When it's worth having a person review a result by hand

Not every result needs a manual review, but some do deserve one. It’s worth looking at a case in depth when the integrity signals suggest context, when the decision is high-impact, or when a result doesn’t fit with the rest of what you know about the candidate. The technology tells you where to focus attention; the person decides what to do with that case.

The balance: neither all nor nothing

There are two unhelpful extremes. One is manually reviewing every result, which overloads the team and adds no value in clear-cut cases. The other is never reviewing and treating the result as an automatic verdict, which would be unfair and lacking in rigor.

The healthy point is in the middle: use the signals and the comparable result to identify which cases deserve a second look, and reserve human time for those. That way manual review stops being a burden and becomes a tool for judgment.

When it’s worth a deeper review

These are situations where it’s worth having a person dig into the case:

• Notable integrity signals: a response time or latency that draws attention, or facial behavior signals that call for context.
• High-impact decisions: critical roles, positions of trust, or processes where a mistake is costly.
• A result that contradicts the rest: when the assessment doesn’t fit with the interview, the résumé, or references.
• Borderline cases: candidates very close to the threshold, where a bit more context changes the reading.

When it isn’t needed

Just as important is knowing when manual review adds nothing:

Review in depth	Trust the comparable result
Signals that call for context	A completion with no notable signals
High-impact decision	A broad initial screening stage
A result that doesn’t fit	A result consistent with the process
Borderline case near the threshold	A clear case, far from the threshold

In clear cases, the comparable result is already a good support. Insisting on reviewing everything only dilutes the attention the cases that do matter need.

Why the final decision is always human

Kokoro’s integrity signals—response time, snapshots with consent, facial behavior analysis—exist to support human review, not to replace it. They are observations that provide context; they are not a diagnosis or an automatic disqualification.

That’s why manual review isn’t a patch: it’s part of the design. The technology organizes, prioritizes, and provides comparable signals; the person brings the judgment and makes the decision. That combination is what makes the process both rigorous and fair.

In short

You don’t need to review every result by hand, but you do need to review the ones that matter: cases with notable signals, high-impact decisions, results that don’t fit, or candidates on the edge. The integrity signals indicate where to focus attention, but they never decide on their own. The technology prioritizes; the person decides. See how it works or learn about the product.