AI products tend to fall into a small set of shapes

If you look at AI-powered products built in the last few years, most of them fit one of a small number of patterns. Not because anyone decreed it, but because the underlying technology lends itself to certain shapes — and certain shapes have proven more useful than others.

Knowing the patterns is useful for two reasons. It helps you understand what other products are really doing under the surface. And it gives you a starting library of shapes to draw on when designing your own AI features.

This topic walks through the three patterns that show up most often: RAG, agents, and AI-supported workflows.

RAG: retrieval-augmented generation

Retrieval-augmented generation is one of the most common patterns in real AI products. The idea is simple....

Real AI patterns (RAG, agents, workflows)