AI-Driven Talent Acquisition: Enhancing Recruitment Efficiency Through Intelligent Screening and Candidate Matching
Annotatsiya
Talent acquisition is the essential first step of human capital formation within an organization-but talent acquisition at most organizations involves high volume (thousands of applications for each position), high risk (bad hires can be 30-150% of the employee's annual salary), and high potential for bias (decades of research confirm the prevalence of demographic bias in hiring, even by trained recruiters). This paper proposes an AI-Driven Talent Acquisition Framework (ATAF) integrating an intelligent résumé parsing and multi-modal candidate profiling engine, a transformer-based candidate matching and predictive hire quality module, and a bias detection, DEI analytics, and regulatory compliance gateway, evaluated across 38 organisations (enterprise$\mathbf{n}=\mathbf{1 4}$, mid-market$\mathbf{n} \boldsymbol{=} \mathbf{1 6}$, SME$\mathbf{n} \boldsymbol{=} \mathbf{8}$) processing 4,480,000 candidate applications (3,136,000 training / 1,344,000 test, 70/30 stratified split) and 186,400 job postings across 6 industry sectors over a 37-month evaluation horizon. The ATAF achieves a composite recruitment effectiveness score of r = 0.941 against expert-validated hiring outcome benchmarks, outperforming ML-only applicant tracking systems (r = 0.828; +13.6% [(0.941-0.828)/0.828]), rule-based keyword-matching ATS$(\mathbf{r}=\mathbf{0. 6 2 4}; +\mathbf{5 0. 8} \boldsymbol{\%}[(\mathbf{0. 9 4 1}-\mathbf{0. 6 2 4}) / \mathbf{0. 6 2 4}])$, and manual recruiter screening$\quad(\mathbf{r}=\mathbf{0. 5 2 6}; \quad+\mathbf{7 8. 9} \% \quad[(\mathbf{0. 9 4 1}-\mathbf{0. 5 2 6}) / \mathbf{0. 5 2 6}])$. Deployment outcomes: -43.2% time-to-hire reduction, +58.2% résumé screening accuracy, +42.2% candidate quality score, +28.2% hiring manager satisfaction, +33.8% DEI hiring improvement, and +21.8% offer acceptance rate. Sub-domain performance: Intelligent Résumé Screening$(\mathbf{r}=\mathbf{0. 9 6 2})$, Candidate Matching and Ranking (0.957), Predictive Hire Quality (0.948), and Bias Detection and DEI (0.930). Algorithmic bias (from demographically skewed historical training data) degrades ATAF accuracy to 0.894 (-5.0% [(0.941-0.894)/0.941]), and sparse candidate data in niche professional roles to 0.874 (-7.1% [(0.941-0.874)/0.941]), with mitigation protocols for both conditions.
Hali tarjima qilinmagan