What the AI Couldn't See

The process was rigorous: we analyzed tickets, tasks, logged time, and output over several weeks. Three people consistently showed up as underutilized. Based on that analysis, we restructured their roles.

The effect wasn't immediate. It took months to surface, and when it did, it was gradual: communication between departments started to break down. Escalations that used to be resolved in 24 hours began taking 72. The volume of meetings increased for no clear reason. The fluidity the operation once had was gone, but no one could point to a cause in the numbers.

When I went to investigate what had changed, the pattern became clear.

Those three people did the work that no system could capture: they were the operation's informal translators. They intercepted problems before they became tickets. They kept information flowing between teams before it ever needed to become formal data. They sustained the trust that allowed one manager to call another without having to open a request.

That function never appeared in any report because it exists precisely outside the formal workflows. It's invisible by design.

The AI did exactly what it was supposed to do: it optimized what was visible and measurable. The problem wasn't the analysis itself, but the assumption that what is measurable is what holds the operation together.

In many companies, what holds things together is exactly what leaves no record. And when you optimize without understanding that, you strip away the glue without even noticing.

Save this post if you've seen something similar. And tell me in the comments: what invisible work exists in your company that no system can measure?