I will not promote. Not every AI workflow needs to be an AI agent.

I think people are making this stuff more complicated than it needs to be. Sometimes you just need an API call, a few rules, and an LLM for one specific step. Instead, people build multi-agent systems that are harder to debug, more expensive to run, and somehow still need a human to check everything. Obviously, some workflows do need agents. But figuring out where autonomy actually makes sense seems like the harder part. Where do you draw the line?