Most AI screening tools hand you a number. Candidate A scores 82, candidate B scores 61, and you are expected to move on. The question nobody asks loudly enough is: 82 out of what, and why? If the tool cannot answer that, you are not screening with AI. You are trusting it, blind, and hoping it agrees with you.
That distinction, between a score you can see inside and a score you cannot, is the single most important thing to understand before you buy or rely on any CV screening tool. It is the difference between a black box and a glass box.
What a black box actually costs you
A black box gives an output with no reasoning. It feels efficient, until the moment you need it to be more than a number. Then the hidden costs surface:
- You cannot defend the decision. When a client asks why a candidate was not put forward, or a candidate asks why they were rejected, "the system scored them low" is not an answer you want to give.
- You cannot catch mistakes. If you cannot see the reasoning, you cannot see when it is wrong. A strong candidate scored low for a bad reason looks identical to one scored low for a good reason.
- You cannot improve. Without knowing what drove a score, you cannot refine your criteria. You are guessing at a dial you cannot see.
- You carry compliance risk. UK rules expect a human to be meaningfully involved in hiring decisions. A reviewer who cannot see the reasoning cannot meaningfully review anything, a point we cover in our guide to ICO-compliant AI screening.
A glass box shows its working
Explainable, or glass box, screening gives a plain-English reason alongside every score. Instead of "82," you see something like: strong match on required sector experience, ten years relevant and recent, missing one preferred qualification, so ranked highly but not top. You can read it, question it, and repeat it to anyone who asks.
That transparency is not a cosmetic feature. It changes what the tool is for. A black box replaces your judgement. A glass box informs it, and leaves you in charge.
Why explainability makes you better, not just safer
The compliance case for explainability is real, and the UK regulator, the Information Commissioner's Office, is clear that people can ask how an automated decision about them was reached. But the stronger argument is about quality, not just cover.
When you can see why each candidate ranked where they did, you catch the tool's errors instead of inheriting them. You override with confidence, because you know exactly what you are correcting. You spot when your own criteria were too narrow. And you can walk a client through your shortlist with reasoning that holds up, which is often the difference between winning the brief and losing it. Explainability turns AI from something you quietly hope is right into something you can actually manage.
How to tell a glass box from a black box before you buy
Ask any vendor these questions, and watch how quickly they answer:
- Can I see why each candidate scored what they did, in plain language, for every result and not just the top few?
- Can I override the ranking, and does the tool make that easy?
- What happens when a CV is unclear? Does it flag it for review, or quietly guess a score?
- Can I export the reasoning to share with a client or keep on file?
If the honest answers are vague, you are looking at a black box, whatever the marketing says. Our guide to what to look for in AI CV screening software goes further on this.
Where Lucuma fits
Lucuma is built as a glass box on purpose. Every candidate gets a score and a plain-English reason for it, across the whole batch, not just the shortlist. You can audit any result, override it, and export the reasoning. When a CV lacks the detail to judge fairly, Lucuma flags it for review rather than inventing a score, so nothing strong is silently dropped. It is the same principle behind reading real evidence instead of keywords.
The fastest way to understand it is to look. Our sample shortlist shows exactly what an explained score looks like on the page, reasoning and all.
Frequently asked questions
What is black box AI in recruitment?
A black box tool produces an output, such as a score or a ranking, without showing the reasoning behind it. You get a number but no explanation of why a candidate scored that way, which makes the result impossible to check, defend, or improve.
What does explainable AI CV screening mean?
Explainable, or glass box, screening gives a plain-English reason for every score. You can see which parts of a candidate's experience drove the ranking, question it, override it, and repeat it back to a client or candidate. The reasoning is visible, not hidden.
Why does explainability matter for recruiters?
It lets you catch the tool's mistakes, defend decisions to clients and candidates, keep meaningful human control for compliance, and actually trust your shortlist. A score you cannot explain is a score you cannot stand behind.
Can I override an AI shortlist?
With an explainable tool, yes. Because you can see the reason for each score, you can apply context the AI does not have and change the ranking with confidence. The AI ranks and explains; you make the final decision.