RAD-AI: Rethinking Architecture Documentation for AI-Augmented Ecosystems
2026-03-30 • Software Engineering
Software EngineeringArtificial Intelligence
AI summaryⓘ
The authors explain that current ways to document software systems don’t work well for complex AI ecosystems where multiple AI parts interact and change over time. This is a problem because new laws, like the EU AI Act, require detailed technical documentation for AI systems, especially high-risk ones. To fix this, the authors created RAD-AI, which adds AI-focused sections to existing documentation methods and helps meet these legal requirements. Their tests with experts and real AI platforms show RAD-AI covers much more important AI details than current frameworks. They also demonstrate new issues that appear only in large AI ecosystems and aren’t captured by standard documentation tools.
AI-augmented ecosystemsarc42C4 modelEU AI ActAnnex IVtechnical documentationprobabilistic behaviormachine learning lifecyclesmart citiesregulatory compliance
Authors
Oliver Aleksander Larsen, Mahyar T. Moghaddam
Abstract
AI-augmented ecosystems (interconnected systems where multiple AI components interact through shared data and infrastructure) are becoming the architectural norm for smart cities, autonomous fleets, and intelligent platforms. Yet the architecture documentation frameworks practitioners rely on, arc42 and the C4 model, were designed for deterministic software and cannot capture probabilistic behavior, data-dependent evolution, or dual ML/software lifecycles. This gap carries regulatory consequence: the EU AI Act (Regulation 2024/1689) mandates technical documentation through Annex IV that no existing framework provides structured support for, with enforcement for high-risk systems beginning August 2, 2026. We present RAD-AI, a backward-compatible extension framework that augments arc42 with eight AI-specific sections and C4 with three diagram extensions, complemented by a systematic EU AI Act Annex IV compliance mapping. A regulatory coverage assessment with six experienced software-architecture practitioners provides preliminary evidence that RAD-AI increases Annex IV addressability from approximately 36% to 93% (mean rating) and demonstrates substantial improvement over existing frameworks. Comparative analysis on two production AI platforms (Uber Michelangelo, Netflix Metaflow) captures eight additional AI-specific concerns missed by standard frameworks and demonstrates that documentation deficiencies are structural rather than domain-specific. An illustrative smart mobility ecosystem case study reveals ecosystem-level concerns, including cascading drift and differentiated compliance obligations, that are invisible under standard notation.