What is Explainable Talent Intelligence? Explainable AI recruiting, defined
Explainable Talent Intelligence is an approach to AI sourcing in which every candidate recommendation comes with reasons a recruiter can read, check and contest, instead of a bare match percentage. It is built for the transparency duties of recruitment AI, which Annex III, point 4 of the EU AI Act (Regulation (EU) 2024/1689) classes as high-risk.
Definition reviewed on
How does Explainable Talent Intelligence work in practice?
What does a recruiter see next to each recommendation?
Explainable Talent Intelligence replaces a bare percentage with reasons. In TrueCalling, each candidate recommendation shows a score from 0 to 100 together with the skills that match the job description, the skills that are missing, and the gaps that weigh on the result, ranked as major or minor. A short written summary states the verdict in plain language, and a "Why this score?" action produces a longer explanation on demand. The rules are visible too: a candidate missing three or more required skills cannot score above 50 (TrueCalling internal data, 2026), so a high score cannot hide a missing must-have. The method has a limit: it explains the model's reading of a CV, and a CV is an incomplete record of a person, so recruiters still check what an explanation relies on before acting.
Why do regulators ask for explanations in hiring?
Explainable Talent Intelligence follows a legal trend, not only a design preference. The EU AI Act lists AI systems used to recruit, filter applications or evaluate candidates as high-risk in Annex III, point 4, and the AI Omnibus, in force since 27 July 2026, applies those rules from 2 December 2027 (European Commission, July 2026). Article 86 gives people affected by such decisions a right to clear and meaningful explanations of the AI system's role. The GDPR already grants, in Article 22, a right not to be subject to decisions based solely on automated processing. In New York City, Local Law 144 has required bias audits of automated employment decision tools since 5 July 2023 (NYC DCWP). Explanations alone do not make a tool compliant: the employer keeps its own duties, set out in our AI Act recruiting compliance guide.
How can an explanation be audited later?
An explanation helps only if it can be found again. In TrueCalling, every AI decision that touches candidate data is logged with the exact model version, a prompt identifier, the input and the output, and kept for 6 months by default (TrueCalling internal data, 2026). That record lets a team answer a candidate's question weeks later with the reasoning that applied at the time, not a fresh one, and the log can be exported for a review. Explanations also make bias testing possible: when the reasons behind scores are visible, a team can check whether a proxy, such as a career gap, pushes a group down and then change the rule. Our guide to AI bias in recruiting describes the tests. The limit: a log proves what the system did, not that it was fair.
How does Explainable Talent Intelligence differ from black-box scoring?
| Legacy approach | Explainable Talent Intelligence (TrueCalling) |
|---|---|
| A match percentage with no supporting reasons. | Score with matched skills, missing skills and ranked gaps. |
| Reasons reconstructed afterwards, if anyone asks. | Written summary at scoring time; full explanation on demand. |
| No trace of which model produced a score. | Model version and prompt identifier logged per decision. |
| Bias reviewed on final hiring outcomes only. | Visible reasons let teams test whether a proxy drives scores. |
Related terms
- TrueFit 360
TrueCalling's candidate fit score: 0 to 100, with matched skills, gaps and a written reason.
- Autonomous Sourcing Copilot
AI agent that runs brief → shortlist → first contact while the recruiter approves each action.
- Semantic Talent Discovery
Search method that reads a job brief for meaning, not keywords, and finds profiles Boolean misses.
FAQ: Explainable Talent Intelligence
Is explainable AI required for recruiting tools in the EU?
For high-risk uses, yes, from 2 December 2027. The EU AI Act classes recruitment AI as high-risk: such systems must be transparent enough for users to interpret their output and must allow human oversight. Article 86 adds a right to an explanation of the AI's role in a decision.
Does an explanation mean the AI's recommendation is correct?
No. An explanation shows how the model reached its output, so a recruiter can spot a wrong reading of a CV, but it does not prove the output right. Explainable Talent Intelligence therefore keeps a human decision after every score and records what the model saw.
Can a candidate ask why an AI ranked them low?
In the EU, yes. The GDPR's right of access (Article 15) covers the personal data used and, for solely automated decisions, meaningful information about the logic involved. Article 86 of the AI Act adds a right to an explanation for decisions based on high-risk AI.
Does Explainable Talent Intelligence remove bias from hiring?
No. Explanations make bias easier to detect, not impossible. A model can still learn from skewed data or reward proxies, such as uninterrupted careers. Explainability adds evidence: visible reasons and logged decisions let a team compare scores across groups and change the rule behind a gap.
Sources
- Regulation (EU) 2024/1689 (AI Act): Annex III point 4, Articles 12, 14 and 86 (EUR-Lex) (accessed 5 Oct 2026)
- European Commission, "AI Omnibus enters into force" (27 July 2026) (accessed 5 Oct 2026)
- Regulation (EU) 2016/679 (GDPR): Articles 14, 15, 17, 20 and 22 (EUR-Lex) (accessed 5 Oct 2026)
- NYC Department of Consumer and Worker Protection: Automated Employment Decision Tools (Local Law 144) (accessed 5 Oct 2026)
- TrueCalling internal data, 2026: platform figures and product behaviour (not third-party benchmarks) (accessed 5 Oct 2026)
- TrueCalling candidate privacy notice (accessed 5 Oct 2026)
See Explainable Talent Intelligence in action
30-minute demo of TrueCalling against one of your open roles: sourcing, scoring, and outreach end-to-end.