[
  {
    "term": "Agentic workflow",
    "slug": "agentic-workflow",
    "category": "workflow",
    "definition": "An agentic workflow is a multi-step process where an AI system plans, calls tools, checks results, and iterates toward an outcome.",
    "whyItMatters": "This is the practical difference between one-off content generation and automation that can handle more complex business processes.",
    "inPractice": "A coding workflow can inspect a repo, edit files, run checks, and revise based on failures.",
    "relatedTerms": [
      "ai-agent",
      "mcp"
    ],
    "links": [
      {
        "href": "/best/ai-coding-assistants/",
        "label": "Compare coding assistants"
      },
      {
        "href": "/prompts",
        "label": "Use prompt templates"
      }
    ]
  },
  {
    "term": "AI agent",
    "slug": "ai-agent",
    "category": "workflow",
    "definition": "An AI agent is a system that can decide what step to take next, use tools, and keep moving toward a goal instead of answering only one prompt at a time.",
    "whyItMatters": "Teams evaluating agent products need to separate simple chat wrappers from systems that can actually plan, retrieve data, and execute tasks.",
    "inPractice": "A support agent might look up order data, draft a response, and escalate edge cases without a human manually switching tools.",
    "relatedTerms": [
      "agentic-workflow",
      "mcp"
    ],
    "links": [
      {
        "href": "/tools",
        "label": "Browse AI tools"
      },
      {
        "href": "/agents",
        "label": "Browse AI agents"
      }
    ]
  },
  {
    "term": "Benchmark",
    "slug": "benchmark",
    "category": "evaluation",
    "definition": "A benchmark is a standardized test used to compare model performance on tasks such as reasoning, coding, math, or factual recall.",
    "whyItMatters": "Benchmarks can help narrow a shortlist, but they rarely map perfectly to your real workflow or internal data.",
    "inPractice": "A model that leads on public reasoning tests may still underperform on your support tickets or pricing extraction workflow.",
    "relatedTerms": [
      "llm",
      "inference"
    ],
    "links": [
      {
        "href": "/tools/ai-model-comparison",
        "label": "Compare model tradeoffs"
      },
      {
        "href": "/data/llm-pricing",
        "label": "Balance quality and cost"
      }
    ]
  },
  {
    "term": "Context window",
    "slug": "context-window",
    "category": "models",
    "definition": "A context window is the amount of text, code, or multimodal input a model can consider in one request.",
    "whyItMatters": "Context limits determine whether a model can handle large contracts, research packets, or codebases without splitting the work up.",
    "inPractice": "If your prompt and source material exceed the model context, quality drops or the request fails outright.",
    "relatedTerms": [
      "token",
      "llm"
    ],
    "links": [
      {
        "href": "/tools/context-window-checker",
        "label": "Check context fit"
      },
      {
        "href": "/data/llm-pricing",
        "label": "Compare model limits"
      }
    ]
  },
  {
    "term": "Embedding",
    "slug": "embedding",
    "category": "knowledge",
    "definition": "An embedding is a numerical vector that represents the meaning of a piece of text, image, or other data in a way that machines can compare and search.",
    "whyItMatters": "Embeddings power semantic search, recommendation systems, and RAG pipelines by enabling similarity matching beyond keyword overlap.",
    "inPractice": "A support system embeds incoming tickets and matches them against a knowledge base to surface the most relevant articles.",
    "relatedTerms": [
      "vector-database",
      "rag"
    ],
    "links": [
      {
        "href": "/datasets",
        "label": "Explore datasets"
      },
      {
        "href": "/tools",
        "label": "Browse AI tools"
      }
    ]
  },
  {
    "term": "Fine-tuning",
    "slug": "fine-tuning",
    "category": "customization",
    "definition": "Fine-tuning means training a base model further on a specialized dataset so it behaves better for a narrow job or domain.",
    "whyItMatters": "It can improve consistency for repeated tasks, but it is usually more expensive and slower to maintain than a prompt-plus-RAG approach.",
    "inPractice": "A company might fine-tune a model for niche classifications after prompt engineering stops improving accuracy.",
    "relatedTerms": [
      "rag",
      "prompt-engineering"
    ],
    "links": [
      {
        "href": "/datasets",
        "label": "Find training data sources"
      },
      {
        "href": "/prompts",
        "label": "Start with prompts first"
      }
    ]
  },
  {
    "term": "Grounding",
    "slug": "grounding",
    "category": "risk",
    "definition": "Grounding means tying a model response to verified sources, live system data, or retrieved documents instead of relying only on model memory.",
    "whyItMatters": "Grounded outputs are usually easier to audit, safer for production, and more defensible in buyer-facing or regulated workflows.",
    "inPractice": "A finance workflow can ground a response in policy docs and ERP data before drafting an answer.",
    "relatedTerms": [
      "hallucination",
      "rag"
    ],
    "links": [
      {
        "href": "/tools/business-finance",
        "label": "Business and finance tools"
      },
      {
        "href": "/data/marketing-tools-matrix",
        "label": "Structured comparison example"
      }
    ]
  },
  {
    "term": "Guardrails",
    "slug": "guardrails",
    "category": "risk",
    "definition": "Guardrails are rules, filters, or validation layers that constrain AI system behavior to prevent harmful, off-topic, or incorrect outputs.",
    "whyItMatters": "Production AI systems need guardrails to maintain safety, compliance, and quality standards at scale.",
    "inPractice": "A financial chatbot uses guardrails to prevent investment advice, block PII from being stored, and ensure regulatory disclosures are included.",
    "relatedTerms": [
      "hallucination",
      "grounding"
    ],
    "links": [
      {
        "href": "/tools",
        "label": "Browse AI tools"
      },
      {
        "href": "/ai-glossary",
        "label": "Full glossary"
      }
    ]
  },
  {
    "term": "Hallucination",
    "slug": "hallucination",
    "category": "risk",
    "definition": "A hallucination is when a model states something false, unsupported, or invented as if it were correct.",
    "whyItMatters": "Hallucination risk shapes where AI can be safely automated and where human review or grounded retrieval is still required.",
    "inPractice": "Models may invent sources, quote nonexistent policies, or confidently misstate numbers when no trusted reference is supplied.",
    "relatedTerms": [
      "grounding",
      "rag"
    ],
    "links": [
      {
        "href": "/tools/customer-support",
        "label": "Support tools with grounding"
      },
      {
        "href": "/tools/context-window-checker",
        "label": "Check prompt fit"
      }
    ]
  },
  {
    "term": "Inference",
    "slug": "inference",
    "category": "evaluation",
    "definition": "Inference is the act of running a trained model to generate an output from a given input.",
    "whyItMatters": "Inference cost and latency become central once you move from experimentation to production traffic.",
    "inPractice": "A cheap model may still be the wrong choice if repeated inference calls and retries create hidden cost.",
    "relatedTerms": [
      "token",
      "benchmark"
    ],
    "links": [
      {
        "href": "/tools/token-calculator",
        "label": "Estimate inference cost"
      },
      {
        "href": "/data/llm-pricing",
        "label": "Review pricing inputs"
      }
    ]
  },
  {
    "term": "LLM",
    "slug": "llm",
    "category": "models",
    "definition": "LLM stands for large language model, a system trained on massive text or code corpora to predict and generate language.",
    "whyItMatters": "Most AI buying decisions start with model choice because model quality, cost, speed, and tool access shape the rest of the stack.",
    "inPractice": "Teams often compare frontier models for reasoning and cheaper models for background automation.",
    "relatedTerms": [
      "token",
      "context-window",
      "benchmark"
    ],
    "links": [
      {
        "href": "/tools/ai-model-comparison",
        "label": "Compare LLMs"
      },
      {
        "href": "/data/llm-pricing",
        "label": "View model pricing"
      }
    ]
  },
  {
    "term": "MCP",
    "slug": "mcp",
    "category": "integration",
    "definition": "MCP stands for Model Context Protocol, a standard way for models or agents to connect to tools, data sources, and external systems.",
    "whyItMatters": "Standardized tool connections reduce brittle one-off integrations and make agent systems easier to extend.",
    "inPractice": "An MCP server can expose CRM data, ticket history, or internal docs to an AI assistant through a consistent interface.",
    "relatedTerms": [
      "ai-agent",
      "agentic-workflow"
    ],
    "links": [
      {
        "href": "/mcp-servers",
        "label": "Browse MCP servers"
      },
      {
        "href": "/agents",
        "label": "See AI agents"
      }
    ]
  },
  {
    "term": "Multimodal model",
    "slug": "multimodal-model",
    "category": "models",
    "definition": "A multimodal model can work across more than one input or output type, such as text, images, audio, or video.",
    "whyItMatters": "Multimodal capability matters when you need one system to interpret screenshots, documents, voice, or visual creative assets.",
    "inPractice": "A multimodal workflow might read a PDF, inspect charts, and answer questions in plain language.",
    "relatedTerms": [
      "llm"
    ],
    "links": [
      {
        "href": "/tools/ai-model-comparison",
        "label": "See multimodal models"
      },
      {
        "href": "/datasets",
        "label": "Browse AI datasets"
      }
    ]
  },
  {
    "term": "Prompt engineering",
    "slug": "prompt-engineering",
    "category": "customization",
    "definition": "Prompt engineering is the practice of structuring instructions, examples, context, and constraints so a model produces more reliable output.",
    "whyItMatters": "For most teams, better prompts are the fastest and cheapest performance improvement available.",
    "inPractice": "Adding role, task, format, and examples often lifts quality before you need model changes.",
    "relatedTerms": [
      "fine-tuning",
      "token"
    ],
    "links": [
      {
        "href": "/prompts",
        "label": "Browse prompt library"
      },
      {
        "href": "/tools/prompt-generator",
        "label": "Generate a prompt"
      }
    ]
  },
  {
    "term": "RAG",
    "slug": "rag",
    "category": "knowledge",
    "definition": "RAG stands for retrieval-augmented generation. It means fetching relevant external information before the model writes an answer.",
    "whyItMatters": "RAG is one of the most common ways to make AI systems more useful on company-specific knowledge without retraining a model.",
    "inPractice": "A team can connect docs, help center articles, or internal policies so the model answers from fresh source material.",
    "relatedTerms": [
      "vector-database",
      "fine-tuning"
    ],
    "links": [
      {
        "href": "/datasets",
        "label": "Browse datasets"
      },
      {
        "href": "/tools/customer-support",
        "label": "See support AI tools"
      }
    ]
  },
  {
    "term": "System prompt",
    "slug": "system-prompt",
    "category": "customization",
    "definition": "A system prompt is an instruction set given to a model at the start of a conversation that defines its role, behavior, and constraints.",
    "whyItMatters": "System prompts are the primary way to customize model behavior without fine-tuning, making them critical for production applications.",
    "inPractice": "A customer support bot uses a system prompt to enforce brand voice, restrict topics, and specify escalation rules.",
    "relatedTerms": [
      "prompt-engineering",
      "fine-tuning"
    ],
    "links": [
      {
        "href": "/prompts",
        "label": "Browse prompt library"
      },
      {
        "href": "/tools/prompt-generator",
        "label": "Generate a prompt"
      }
    ]
  },
  {
    "term": "Temperature",
    "slug": "temperature",
    "category": "models",
    "definition": "Temperature is a parameter that controls how random or creative a model's output is. Lower values produce more deterministic responses, higher values increase variation.",
    "whyItMatters": "Choosing the right temperature affects whether output is consistent and reliable or diverse and exploratory.",
    "inPractice": "A code generation task typically uses low temperature for correctness, while brainstorming tasks use higher temperature for variety.",
    "relatedTerms": [
      "llm",
      "inference"
    ],
    "links": [
      {
        "href": "/tools/ai-model-comparison",
        "label": "Compare models"
      },
      {
        "href": "/prompts",
        "label": "Prompt library"
      }
    ]
  },
  {
    "term": "Token",
    "slug": "token",
    "category": "models",
    "definition": "A token is a small unit of text that models use for reading and generating content. Pricing and context limits are usually measured in tokens.",
    "whyItMatters": "Token counts directly affect API cost, latency, and whether a workload fits into a model at all.",
    "inPractice": "A long prompt with a large knowledge base chunk can become expensive even before the model writes any output.",
    "relatedTerms": [
      "context-window",
      "inference"
    ],
    "links": [
      {
        "href": "/tools/token-calculator",
        "label": "Estimate token cost"
      },
      {
        "href": "/data/llm-pricing",
        "label": "See pricing data"
      }
    ]
  },
  {
    "term": "Tool use",
    "slug": "tool-use",
    "category": "integration",
    "definition": "Tool use is a model capability where the LLM can call external functions, APIs, or services as part of generating a response.",
    "whyItMatters": "Tool use transforms a model from a text generator into a system that can take real actions like querying databases, sending messages, or reading files.",
    "inPractice": "A coding agent uses tool calls to read files, run tests, and write code rather than just suggesting edits in chat.",
    "relatedTerms": [
      "mcp",
      "ai-agent"
    ],
    "links": [
      {
        "href": "/mcp-servers",
        "label": "Browse MCP servers"
      },
      {
        "href": "/tools",
        "label": "See AI tools"
      }
    ]
  },
  {
    "term": "Vector database",
    "slug": "vector-database",
    "category": "knowledge",
    "definition": "A vector database stores embeddings so a system can retrieve semantically similar content rather than matching only exact keywords.",
    "whyItMatters": "Vector retrieval is foundational for many RAG systems, especially when knowledge is spread across large document collections.",
    "inPractice": "Instead of searching for exact phrasing, the system can find passages that are conceptually similar to the user question.",
    "relatedTerms": [
      "rag"
    ],
    "links": [
      {
        "href": "/datasets",
        "label": "Explore source datasets"
      },
      {
        "href": "/prompts/research-deep-dive/",
        "label": "Research prompt example"
      }
    ]
  }
]