Human in the loop is not a nice-to-have

We are starting to hear about the first redundancy plans triggered by AI. And yet, we cannot stop stressing how critical it is to keep humans in the loop.

Some say it because they have to. Some say it to tick a compliance box. Others genuinely mean it. But the truth is that we must control what our AI tools produce, even when their output far exceeds our capacity to review it.

Let us give you a real example.

Look at this Google Cloud cost graph. You can see a steady upward trend — then it suddenly flatlines to zero. What happened?

A team vibe-coded an internal tool with AI. Two days, done. Impressive. But the costs kept creeping up, day after day, until one sharp operator thought: this doesn't feel right.

When he flagged it, the dev team found a flawed implementation decision made by the AI. The fix? A couple of lines of code. The result? The cost did not just stop growing — it disappeared entirely.

One person's gut feeling saved the company real money. That is what human in the loop actually means.

Have you experienced something similar? We would love to hear your story. And if you are curious about what the specific issue was here, drop a comment or reach out — we would be happy to share.