Collibra unifies governance for structured and unstructured data in a shared knowledge graph. Documents, PDFs, transcripts, presentations, contracts, emails, and cloud files can be governed with the same rigor as structured data products.
Unstructured data often makes up 80% to 90% of an organization's total data estate, but without metadata it tends to be underutilized, inaccessible, or unsuitable for AI applications, and invisible to governance and analytics programs. Deasy Labs closes that gap by automatically tagging and enriching unstructured content, and Collibra governs the result.
Collibra and Deasy Labs
Deasy's tagging engine and Collibra's governance layer are integrated to bring control to unstructured data. Deasy Labs scans unstructured files, such as documents, contracts, and transcripts, and tags them automatically, generating structured metadata (document type, date, PII presence, and other attributes) without manual effort.
A Document Slice is the output of a query against those tags, for example document type = contract AND date > 2025 AND PII = none. It's a curated, named subset of documents, not the documents themselves.
Document Slices are published into Collibra, where they're governed like any other asset: access is controlled and routed to owners, slices can be published to the Data Marketplace to improve discoverability, and they're monitored for ongoing compliance. Collibra never ingests the underlying file content, only the slice definition, its tags, and who can access it. Files stay out of the model by design, for scale.
Three asset types make this governable:
- Document Slice: a technology asset representing a defined subset of documents from a source system, published and governed as a named collection. It's a query result over documents, not the documents themselves.
- Tag Definition: a business asset describing a tag used to classify unstructured content, defined using a Smart AI, Pattern Match, or Rules-based strategy.
- Tag Value: a data asset holding a value produced by a Tag Definition, used to label and categorize documents within a Document Slice.
Tag Definition and Tag Value are stored in a Tag Library domain. For the full asset type, attribute, and relation list, go to Governance for unstructured data: operating model.
Beneficiaries
Deasy Labs is designed for organizations, particularly those in regulated industries, such as financial services and healthcare, that are building Generative AI (GenAI), Retrieval-Augmented Generation (RAG), and agentic workflows. Key stakeholders include:
- AI Engineers: Improve retrieval accuracy for GenAI and RAG systems by automatically tagging and enriching unstructured data with semantic metadata.
- Data Scientists: Ensure documents used to train or augment models meet governance and lineage requirements, while filtering out noisy or duplicate documents before ingestion.
- Knowledge Managers: Make unstructured content discoverable and reusable, and curate documents by themes or compliance needs.
- Compliance Analysts: Monitor unstructured content for policy violations, such as PII exposure, and use traceable metadata for AI audit and transparency requirements.
- Data Platform Engineers: Automate the ingestion and enrichment of unstructured files at scale, reducing manual tagging and accelerating integration with downstream systems.
Key use cases
Capabilities
Deasy Labs automates the conversion of unstructured content into structured knowledge assets, eliminating the need for custom AI pipelines or extensive manual labeling. The automated workflow connects to unstructured data, such as raw files or vector databases, performs AI-assisted schema generation, applies LLM-based tagging, curates Document Slices, and syncs the structured metadata. It provides confidence scores and traceability for every generated tag, supporting explainability and enabling human-in-the-loop validation.
How this fits with what you already govern
A Document Slice is a member of Collibra's existing Data Product Port Asset asset type group, so it can be used directly as a Data Product port, the same mechanism structured data already uses, with no new relation type required. In practice, a Document Slice can sit behind a Data Product in the Data Marketplace with the same request-and-access flow as any structured data product, serve as an input port for a Data Product, or be consumed directly by an AI use case. A Tag Definition can also represent a Data Attribute, linking tag-based classification directly to Collibra's semantic model.
Integration with Collibra Platform
Deasy Labs capabilities are deeply integrated with Collibra to provide unified governance across all data types:
- Unified governance: Enable governance across structured and unstructured files, transforming them into reusable, governed data assets.
- Lifecycle management: Govern the full lifecycle of AI and data use cases, including discovery, enrichment, policy enforcement, risk management, and lineage tracking.
- Technical stacks: Access functionality through a user-friendly platform for integration with existing AI stacks, such as vector databases and embedding models.
- Deployment: Built to operate across multi-cloud environments.
Deasy Labs documentation
Collibra governs what Deasy Labs produces. The documentation for the tagging engine itself is currently maintained on the Deasy Labs website. For configuration and technical details beyond what you find here in the Collibra Documentation Center, go to the Deasy Labs documentation.