Vacancy matching that recruiters approve instead of search
A semantic matching engine pairs every new vacancy with the right companies and candidates, drafts the outreach, and puts a recruiter on the final decision.
- Client
- HR and recruitment services provider
- Sector
- Recruitment
- Type
- AI agent and automation
- Services
- AI agent, Semantic search, n8n orchestration, Approval workflow
Outcomes
Challenge
Recruiters searched a growing database of companies, roles and past feedback by hand. A good match depended on who was on shift and how much they remembered. New vacancies waited hours for a first shortlist.
The data existed. It was the moving of work between the database, the recruiter, the hiring manager and the inbox that cost the time.
Approach
Vacancies, companies, services and historical feedback are embedded with OpenAI text-embedding-3-large and searched semantically in PostgreSQL. An agent enriches each vacancy with insights, ranks candidates by fit and distance, and writes the reasoning down.
Nothing leaves the system on its own. Matches land in an approval queue; the recruiter sees the shortlist and pre-drafted outreach in Gmail, with the right manager already looked up, and approves or edits. A dashboard and Telegram notifications keep the team informed, and the whole flow runs in n8n.
Outcome
Matching turned from a search task into a review task. A new vacancy has a ranked shortlist and drafted outreach within minutes, and every message is still sent by a person who agreed with it.