Enterprise AI glossary
Guides for executives, IT and data owners. Understand visible risks, then define your Private AI scope together.
Private AI
Private AI is an approach where an organization defines data boundaries, users, models and deployment. Privacy results from architecture and operations, not the product name alone.
Learn moreLocal LLM
A local LLM runs inference on a chosen computer or server. Locating the model near data helps define processing paths, but OCR, embedding and log services still need separate review.
Learn moreOn-premise AI
On-premise AI runs on infrastructure operated by the organization. Scope includes more than GPUs: networking, storage, identity, backup and incident response matter too.
Learn moreEnterprise RAG
RAG retrieves relevant information and supplies it as context for model-generated answers. Enterprise RAG adds access, version and audit requirements so retrieval uses knowledge the asker is allowed to access.
Learn moreAI Agent
An AI agent uses a model to select steps and tools toward a goal. Action permissions matter beyond document access: a model proposing an action does not authorize the user to perform it.
Learn moreRole-based access control (RBAC)
RBAC assigns permissions by role, linked to workspaces, sources and agents. Hiding buttons is not access enforcement.
Learn moreWorkspace
A workspace groups users, documents and agents for one department, client or project, making scope visible and reducing accidental context mixing.
Learn moreShadow AI
Shadow AI is AI use outside organizational approval or visibility. Risk depends on data, tools, accounts and contracts. It does not mean every public AI service is unsafe.
Learn moreAudit log
AI records should identify who did what, with which data and model, when, and who may review it. Balance useful evidence against sensitive information in logs.
Learn moreEnterprise AI
Private AI is an approach where an organization defines data boundaries, users, models and deployment. Privacy results from architecture and operations, not the product name alone.
Learn morePrivate LLM
A local LLM runs inference on a chosen computer or server. Locating the model near data helps define processing paths, but OCR, embedding and log services still need separate review.
Learn moreSovereign AI
Data control covers originals, chunks, embeddings, indexes, caches, answers and logs. Know where each lives, who may access it and how to delete or restore it.
Learn moreRAG
RAG retrieves relevant information and supplies it as context for model-generated answers. Enterprise RAG adds access, version and audit requirements so retrieval uses knowledge the asker is allowed to access.
Learn moreGrounding
Citations connect answers to documents, sections, pages and versions so users can check support for a conclusion. A filename alone does not guarantee correctness.
Learn moreSemantic search
Enterprise search retrieves across organizational sources under user permissions. Keywords help with identifiers and exact names; semantic retrieval helps with different wording. Test both on actual work.
Learn moreLLM
A large language model generates text from learned patterns and context; it is not a verified fact database.
Embedding
A numeric vector representation used to compare relationships. Evaluate the model on actual languages and tasks.
Vector database
A store for vector retrieval. Manage permissions, backups and deletion when source documents change.
Vector search
Finding vectors near a query. Similarity does not establish correctness or permission to read.
No results. Try another term.
References
Public sources explain principles; they do not certify or endorse RAGBOX.
- NIST AI Risk Management Framework
- Microsoft: Retrieval-augmented generation
- OWASP: RAG Security Cheat Sheet
- OWASP: Prompt Injection Prevention
Expert review is required before operational use. This is not case-specific legal or tax advice.