When discussing the production use of LLM-based agents for business-critical systems, there is often talk about adding a “human-in-the-loop.” But the human is already in the loop. People have always done things for themselves.

When computers first outperformed humans in chess, people asked: “What will humans do now? We are all doomed.” It felt like there was no reason to play chess anymore because computers could do it better. Yet, chess is still a very popular game, and people still compete with people.

Throughout history, whatever people do, they do it for other people - because humans need humans. Initially, we did everything by hand. Then we added powerful machines. Then we got smart machines and computers. It always seemed like people wouldn’t be needed anymore.

Now we have sophisticated, intelligent language models. And once again, it seems like this is the end. But the fact is, humans just need humans.

We aren’t adding a human to the AI’s process. We are just adding an LLM to the loop, not the other way around.