I Built an AI Ops Center for My Job Search. Here's What the Data Actually Says.

I built a real operating system to run my executive job search: a scored pipeline, a weekly review, a tracked follow-up queue. I wrote about it as a system worth copying. Then I asked myself the harder question: is it actually working? I don't have a job yet, so I went and checked, instead of assuming the system gets credit just because it's well built.

Here's what I found, and it wasn't flattering.

Looking at my own tracked pipeline, the majority of roles I've evaluated never became an actual application. A lot of hours went into scoring and researching roles that, on reflection, were never going to be worth pursuing that closely in the first place. Meanwhile, the handful of roles that actually moved, real conversations, real interviews, almost all of them moved because a specific person was involved, not because of a cold application.

That matches what the outside data says, and the outside data is worse than most people assume right now. Cold-application-to-interview conversion has reportedly dropped from roughly 8% a couple of years ago to around 3% today. The median time from starting a search to a first offer is running near 108 days as of early this year, up sharply from the year before. An estimated 80% of executive-level roles are never publicly posted at all, they get filled through referrals and search relationships. And referrals account for something like a third to half of all hires while representing only a small single-digit percentage of applicants.

So the honest read on my own system: the tracking wasn't the problem. It was accurately recording a mix of activity that was weighted toward the channel that converts at 3%, and underweighted toward the one that produces a third to half of actual hires. That's not a tooling failure. That's a behavior gap between what good tracking told me and what I actually spent my hours doing.

There's a reason for that gap worth naming honestly: scoring and evaluating roles against a rubric is exactly the kind of structured, solvable work I'm energized by. Sustained relationship-building, the repeated follow-up, staying present with people without a clean resolution, is the kind of work that drains me even when it's working. So there's a real chance I built an elaborate, genuinely useful tracking system partly because it's the work I enjoy, while the harder, more effective lever quietly got less of my time.

What's changing: a hard cap on new-role evaluation time each week, and a minimum number of direct outreach touches that gets tracked with the same discipline the scoring system gets now, not as a nice-to-have, as the primary weekly number I look at. If you're running a job search and it's not producing results, don't just ask whether you're organized. Ask what your system is actually optimizing for, and whether that matches what the data says actually works.