Will they actually take the offer? Stop scoring qualification, start scoring intent.

A candidate aced every interview round. The panel was unanimous. The offer went out the next morning.

By Friday, they had accepted a competing offer.

If this has happened to you, you already know the math. Three rounds of panel time, four panelists per round, an hour each. Plus the recruiter coordinating, the hiring manager debriefing, the offer prep. Roughly thirty person-hours sunk into a candidate who was never going to take the role.

Most hiring tools do not catch this. They optimize for the wrong question.

The two questions, and why they get conflated

There are two distinct questions you can ask about a candidate:

1. Can they do the job? (Qualification.)
2. Will they take the offer if we make one? (Intent.)

These are unrelated dimensions. A senior engineer who's bored at her current job and excited about your mission scores high on both. A senior engineer who's two weeks from a competing offer they're going to accept scores high on (1) and low on (2). The pipeline cost of mistaking the second for the first is enormous.

Most scoring systems on the market today, including the "AI-powered" ones, only score (1). They tell you the candidate is qualified. They do not tell you the candidate is coming.

What an intent signal actually looks like

The data you need is already in the conversation. The candidate told you, in their own words, things like:

"I'd want to give my current team a proper handoff, so probably six weeks notice."

That's a notice period signal. Six weeks is fine. Twelve weeks would be a flag.

"I'm pretty far along with one of the other processes I'm in."

That's a competing-offer signal. Now you know they're not exclusive.

"Honestly the comp would need to land north of one-fifty for me to leave my current role."

That's a salary signal. If your band tops out at one-forty, you have a problem nobody has named yet.

These are not subtle cues. They're explicit statements the candidate made, in their voice, on the record. The issue is not that the data is hidden. The issue is that nobody is structuring it.

Why three pillars, not one number

A single overall score hides too much. We organize the signal into three pillars:

Stated Intent (40 points). What they said about timing, money, motivation, competing processes, and whether they had questions for the team. Notice period, salary alignment, push vs pull motivation, and how engaged they were when asked about your specific role.

Behavioural (30 points). What they did. How quickly they completed each stage. Whether they followed up. Whether they showed up to scheduled calls. Whether they responded inside business hours or at 11pm on a Sunday.

Linguistic (30 points). How they talked. Hedge words ("I think I could be open to..."), specificity when describing why your company in particular ("...the way you've structured the engineering org around..." vs "...sounds like a great opportunity..."), and enthusiasm markers.

Each pillar carries its own evidence. When the system flags a candidate as low-intent, it can show you the exact sentence the candidate said that drove the score. That matters because:

• Hiring managers want to argue the score. Letting them see the citation turns the argument from "I disagree with your AI" to "let's look at what she actually said".
• Recruiters need to defend the call. "We deprioritized this candidate because she said she's pretty far along with another process" is defensible. "Our AI said no" is not.

What this changes in the pipeline

Once you have a defensible intent score after the first conversation, three things change.

You stop scheduling panels for low-intent candidates. The panel time you reclaim, you spend on candidates with both high fit and high intent. Your panel-to-offer rate goes up because the funnel narrowed in the right place.

You start having different conversations with high-intent borderline-fit candidates. They're worth bending the role around. You'd be surprised how often the candidate you almost rejected for fit was the one who would have actually accepted.

Your offer-acceptance rate becomes a real number, not a vanity metric. When you only extend offers to candidates the signal said would accept, your acceptance rate climbs and stays there. That stat is the one your VP cares about.

What you should not expect

A score is a tool, not a verdict. Intent signals are calibrated against past pipeline outcomes; they get sharper as your data grows. They are wrong on individual candidates more often than the marketing copy of any vendor (us included) will admit.

What they will not be wrong about is the trend. Across your pipeline, the high-intent candidates accept at a much higher rate than the low-intent ones. Stop investing time at the same level in both groups and your time-to-hire collapses without your acceptance rate moving.

That's the trade. You give up the comfort of "every candidate gets full investment". You get back the weeks of your team's life that were going into candidates who were never coming.

The next question to ask in your pipeline

Forget AI. Forget tooling. The next time you debrief a hire that did not take the offer, look at the transcript of the first conversation you had with them. The signal was almost certainly there, in their own words, in the first fifteen minutes.

If it wasn't, you weren't asking the right questions. That's the easier fix.

If it was, and you missed it, that's the harder one. The harder one is the one this product solves.