Explainable AI in hiring
Explainable AI in hiring means the software shows why each candidate got the score they did: which requirements they met, which they missed, and how those added up. If a tool gives you a number and no reasons, you cannot check it, correct it, or defend it to a candidate or a regulator.
Why a score alone is not enough
Older applicant tracking systems filter on exact keywords, so a strong candidate who words their experience differently gets missed. Newer AI tools can read for meaning, which helps, but many return only a match percentage. That moves the opacity from the keyword list into the model.
A hidden score creates three problems. Recruiters cannot tell a good rejection from a bad one. Hiring managers do not trust a ranking they cannot interrogate. And when a candidate or regulator asks why someone was screened out, there is nothing to show.
What good explainability looks like
- A per-candidate breakdown. Requirements matched, requirements missed, and how the score was calculated, in plain language.
- Criteria your team controls. The score is measured against the role you wrote, not a generic profile.
- Excluded inputs. Name, age, photo, school prestige and demographic signals are not used.
- Human override. A recruiter can change or ignore the score, and the change is recorded.
- No automatic rejection. The AI ranks and flags. People decide.
- An audit trail. Scores and reasons can be exported when someone asks.
An illustrative score breakdown
This is an invented example to show the shape of a useful explanation. It is not a real candidate.
Role: Customer Success Manager. Fit score 78 / 100
| Requirement | Result | Evidence |
|---|---|---|
| 3+ years in B2B SaaS support or success | Met | Four years across two SaaS companies |
| Owns renewals | Partly met | Led renewals for a book of accounts; no quota stated |
| Experience with a CRM | Met | Names a CRM in two roles |
| Spanish fluency | Not shown | Not mentioned; worth asking at screen |
What the rules ask for
Regulators are converging on transparency for automated hiring tools. This is a summary, not legal advice, and the details change, so check the current text for the places you hire.
- New York City Local Law 144. Employers using automated employment decision tools for NYC roles need an independent bias audit each year, a public summary of the results, and notice to candidates before the tool is used.
- EU AI Act. AI used for recruitment and candidate selection is classed as high-risk, which brings requirements around risk management, documentation and human oversight. The Digital Omnibus changes adopted in 2026 deferred the date those obligations apply.
In both cases a tool that already shows its reasoning and keeps a human in the loop is much easier to document than one that does not.
Eight questions to ask any vendor
- For a single candidate, can I see which requirements matched, which did not, and how the score was built?
- Which inputs does the model never use (name, age, school, photo, address)?
- Can a recruiter override the score, and is the override recorded?
- Does the system ever reject a candidate automatically, with no human step?
- Is the scoring criteria set by my team, or is it a generic model I cannot see into?
- How are career gaps and non-linear careers treated?
- Can I export the scores and reasons for an audit or a candidate request?
- What independent bias testing exists, and can I see the summary?
How Jatura approaches it
Jatura scores every applicant against the criteria you set and attaches a plain-English explanation covering strengths, gaps and how the candidate stacks up. Name, age, school prestige and demographic signals are not scoring factors. Career gaps are not penalised. No candidate is auto-rejected, and every AI recommendation can be overridden.
Plans start at $79 a month with unlimited team members and a 14-day free trial, so you can run the checklist above on your own roles. See how the AI applicant tracking system works or pricing.
Questions people ask
What is explainable AI in hiring?
Explainable AI in hiring means the system shows, for each candidate, why it produced a score or ranking: which requirements the candidate met, which they did not, and how those combined into the result. A recruiter can read the reasoning, check it against the resume and override it.
Is AI resume screening legal?
In most places, yes, but it is regulated. New York City Local Law 144 requires an annual independent bias audit and candidate notice for automated employment decision tools used on NYC roles. The EU AI Act classifies AI used in recruitment and candidate selection as high-risk, with obligations that were deferred in 2026. Rules vary by location, so take legal advice for your situation.
Does explainable AI remove bias?
No. Explanations make bias visible so people can catch it, and keeping protected or proxy attributes out of scoring reduces it. You still need human review, regular testing of outcomes, and a way to challenge a decision.
Should AI make the final hiring decision?
No. AI should rank and surface candidates and show its reasoning. People should decide who moves forward, and every AI recommendation should be easy to override.
How does Jatura explain its scores?
Every applicant is scored against your criteria, and the score comes with a plain-English breakdown of strengths, gaps and how the candidate compares. Name, age, school prestige and demographic signals are not scoring factors, career gaps are not penalised, and no candidate is auto-rejected.