GLOSSARY
Vector Database (vector search)
A vector database stores data (resumes, roles) as numerical vectors capturing their meaning. It enables semantic search: finding profiles by similarity of meaning rather than exact keywords.
IN DEPTH
A vector database stores each document (resume, candidate record, role description) as an embedding: a vector of numbers produced by an AI model that captures the text's meaning. Two texts close in meaning produce close vectors, even if they share no words: "fullstack JS dev" and "React/Node engineer" end up neighbors in vector space. This is the technology behind semantic search and AI matching: instead of filtering on exact keywords (the Boolean approach, which plateaus around 65% precision), the engine ranks profiles by similarity of meaning with the role, with precision exceeding 90% on complete data. For an agency or staffing firm, the stakes are concrete: a vector-indexed database can be queried in natural language and reveals profiles classic filters miss. AI-native platforms like Cobalt build vector indexing into the core of their architecture, where legacy ATS graft it as an add-on with partial data access.
Frequently asked questions
Because it understands synonyms and business context with no configuration: a search for "JavaScript backend architect" also surfaces "Node lead developers" that Boolean search would have missed. Result: more relevant profiles found in your own talent pool, before even sourcing externally.
If you want reliable semantic search and AI matching, yes: it is the technical foundation. The question for your vendor: is vector indexing native and applied to all data, or limited to an add-on matching module?
Related terms
- Semantic SourcingSemantic sourcing uses AI to understand natural language queries (e.g., "DevOps AWS 5 years Paris available March") and retrieve relevant candidates without Boolean search.
- AI Matching (candidate-mission)AI matching uses machine learning models to evaluate a candidate's relevance to a mission by crossing skills, experience, availability and cultural fit. It replaces Boolean search.
- CV ParsingCV parsing automatically extracts information from a CV (contact, experiences, skills, education) to structure it in an ATS. Modern AI reaches >95% accuracy.
- Boolean SearchBoolean search uses AND, OR, NOT operators and quotation marks to filter candidates in an ATS or on LinkedIn. In 2026, it's largely replaced by AI semantic search.
Ready to upgrade to an AI-first ATS?
Book a 30-min Cobalt demo to see how these concepts apply to your staffing firm or agency.
Book a demo