AI glossary
— the terms, explained simply.
The most important AI terms for businesses, explained briefly and without buzzwords. From AI agent and RAG to human-in-the-loop and GDPR — so you know what we mean when we get started.
- AI agent
- A software system that pulls data from several sources, makes decisions and carries out tasks across multiple steps under clear rules. Unlike a chatbot, an agent can act across systems such as CRM, email and ERP — with a human-in-the-loop where the risk is high.
- Process automation
- Letting software carry out repetitive, rule-based tasks automatically, such as data entry, notifications or reporting. Reduces manual work and errors. The difference from an AI agent is that automation follows fixed rules, while an agent can weigh context.
- Large language model (LLM)
- An AI model trained on vast amounts of text that understands and generates natural language. Examples are GPT, Claude and Gemini. Language models power chatbots, text analysis and AI agents.
- RAG (Retrieval-Augmented Generation)
- A technique where a language model retrieves relevant information from your own sources, such as documents and databases, before it answers. Gives more precise, up-to-date answers grounded in the business's actual data — and fewer hallucinations.
- Human-in-the-loop
- A principle where a person approves or overrides the AI's decisions before they take effect, especially where the risk is high. Ensures control and accountability in automated workflows.
- Audit log
- A traceable, timestamped record of what a system has done — which decisions were made, when and why. Makes AI actions verifiable and is central to trust and compliance.
- Confidence threshold
- The bar for how certain an AI model must be before it acts on its own. Below the threshold, the case is handed to a person. A tool for balancing automation against risk.
- ICP (Ideal Customer Profile)
- A precise description of the company that benefits most from your product — based on industry, size, finances and needs. Used to prioritise sales effort. ICP Hub builds such profiles for Norwegian businesses.
- Prompt
- The instruction or question you give a language model. The quality of the prompt shapes the quality of the answer. “Prompt engineering” is the craft of writing effective instructions.
- Hallucination
- When a language model generates information that sounds credible but is wrong or made up. Reduced with techniques such as RAG, source citing and human-in-the-loop.
- Fine-tuning
- Training an existing language model further on a specific dataset so it performs better on a particular task or domain. An alternative or supplement to RAG.
- MCP (Model Context Protocol)
- An open standard that lets AI agents connect to tools and data sources in a structured way. Noragentic publishes an MCP server card so agents can look up services and pricing automatically.
- Workflow
- A defined sequence of steps that completes a task from start to finish. Automated workflows connect systems so data flows without manual handling.
- Data processing agreement
- An agreement governing how a vendor processes personal data on a company's behalf, in line with GDPR. Noragentic only uses vendors with clear data processing agreements and processes data in the EU.