Skip to main content

AI-era Architecture for Software Architects

AI-era architecture covers designing systems that include large language models and using AI in engineering work. It involves handling non-determinism, latency and cost of model calls, retrieval, evaluation and guardrails, plus governance for AI-assisted coding, so generated code is reviewed, tested and traceable.

Stage: Modern Concerns · Part of the Software Architect roadmap · Last reviewed 19 November 2026

What to learn in AI-era Architecture

AI-era Architecture interview questions

What new risks does an LLM component add to an architecture?

Output is non-deterministic and can be wrong, latency and cost are variable, prompts can be attacked through injection, and data may leave your boundary. Designs need validation, fallbacks, monitoring, evaluation sets and strict control of tool access.

Where does retrieval-augmented generation fit in a system architecture?

RAG retrieves relevant documents from your own data, often via embeddings in a vector store, and adds them to the prompt so the model answers from that context. It reduces hallucination and keeps knowledge updatable without retraining the model. It is a retrieval service in front of the model.

Where AI-era Architecture fits in the Software Architect roadmap

AI-era Architecture belongs to the Modern Concerns stage (advanced level), which takes about 5 hours. Where the field is heading.

Back to the full Software Architect roadmap