AI Glossary: Key Terms & Definitions

AI systems that can autonomously think, plan, decide, and execute multi-step tasks. Unlike traditional AI that only generates responses, agentic AI agents can use tools, access databases, coordinate with other agents, and complete complex workflows without human intervention.

An AI system that operates independently to achieve goals. Autonomous agents perceive their environment, make decisions, take actions, and learn from outcomes — all without continuous human oversight.

A system where multiple AI agents work together, each with specialized roles, to solve complex problems. Agents communicate, delegate tasks, share context, and coordinate on workflows. Multi-agent systems enable parallel processing and role-based specialization.

The coordination and management of multiple AI agents working together. Orchestration handles task delegation, agent communication, conflict resolution, and workflow execution across the agent team.

A deep learning model trained on massive text datasets to understand and generate human-like language. LLMs like GPT, Claude, and Gemini power conversational AI, content generation, and reasoning capabilities.

A technique that combines LLMs with external knowledge retrieval. RAG systems fetch relevant information from databases or documents before generating responses, improving accuracy and reducing hallucinations.

The process of further training a pre-trained LLM on domain-specific data to improve performance on specialized tasks. Fine-tuning customizes model behavior without training from scratch.

The practice of designing effective instructions (prompts) for LLMs to produce desired outputs. Includes system prompts, few-shot examples, chain-of-thought reasoning, and structured output formatting.

The ability of AI agents to call external APIs, query databases, and execute code. Function calling allows agents to interact with real-world systems and perform actions beyond text generation.

The maximum amount of text (measured in tokens) an LLM can process in a single request. Larger context windows enable longer conversations and more comprehensive document analysis.

A database optimized for storing and querying vector embeddings. Vector databases enable semantic similarity search, which is essential for RAG systems and AI-powered recommendations.

Numerical representations of text, images, or other data as high-dimensional vectors. Embeddings capture semantic meaning, enabling similarity comparisons and semantic search.

When an LLM generates false or fabricated information that appears plausible. Hallucinations are a key challenge in AI reliability, mitigated through RAG, grounding, and output validation.

A design pattern where human reviewers approve, reject, or modify AI agent decisions before execution. HITL ensures oversight for high-stakes actions and builds trust in autonomous systems.

The framework of policies, processes, and controls that ensure AI systems are deployed responsibly. Governance covers accountability, transparency, fairness, compliance, and risk management.

A standardized protocol for connecting AI models to external data sources and tools. MCP enables agents to access documents, databases, and APIs through a unified interface.

The ability to monitor, trace, and debug AI agent behavior in production. Observability includes logging, metrics, tracing, and replay capabilities for understanding agent decisions and failures.

Platforms that enable non-technical users to build and deploy AI agents through visual interfaces — no programming required. No-code AI democratizes AI development for business users.

The ability of AI agents to reason through complex problems using deduction, induction, and analogical thinking. Cognitive reasoning enables agents to break down problems and plan multi-step solutions.

How AI agents store and retrieve information across interactions. Includes short-term (conversation context), long-term (persistent knowledge), and episodic (past experiences) memory systems.