Loading…
Definitions for the most important terms in agentic AI — written for both technical and business audiences.
Hand-picked pieces from our architects, customers and partners.
Systems that take goals as input and deliver outcomes.
Read more →Accept a goal, deliver an outcome without step-by-step direction.
Read more →Specialized agents coordinated to complete workflows.
Read more →Retrieval-augmented generation.
Read more →RAG with access policy enforced at retrieval.
Read more →Role-based access control.
Read more →Immutable record of agent actions.
Read more →Plausible but unsupported model output.
Read more →Cloud within a jurisdiction with government-grade controls.
Read more →We publish what we'd want to read if we were building agents in your shoes — short on hype, long on practice.
The agentic AI vocabulary is moving fast. This glossary collects working definitions from practitioners so teams can communicate without re-litigating terms in every meeting.
A one-paragraph plain-English definition. No jargon, no marketing.
When and why the concept matters in enterprise practice.
A brief note on how the concept maps to xyner — without overselling.
Pointers to related vocabulary so you can navigate the conceptual map.
Terms are revised as the field evolves; definitions are dated where it matters.
The principles that distinguish how we approach this — calibrated for senior enterprise audiences.
Definitions reflect how engineering, operations and risk teams actually use the terms — not how vendors would prefer they were used.
Where a term has multiple working definitions in the industry, we say so and give you both — so your team can pick consciously.
Categories and patterns that recur across regulated industries.
Foundational concepts — RAG, vector database, control/data plane, model routing.
RBAC, ABAC, audit trail, approval gate, sovereignty, EU AI Act.
Self-healing, compensating action, observability, OpenTelemetry.
Hallucination, prompt injection, grounding, guardrails.
Agentic AI, autonomous goal execution, multi-agent orchestration, tool calling.
Session memory, long-term memory, organizational memory.
Most vendor glossaries define each term to fit their product. We aim to define each term as practitioners actually use it.
Plays well with
See xyner in your environment with a guided executive demo.