HRTech // Client Project

HR Policy Support:
85% of Questions Answered Automatically.

We built a practical HRTech solution combining semantic retrieval system, multi-lingual AI-powered logic, and personal data-gated audit trails. The project delivered 85% Query Resolution and 12k Hours Saved.

Key Results RESULTS
85%
Query Resolution
Manual Support
12k
Hours Saved
ROI: 14 Weeks
2.4s
p95 Response
SOC2 GATED

Trusted by teams building ambitious products

What We Set Out to Improve.

Industry
HRTech

Distributed workforce across 14 countries requiring instant, localized policy clarity in 8 languages.

How We Worked
Dedicated product team

AI Architect + Language AI specialist + MLOps Lead embedded within Global People Operations.

Goal
Day-to-day Efficiency

Eliminating manual HR prioritization for repetitive benefits, tax, and compliance queries.

What We Built
RAG Agent Platform

A practical solution designed around the client’s existing tools, teams, and day-to-day workflow.

What Was Getting in the Way.

The client’s HR knowledge was trapped in 1,200+ static PDF policies, local law addendums, and disconnected Slack histories. For 20,000 employees, finding specific benefits answers required manual HR intervention, resulting in 48-hour ticket backlogs and shadow work for managers.

The biggest issues were hr burnout, policy non-compliance, and personal data exposure. The team needed a faster, clearer way to manage the work while keeping the right checks in place.

HR Burnout
Repetitive query volume led to high turnover in the support tier and delayed business People initiatives.
Policy Non-Compliance
Inaccreate local answers from unverified peer chats created legal risks in regulated labor markets.
Personal Data Exposure
Older chatbots lacked reliable data masking, risking the exposure of sensitive employee data during training/inference.
The Solution

What We Built.

01
Semantic Retrieval System
Added an advanced RAG (Retrieval-Augmented Generation) workflow using semantic chunking to ensure policy context is preserved across languages.
Key details
Storage Vector Native
Chunking Context Aware
Response time Sub Second
02
Multi-Lingual AI-powered Logic
Introduced agents capable of cross-referencing global master policies with local labor law addendums to provide localized, accreate answers.
Key details
Type AI Assisted
Language AI Cross Lingual
Verification Fact Checked
03
Personal Data-Gated Audit Trails
Every employee interaction is scrubbed of personal information before persisting in secure and traceable logs for quality auditing and policy drift analysis.
Key details
Security SOC2 Ready
Privacy personal information Masking
Controls Cloud Optimized
Before and After

How the Workflow Improved.

Area
Before
After
User Experience

Keyword Search

Employees manually sifted through long PDFs to find specific clauses.

Natural Inquiry

Context-aware agents provide direct answers with citations to the source policy.

Compliance

Manual Updates

Policies often out-of-sync with local law changes across different regions.

Automated Sync

Automatic updates to the vector store ensure the AI always has the latest version.

Response Time

24-48 Hours

Support tickets queued behind manual prioritization and regional timezone gaps.

Instant (<3s)

Automated resolution as the business grew, available 24/7 across all global regions.

Key Features

What Made the Solution Useful.

AI Assisted

Self-Correcting RAG

The agent performs a Fact-Check loop, comparing the generated answer against the source data before responding to the user.

Business impact
Zero Hallucination
Traceable Data

Citations & Transparency

Every answer includes a Deep Link to the specific paragraph in the official handbook, building employee trust and a clear audit trail.

Business impact
100% Policy Accuracy
Live Updates

Early Notification

System identifies common query spikes (e.g., tax season) and earlyly suggests policy updates to the HR leadership team.

Business impact
65% Ticket Reduction
Faster Delivery

How We Reduced Build Time.

Tested foundations helped the team spend less time on setup and more time on the parts that made this product useful.

What accelerated the work

4 reusable building blocks
01

Knowledge Platform Module

A tested starting point for knowledge platform module reduced repeated setup work.

02

Personal Data Masking Platform

Reusable work for personal data masking platform let the team focus more time on the client’s specific needs.

03

AI Monitoring Platform

This made it easier to add ai monitoring platform without rebuilding common foundations.

04

Multi-Lingual Bridge

A tested starting point for multi-lingual bridge reduced repeated setup work.

Results

The Business Difference.

A straightforward before-and-after view of what changed for the team and their customers.

RESULT: AUTOMATION01

Auto-Resolution Rate

Percentage of employee queries resolved fully by the AI without escalating to an HR agent.

Older BaselineManual
Coretus Platform85% Auto
Outcome85% Resolution Rate
RESULT: OPS EFFICIENCY02

HR Support Capacity

Reduction in manual ticket volume, allowing HRBPs to focus on high-impact higher-value roles.

BeforeCongested
After65% Lower
Outcome65% Less Support Load
RESULT: RELIABILITY03

Response Response Time (p95)

High-concurrency streaming ensures answers are delivered instantly even during peak benefit windows.

Target< 5.0s
Coretus2.4s
Outcome2.4s p95 Response Time
Results85% Resolution Rate • 65% Less Support Load
Trust and Control

How We Kept It Safe and Reliable.

01
Data Masking Protocol
All prompts undergo live Removal of unnecessary personal data to ensure no employee identity data reaches the LLM providers.
Business SECURE
02
Reason Code Lineage
The system logs the exact policy version and chunk ID used to generate every response for legal verification.
VERIFIABLE DATA
03
Growth
Cloud-based inference pods ensure reliable scaling for 20,000+ concurrent staff members.
Ready to Grow
04
Code and IP Ownership
Coretus provides 100% IP ownership of the Knowledge Platform, vector embeddings, and custom agent logic.
100% OWNED
Client Testimonial

In their own words.

Coretus didn't just build a chatbot, they built a multi-lingual Knowledge Platform that understands the nuances of our global policies. It resolved 85% of queries instantly, giving our HR team 12,000 hours back for higher-value work.

Scale HR Support with AI assistants.

Have a similar challenge? We can help you plan and build a practical HRTech solution around your goals, budget, and existing systems.

personal information-Gated RAG System design

Multi-Lingual Policy Consistency

100% IP & Model Ownership