Thanks! From my experience, yesβAI is one of the areas where Event-Driven Architecture fits really well.
We recently used it for bulk LLM processing on database records. Instead of processing 1,000 records one by one, each record became an independent event that workers processed in parallel. It made the pipeline much faster and much easier to scale.
That said, I'd still choose it only when the problem justifies it. Not every AI application needs an event-driven architecture.