AI-Powered Internal HR Knowledge & Talent Search Chatbot
An internal chatbot that searches HR policy and talent profiles so recruiters find answers in seconds.
Key Details
| Challenge | HR teams could not find policy answers or candidate skills quickly across fragmented drives and ATS notes. |
|---|---|
| Solution | Axtra Labs built a RAG chatbot over HR documents and talent profiles, with access control and citations. |
| Technologies | Azure OpenAI, LangChain, SharePoint, Elasticsearch |
Technologies used
Azure
OpenAI
LangChain
SharePoint
Elasticsearch
Python
Client background
A multi-country professional-services firm with 4,000+ employees managed HR policy across SharePoint libraries, email threads and an aging ATS. Recruiters spent hours hunting for answers, and people-ops could not trust a single source of truth when policy changed.
Key challenges
- Policy PDFs, FAQs and ATS notes lived in separate systems with inconsistent permissions.
- Recruiters asked the same questions repeatedly; answers depended on who was online.
- No citations — employees could not verify whether an answer matched the current handbook.
- Talent search was keyword-only, so skill matches and past project experience were easy to miss.
What we built
- RAG chatbot over SharePoint HR docs and curated talent profiles, with role-based access.
- Citation panel that links every answer back to the source paragraph.
- Talent search that ranks candidates by skills, projects and availability signals.
- Admin console for ingesting new policies and retiring outdated versions.
Project team: 6 engineers across AI/ML, backend and domain specialists — delivery over 14 weeks.
How we delivered
Discover
Mapped HR workflows, permission boundaries and the highest-volume question types.
Ingest
Normalized SharePoint and ATS exports into a searchable index with ACL metadata.
Pilot
Rolled out to HR ops first, then recruiters, with feedback loops on answer quality.
Harden
Added eval sets, access audits and a change process for policy updates.