AI Glossary 2026
All key AI terms clearly explained — from chatbot to RAG, from LLM to prompt engineering. Everything you need for AI in your business.
Artificial Intelligence (AI)
Computer systems that perform tasks normally requiring human intelligence, such as learning, reasoning and understanding language.
Machine Learning (ML)
A branch of AI where systems automatically improve through experience, without being explicitly programmed.
Deep Learning
A form of machine learning based on neural networks with multiple layers, inspired by the human brain.
Neural Network
A mathematical model inspired by the brain, made of layers of connected nodes (neurons).
Natural Language Processing (NLP)
The ability of AI to understand, interpret and generate human language in both text and speech.
Generative AI
AI that creates new content — text, images, code or audio — rather than only analyzing existing data.
LLM (Large Language Model)
AI models trained on massive amounts of text, such as GPT-4, Claude and Gemini. Understand and generate text.
GPT
A family of language models from OpenAI, including ChatGPT. GPT stands for Generative Pre-trained Transformer.
Transformer
The neural network architecture behind modern LLMs. Introduced by Google in 2017.
Parameters
The internal variables an AI model learns during training. More parameters often (not always) means a more capable model.
Token
The basic unit of text an LLM processes — usually part of a word. 1 token ≈ 4 characters.
Context Window
The amount of text a model can process at once. Large windows (100K+ tokens) can handle entire documents.
Fine-tuning
Specializing an AI model on your specific data or task, on top of its general training.
Embedding
A numerical representation of text that lets AI compare meaning. The foundation of semantic search.
RAG (Retrieval Augmented Generation)
AI technique where the model retrieves relevant info from your documents before answering. Ensures accurate, source-specific responses.
Vector Database
A database specifically for storing and searching embeddings. Essential for RAG systems.
AI Agent
An autonomous AI that performs tasks independently — thinking, deciding, using tools — rather than only answering questions.
Chatbot
An AI program that communicates with users in natural language via text. Usually for customer service and lead generation.
Voice Agent
An AI that can hold phone conversations with a natural voice, schedule appointments and qualify leads.
RPA (Robotic Process Automation)
Software robots that automate repetitive computer tasks like data entry, email processing or invoicing.
Workflow Automation
Automating business processes by connecting multiple tools and systems together.
Copilot
An AI assistant that helps people work faster without operating fully autonomously. E.g. GitHub Copilot or Microsoft 365 Copilot.
Function Calling
An LLM’s ability to call external functions or APIs, like checking a calendar or updating a database.
Multi-agent Systeem
Multiple AI agents working together, each with its own specialty, to solve complex tasks.
Prompt
The instruction or question you give to an AI model. Better prompts produce better results.
Prompt Engineering
The craft of writing effective prompts to get optimal AI results.
System Prompt
Background instructions that determine an AI’s behavior, such as personality, role and rules.
Hallucinatie
When an AI generates convincing-sounding but factually incorrect information.
Temperature
A setting that controls how creative or predictable AI output is. Low = consistent, high = creative.
Chain-of-Thought
A technique where AI reasons step-by-step before answering. Improves accuracy.
Few-shot Learning
Giving an AI examples in the prompt so it learns what you want, without fine-tuning.
Zero-shot
Having an AI perform a task without examples — based on instructions alone.
Grounding
Connecting an AI to real data (documents, databases) to give factually correct answers.
ROI (Return on Investment)
The return on an AI investment. E.g. a voice agent costs €5K but saves €20K/year = 4× ROI.
MVP (Minimum Viable Product)
A first working version of an AI solution to test before full investment. Similar to our pilot.
Pilot
A small-scale test of an AI solution in your business, usually 1–2 weeks, to prove value before scaling.
API Integratie
Connecting AI to your existing systems (CRM, calendar, email) via programming interfaces (APIs).
Human-in-the-Loop
System design where AI prepares decisions but humans give final approval. Essential for high-risk tasks.
Time-to-Value
The time between implementation and measurable business results. Utomatic delivers this within 2–4 weeks.
SLA (Service Level Agreement)
Contractual agreements about performance, e.g. uptime guarantee or response time.
AVG / GDPR
European privacy law governing how companies process personal data. All Utomatic solutions are GDPR-compliant.
EU AI Act
European AI legislation (2024) that classifies AI systems by risk and imposes corresponding obligations.
Bias
Systematic skewed judgments by AI, often due to biased training data. Important to mitigate.
Data Sovereignty
The principle that data is subject to the laws of the country where it’s stored. Utomatic uses EU servers.
Verwerkersovereenkomst
Contract between a company and AI provider about how personal data is processed. Required under GDPR.
Explainability
The ability to explain why an AI made a particular decision. Required for high-risk applications.
Speech-to-Text (STT)
Technology that converts spoken words into written text. Foundation of voice agents.
Text-to-Speech (TTS)
Technology that converts text into natural-sounding speech. Modern TTS is barely distinguishable from humans.
Voice Cloning
Creating an AI copy of someone’s voice from a short recording.
Intent Recognition
AI that understands what a user wants to achieve, regardless of exact wording.
Semantic Search
Search method that understands meaning, not just exact words. “Car” also finds “vehicle”.
More questions about AI?
Book a free consultation and discover how AI can help your business.
