Your AI sourcing tool gives you a score it can't explain. That's the whole problem.

Open almost any AI sourcing tool and you will see the same thing: a candidate, and a number. 95% match. 87% fit. A confident score, printed next to a face.

Now try to answer one question about it. Why 95?

You can't. Neither can the tool. And that quiet fact is the biggest unsolved problem in recruiting technology right now.

Two flavors of the same flaw

The industry has converged on two versions of the same mistake.

The first is the black-box score. A model reads a profile, produces a number, and moves on. The number looks authoritative. It is also unexaminable. You cannot tell whether it rewarded the right skills, penalized the wrong things, or leaned on something it should never have touched.

The second is newer and louder: the agent that decides for you. Sourcing, screening, follow-ups, all handled automatically, while you are promoted from doer to strategist. It sounds like leverage. Read it again. It means the machine made the calls and you approved them. A recruiter who rubber-stamps an agent's picks is not a strategist. They are a signature.

Both take the judgment away from the person and hide the reasoning. That is the flaw.

Why an unexplainable score is a real problem, not just an annoying one

Three reasons, and they get more serious as they go.

You can't defend it. A hiring manager asks why you passed on their favorite candidate, or why you are pushing this one. "The tool scored them high" is not an answer. It is the moment your judgment looks outsourced.

It doesn't make you better. A number you cannot interrogate teaches you nothing. You run the same searches, trust the same opaque output, and your instincts never sharpen because the tool never shows you its reasoning to argue with.

It is becoming a liability. Courts are now looking hard at algorithmic screening that advances some applicants and drops others, on scores no one can account for, using signals no one audited. When the reasoning is a black box, "the tool did it" is not a defense. It is the problem.

What we built instead

We started from the opposite premise. The machine should do the legwork and show its work. The human should make the call.

So every candidate a search surfaces comes with the reasoning, in plain sight:

• The must-have skills that matched, and the ones that are missing.
• How well the title actually fits the role.
• Whether the person is realistically reachable for the job, including how far they sit from an onsite role.
• Whether their current employer just had layoffs, because timing is often the real reason someone takes your call.
• A timestamped record of that reasoning, so the same inputs always produce the same explanation, and you can go back and see exactly why a candidate scored the way they did.

None of that is decoration. It is the difference between a score you trust because it is confident and a shortlist you trust because you can see the thinking. One you defend to a hiring manager. The other you hope they don't ask about.

Signals, not strings

There is a second habit worth breaking: sourcing by keyword.

A boolean string matches words on a page. It has no idea that a company filed a layoff notice last week, that a startup just ran out of runway, or that a leadership change is about to shake loose a team. Those are the signals that decide who is actually ready to move, and they are invisible to keyword search.

We source on those signals. A candidate whose employer just announced cuts is not a better engineer than they were a month ago. They are a better call this week. Sourcing that understands timing beats sourcing that only understands keywords, every time.

It gets sharper with use, and the sharpness is yours

Most tools measure vanity: opens, clicks, sends. We measure what actually happened. Which outreach framing earned a reply. Which approach led to an interview. Which led to a hire.

That means the system learns from your outcomes, not a vendor's aggregate data. It can tell you which of your angles is working and which one just felt good to write. That feedback loop is the part no amount of bought data can copy, because it is built from your placements, not the market's profiles.

Automate the legwork. Keep the decision.

This is the whole philosophy in one line. The parts of recruiting that are grunt work, finding people, pulling the signals, drafting the first message, we automate without apology. The part that is judgment, who to actually pursue and why, stays with you, because that is the part that is yours to defend and yours to get better at.

We are not trying to replace the recruiter. We are trying to make the recruiter the sharpest one in the room.

It works on any role

None of this is tied to one industry. The engine reasons about skills, adjacency, timing, and fit, so it works on the role in front of you, whatever it is. It happens to be unusually good on the hard, specific, onsite roles the generalist tools fumble, but the machine does not care what you throw at it.