Multi-skill interview assessment: how Mochi scores several skills from one answer

For decades, hiring has run on a quiet bargain: the more you want to know about a candidate, the longer you make them talk. We think that bargain is broken. With overlay skills, Mochi reads a single answer through several lenses at once, so candidates spend less time proving themselves and recruiters walk away with more signal than before.

Ask any experienced recruiter what they learned from a good first conversation and you'll get a long answer. They'll tell you how the candidate handled a tricky question. They'll also tell you how clearly they explained themselves, how they talked about their old team, whether they seemed at ease in the language the job requires.

None of that came from a dedicated question. It came from listening.

Assessment, for most of its history, has worked the other way around. One question, one skill, one score. If you wanted to measure four things, you asked four sets of questions, and the candidate paid for every one of them in time and patience.

We set out to build an interview that listens the way that recruiter does.

Why do interviews keep getting longer?

Most screening processes grow by addition. A hiring manager asks for teamwork, so a teamwork section appears. The role needs Italian, so an Italian section follows. Each request is reasonable on its own. Together, they turn a fifteen-minute conversation into a gauntlet.

The structure made sense when assessment meant paper tests and human graders. A grader could only score what a question was designed to capture, so every skill needed its own question. That logic quietly carried over into digital assessments, then into AI interviews.

The cost lands on the candidate. They answer a question about customer service in Italian, then answer more questions in Italian just to prove they speak Italian. It feels repetitive because it is.

In high-volume hiring, repetition is expensive. Every extra minute is a chance for a strong candidate to close the tab. Some of our customers went as far as running two separate screenings for the same role: one for the core skills, one for the language. Two invitations, two completions, two chances to lose someone good.

How does multi-skill interview assessment work?

A real answer is never about one skill. When a candidate describes the day they calmed down a furious customer, they're showing you their judgment under pressure. They're also showing you how they structure a story, how they speak about other people, and how comfortably they move through the language.

The candidate answers once. The recruiter sees every dimension.

Does a shorter interview mean a weaker assessment?

No. Overlay skills are one of the rare cases where candidate experience and assessment quality improve at the same time, instead of trading off against each other.

Shorter doesn't mean less rigorous, because length was never what made an interview predictive. Structure is. When researchers went back over decades of hiring studies, the result surprised a lot of people:

Overlay skills keep that structure intact. Every skill is still scored against its own criteria; it just no longer needs its own questions to get there.

The signal itself often improves, too. A dedicated language question invites a rehearsed answer. Language assessed across a whole interview captures how someone speaks when their attention is on the substance of the conversation, under a little pressure, thinking on their feet, much closer to how they'll use the language with a real customer on a busy Monday morning.

The same principle extends beyond language. Some skills, like communication, show up in every answer. It never made sense to measure them in a box of their own.

How do we validate each overlay skill?

More signal is only valuable if every signal holds up. Getting more out of each answer raises the bar on how carefully we read it.

Each lens is built and validated by our Science team before it scores a single real candidate. Language proficiency goes through the same process as every competency grid we ship: a dedicated evaluator, tested against its own validation dataset. Until a lens has earned that, it doesn't produce a score.

Every score also shows its working. On the report, each overlay skill states exactly where it was scored from: specific moments in the conversation, or the interview as a whole. A recruiter or hiring manager reviewing a result can see what it rests on, which is what human oversight of AI in hiring should look like in practice.

We also kept the rules simple. Each skill has one clear source. If a recruiter changes where a skill is scored from, results before and after the change are kept apart, so nobody compares numbers that measured different things.

What does this change about how we design interviews?

The principle behind overlay skills is simple, and it shapes how we build everything at Maki: ask candidates only what you need to ask, and learn everything you can from what they tell you.

That changes the question hiring teams ask when they design an interview. It used to be “what else should we test?”, with each answer adding length. Now it can be “what else can we learn from this conversation?”, and the answer costs the candidate nothing.

It's the same instinct behind skills-based hiring: stop asking people to prove themselves twice, and start trusting what they've already shown you.

We think this is where AI in hiring gets genuinely interesting. Automating the old process faster was the first step. The real opportunity is designing assessments that would have been impossible before: conversations that feel natural to the person having them, and still give recruiters a deep, multi-dimensional view of their potential.

The best recruiters have always listened this way. Now every candidate gets that kind of attention.

FAQ