems that degrade, evolve, and influence decisions every day they operate. Each model in production should be tracked with documented ownership, risk classification, performance metrics, and known limitations. Organizations that treat models as deploy-and-forget software create ungoverned decision-making systems that accumulate risk silently over time.
How COMPEL uses it
Model governance spans the entire COMPEL lifecycle from initial development through deployment, monitoring, retraining, and eventual retirement. During Calibrate, existing model governance practices are assessed. The Model stage designs governance controls for each model's lifecycle. The Produce stage implements model registry and monitoring. The Evaluate stage audits governance compliance across all production models, ensuring continuous oversight.
Related articles in the Body of Knowledge
- Model: Designing the Target State
- The COMPEL Operating Model: Roles, RACI, and Decision Rights
- Creating the AI Operating Model Blueprint
- Introduction to the 18-Domain Maturity Model
- Generative AI and Large Language Models
- MLOps: From Model to Production
- Model Governance and Lifecycle Management
- Agentic AI Maturity Assessment: Extending the 18-Domain Model
Related Terms
Other glossary terms mentioned in this entry's definition and context.