HR & Recruiting Copilot
Screen smarter, interview better, and surface internal mobility your TA team would have missed.
The problem
Your TA team receives 4,200 applications a month. They screen 18% of them carefully and reject the rest in 30 seconds. Hiring managers complain that shortlists arrive late and miss the strongest candidates. Internal mobility is non-existent because nobody has a single view of who has the skills you are about to recruit externally.
LinkedIn's 2024 Future of Recruiting report finds that AI-augmented TA functions in APAC reduce time-to-shortlist by 47% and increase quality-of-hire scores by 23 percentage points over 12 months.[^1] McKinsey adds that internal-mobility AI saves mid-market enterprises an average US$18,000 per filled role versus external hiring.[^2]
Our approach
Sources: ATS (Greenhouse / Workday / Lever) + LinkedIn + referrals
│
▼
Application ingest (webhook → Laravel queue)
│
▼
Screening engine
- JD parsing → structured requirements
- resume parsing → structured candidate profile
- LLM scoring (Claude Sonnet 4.6) against requirements
- bias detection layer (Llama Guard 3 + custom rules)
│
▼
Shortlist proposal → human recruiter review
│
▼
Interview prep
- role-tailored question generation
- rubric per question
- prep brief for hiring manager
│
▼
Post-interview
- feedback summary draft (recruiter edits)
- debrief synthesis across panel
- hire/no-hire recommendation
│
▼
Internal mobility scanner (cross-references skills graph against open roles)
│
▼
ATS sync (Greenhouse / Workday / Lever / SmartRecruiters)
Who it is for
- The CHRO of a 1,000-person Korean retail group with 60+ open roles and a 10-person TA team underwater.
- The Head of People at a 400-person Singaporean SaaS company trying to lift the quality of engineering hires while cutting time-to-offer.
- The Head of Talent at a 700-person Indonesian fintech building an internal-mobility program from scratch.
Tech stack
- LLMs: Claude Sonnet 4.6 for reasoning-heavy tasks (screening, interview design), Claude Haiku 4 for high-volume parsing
- Skills graph: Custom-trained skill-extraction model (open-weights) over employee profiles + job descriptions, stored in pgvector
- Bias detection: Llama Guard 3 for prompt-injection defence, plus AIMenta custom rule set for protected-characteristic mentions in screening reasoning
- Backend: Laravel 12 + Filament 3 admin UI, queue workers for async ATS sync
- ATS connectors: Greenhouse, Workday, Lever, SmartRecruiters, Manatal, JobAdder, BambooHR
Integration list
Greenhouse, Workday HCM, Lever, SmartRecruiters, Manatal (popular in Hong Kong / Southeast Asia), JobAdder, BambooHR, LinkedIn Recruiter (via partner connector), Slack, Microsoft Teams, Outlook calendar, Google Calendar, DocuSign for offer letters.
Deployment timeline
| Week | Activity |
|---|---|
| Week 1 | TA workflow audit; pick 3 priority roles for pilot; bias rule baseline |
| Week 2 | ATS integration; resume parsing tuned per role family |
| Week 3 | Screening LLM prompt tuned against 200 historical applications per role |
| Week 4 | Shadow mode launch (LLM scores in parallel; recruiter still decides) |
| Week 5 | Cutover on pilot roles; recruiter review queue live |
| Week 6 | Add interview-prep and feedback summary modules |
| Week 7-8 | Roll out to remaining role families; train hiring managers |
Mini-ROI
A 400-person SaaS company in Singapore deployed the Copilot for engineering hiring in week 6 of 2025. Time-to-shortlist dropped from 11 days to 4. Hiring-manager satisfaction with shortlist quality rose from 6.2/10 to 8.4/10. Quality-of-hire (90-day performance review) lifted 27%. Annualised TA capacity equivalent to 2 additional recruiter FTEs.
LinkedIn's benchmark for AI-augmented TA productivity is US$24,000-US$48,000 saved per recruiter annually, plus 15-30% reduction in cost-per-hire across 12 months.[^1]
Pricing tiers
| Tier | Setup (one-time) | Monthly run cost | Best for |
|---|---|---|---|
| Starter | US$22,000 - US$38,000 | From US$1,400/mo | Single ATS, 1-2 role families, screening + interview prep modules. |
| Scale | US$48,000 - US$95,000 | From US$3,600/mo | 4-8 role families, internal mobility scanner, multi-language outreach. |
| Strategic | US$110,000 - US$220,000 | From US$7,200/mo | Enterprise-wide, custom skills graph, succession-planning module, DEI dashboards. |
All tiers include a quarterly bias-audit refresh and ongoing prompt tuning.
Frequently asked questions
Will the screening introduce bias? The risk is real and we engineer for it explicitly. The screening prompt is reviewed by a TA leader and an external DEI advisor before launch. Protected characteristics never appear in scoring inputs. We measure bias quarterly against placement outcomes by gender, age cohort, and (where collected) ethnicity. Any drift triggers a prompt review.
Is this compliant with regional labour and privacy law? Yes. We deploy with privacy notices and consent capture aligned to Singapore PDPA, Japan APPI, Korean PIPA, Indonesia UU PDP, and Hong Kong PDPO. We do not use candidate data for model training. We retain candidate data per your retention policy.
What if a hiring manager disagrees with the score? The score is a recommendation, not a gate. Recruiters always have final say. We track score-vs-decision divergence and surface patterns to the TA leader monthly — this often reveals real role-criteria mismatches the score caught and the manager intuited.
How do you handle resumes in mixed languages or non-Latin scripts? Resume parsing uses Claude Sonnet 4.6 which handles Japanese, Chinese, Korean, Vietnamese, Thai, Bahasa, and English natively. Mixed-language resumes (e.g., English with Japanese certifications) parse cleanly. Field-level extraction accuracy in our last benchmark sat at 94.6%.
Can we connect this to LinkedIn Recruiter? Yes, via partner connector (LinkedIn restricts direct API access). We pull candidate profiles into the Copilot and write enriched notes back to LinkedIn Recruiter projects.
What does the internal mobility scanner do? It builds a skills graph from your existing employees (via HRIS profiles, performance reviews, internal projects) and matches against open requisitions. When a 75%+ skill match exists internally, the scanner alerts the TA team and the employee's manager (with the employee's consent).
Can hiring managers run their own searches? Yes. Filament 3 dashboards give hiring managers self-service views of applications, scored shortlists, and interview prep — without recruiter intervention. Most clients see hiring manager NPS rise 18+ points within 60 days.
How do you handle data residency for cross-border hiring? Candidate data stays in the region of the requisition. For cross-border placements (e.g., Singapore role with Vietnam-based candidates), we follow the regulatory framework of the higher-residency-bar jurisdiction. Data flow diagrams ship with the deployment.
Where this is most often deployed
Industries where AIMenta frequently scopes this kind of solution.
Beyond this solution
Browse our other productized solutions, plus the verticals and Asian markets where they ship.
Other solutions
By industry
By Asian market
Frequently asked questions
How does the copilot screen CVs without introducing hiring bias?
The model is trained on role-relevant competency frameworks rather than demographic proxies. Names, gender indicators, graduation years, and university rankings outside the job's geographic scope are masked during initial scoring. A bias audit report (disparity analysis by gender and nationality) is generated monthly and reviewed with your HR team.
Can the copilot conduct first-round interviews autonomously?
It can conduct asynchronous video or text-based screening interviews using structured question sets you define. Candidate responses are scored against the competency rubric and ranked for recruiter review. We recommend human-led second-round interviews for all shortlisted candidates — the copilot's role is to surface the right 15% from hundreds of applicants, not to make final hiring decisions.
How does the platform handle data privacy for job applicants across APAC?
Applicant data is stored in jurisdiction-scoped infrastructure (HK, SG, or JP nodes depending on the hiring entity). Retention periods comply with local labour regulations — typically 12 months after rejection under PDPO (HK) and PDPA (SG). Candidates can request deletion; the platform provides a compliant purge workflow that removes PII while retaining anonymised aggregate analytics.
Related solutions
Don't see exactly what you need?
Most engagements start as custom scopes. Send us your problem; we'll tell you whether one of our productized solutions fits — or what a custom build looks like.