Twenty Years of Instinct, Then a Test
For twenty years, I believed I knew how to find the right person for a job.
I never thought of it as arrogance. It felt like knowledge I had earned. I had built recruitment businesses, sat across from hiring managers who needed someone yesterday, and worked alongside teams who stayed late to close a difficult search. After enough of that, you stop second-guessing yourself. You develop a feel for it. You look at a résumé and you just know.
Then we started building recruitment software, and that feel was tested in a way I was not ready for.
Being Asked to Explain My Own Judgment
It began with a question that sounded simple. When we sat down to teach the software how to screen candidates, we had to decide: should someone with exactly the right skills but the wrong job title rank below someone whose résumé matches the job description perfectly?
I had answered versions of that question for years, always in the moment, always by instinct, and never by writing the answer down. Software does not accept instinct. It asks you to say what you mean. And when I tried to say what I meant, I found assumptions I had never examined: about titles, about employers, about career paths, about what a “good candidate” is supposed to look like.
That was uncomfortable. Not because software could not think like a recruiter, but because I realized I had never really questioned how recruiters think. That included me.
After two decades, I was being asked to explain my own judgment, and for the first time I could not do it cleanly.
It made me more curious than defensive, and it changed what I wanted to build.
Two Beliefs I Had to Let Go
I had to let go of two beliefs.
The first was that instinct is a reliable standard. A lot of what I called instinct was pattern recognition, built on candidates I had watched succeed before. That is valuable, but it also narrows what you notice, and it is very hard to see a narrowing from the inside.
The second was that a résumé matching a job description tells you something about fit. It tells you about a match to a document.
If the document is wrong, matching it faster only gives you a more confident wrong answer.
Many Hiring Problems Are Requirements Problems
That second one stayed with me, because it points to something bigger than screening. So much of what we treat as a hiring problem is really a requirements problem. A hiring manager asks for a title when the business needs a capability. A job description is copied from the last opening. The real criteria appear halfway through the search, after the first shortlist disappoints. Everyone is working hard, toward a definition that was never settled.
It is humbling to say this as someone who builds matching technology, because automation can make the problem worse.
A ranked list looks authoritative. A vague requirement behind it starts to look like a precise answer.
Moving the Work to the Start
So we changed how we build. We asked recruiters to work directly with us and help train the model, because I wanted it to reflect how good recruiters actually think, not how a product team imagines they do. But I also know the risk in that: a model trained on recruiters can inherit recruiters’ blind spots, mine included.
So we moved the work to the start of the process. The person who owns the role writes the job description first, including what they will screen for. The definition comes straight from the person who knows the role best, before anyone starts searching.
I think this matters most for large teams. When a company hires across many roles and locations, a loosely defined role gets repeated again and again, and each recruiter brings their own instincts to it. The inconsistency that follows is hard to see and harder to fix.
Writing the criteria down first does not remove judgment. It makes judgment visible, so that when people disagree, they can talk about it.
What Still Needs a Person
None of this means I want to automate everything. I do not. Scheduling interviews and collecting documents are tasks no recruiter misses doing. But understanding why someone is thinking of leaving a job they have held for years, sensing that a hiring manager’s brief does not match what the business really needs, helping a candidate weigh an offer where money is only part of the decision: that is the part of this work I have always valued, and it needs a person who is accountable for the outcome.
After twenty years, I still believe in experience. I just no longer treat it as proof.
I treat it as a place to start asking better questions, and I value it more now that I am willing to question it.
I think the best hiring technology will not be the one that answers fastest. It will be the one that makes us ask better questions first.
Vantage