Home / AI Glossary
AI Glossary: every AI term decoded for SMEs.
AI has its own vocabulary, and it changes fast. Here are the essential terms, decoded for SME leaders, without unnecessary technical jargon.
AI Concepts
- AI agent An AI agent is an autonomous program that chains several tasks — rather than simply answering a question — by orchestrating a language model…
- LLM (Large Language Model) An LLM (Large Language Model) is an AI model trained on massive amounts of text to predict, understand and generate natural language.
- RAG (Retrieval-Augmented Generation) RAG (Retrieval-Augmented Generation) is a method that connects an LLM to your own data: documents, contracts, invoices, customer bases or internal…
- Fine-tuning Fine-tuning means adapting a pre-trained language model to your own data, to specialize it in a specific domain (your jargon, your tone, your…
- Foundation model A foundation model is a massive AI model, pre-trained on huge amounts of data, that serves as a base for countless applications.
- Token A token is the basic unit a language model processes: a piece of text (often part of a word, sometimes a whole word, sometimes punctuation) that…
- Embedding An embedding is a numeric representation of a text (sentence, paragraph, document) as a list of numbers — a vector — that encodes its meaning.
- Vector In AI, a vector is a structured list of numbers representing an object (document, email, image, product) in a space where proximity = meaning…
- Generative AI Generative AI refers to any artificial intelligence that can create new content — text, images, audio, code, video — from patterns learned during…
- Prompt engineering Prompt engineering is the practice of crafting clear, structured and contextual instructions to get reliable results from a language model.
- AI governance AI governance encompasses the rules, processes and responsibilities that frame Artificial Intelligence usage within an organization.
- AI hallucination An AI hallucination is a language model response that seems plausible but is factually false, fabricated or unverifiable.
- AI inference AI inference is the stage where a language model "reasons" over an input (your prompt, your document, your image) to produce an output (response,…
- Private AI cloud A private AI cloud is a dedicated hosting infrastructure — physical server, isolated VM or GPU cluster — where a company runs its AI models and…
- AI security AI security encompasses all practices aimed at protecting Artificial Intelligence systems against attacks, data leaks, manipulation and…
- Data ingestion Data ingestion is the process of collecting, cleaning and integrating data from heterogeneous sources (databases, Excel files, PDFs, emails, APIs)…
- AI automation AI automation is the use of Artificial Intelligence to execute repetitive tasks without human intervention — or with reduced oversight.
- AI technical workflow An AI technical workflow is the coded sequence of processing steps that transforms an input into a usable output, combining AI components (LLM,…
- AI pipeline An AI pipeline is a sequential processing chain where each step prepares data for the next: collection → cleaning → AI processing (classification,…
- Zero-shot learning Zero-shot learning is a language model's ability to perform a task without any specific examples in the prompt.
- Few-shot learning Few-shot learning is the technique of providing the language model with a few concrete examples (2 to 5) of the expected task directly in the…
- AI modality An AI modality is a type of data an AI model can process: text, image, audio, video, code, or a combination of several (multimodal).
- Context window The context window is the maximum amount of text a language model can process in a single request — input (prompt + injected documents) and output…
- Temperature (AI) Temperature is a generation parameter of a language model that controls the randomness of responses.
- System prompt The system prompt is the framing instruction sent to the model before the user messages: it defines the role, posture, constraints, tone, output…
AI Tools & Stacks
- N8N N8N is an open-source workflow automation platform that connects applications, APIs and services with little or no code.
- LangChain LangChain is an open-source framework in Python and JavaScript for building applications powered by language models, connecting them to data…
- Streamlit Streamlit is an open-source Python framework that turns data-science code into interactive web applications — no frontend or HTML required.
- Hugging Face Hugging Face is the go-to platform for open-source AI models: it hosts hundreds of thousands of models (text, image, audio, code), datasets and…
- OpenAI API The OpenAI API is the technical interface that lets a company embed OpenAI's language models (GPT-4, GPT-4o, etc.) directly into its own…
- vLLM vLLM is an open-source inference engine optimized for serving language models in production, with significantly better memory efficiency and…
- Make (formerly Integromat) Make (formerly Integromat) is a no-code automation platform that connects applications through visual scenarios.
- Supabase Supabase is an open-source database platform built on PostgreSQL, adding turnkey features: authentication, file storage, edge functions…
- Claude API The Claude API is Anthropic's technical interface for integrating Claude language models (Claude Sonnet, Claude Opus, Claude Haiku) into…
- Vector database A vector database is a database specialized in storing and searching vectors — numerical representations of the meaning of texts, images or data.
- AI copilot An AI copilot is an AI assistant integrated directly into an existing work tool (text editor, CRM, spreadsheet, development environment) that…
- Gemini API The Gemini API is Google's technical interface for accessing Gemini language models to integrate AI into applications and workflows.
- AI no-code platform An AI no-code platform is a tool that lets you create Artificial Intelligence applications — chatbots, workflows, data extractors, assistants —…
- MCP (Model Context Protocol) The MCP (Model Context Protocol) is an open standard that lets a language model interact with external data sources, tools and services through a…
- LangSmith LangSmith is an observability and evaluation platform for LLM-based applications.
- Firecrawl Firecrawl is a web content collection tool optimized for AI: it navigates websites, extracts structured content (text, tables, links) and returns…
- Resend Resend is a transactional and marketing email API designed for technical teams: it lets you send emails programmatically (via REST API or SDK)…
- ElevenLabs ElevenLabs is an audio AI platform specialized in voice synthesis (text-to-speech), voice cloning and audio translation.
- Vercel AI SDK The Vercel AI SDK is an open-source development kit that simplifies building AI applications in JavaScript: it provides functions for calling LLMs…
- Flowise Flowise is an open-source no-code platform that lets you build Retrieval-Augmented Generation (RAG) applications and AI agents via a visual…
- Whisper (AI transcription) Whisper is an open-source speech transcription model developed by OpenAI that converts audio (meeting, call, dictation, video) into text in many…
- Cursor (AI IDE) Cursor is a development environment (IDE) augmented by AI, fork of VS Code, that assists developers in real time: it completes code, explains an…
AI Roles
- Data scientist A data scientist is a specialist who combines statistics, programming and business expertise to extract actionable insights from data.
- Prompt engineer A prompt engineer is the specialist who designs, tests and optimizes instructions sent to language models to get reliable, reproducible and…
- Chief AI Officer (CAIO) The Chief AI Officer (CAIO) is the strategic leader of Artificial Intelligence within an organization.
- AI consultant An AI consultant is an independent expert or firm that helps businesses adopt Artificial Intelligence: diagnosis, strategy, technical selection,…
- AI auditor An AI auditor is a professional who independently assesses a company's Artificial Intelligence maturity, risks and opportunities.
- Fractional AI leadership Fractional AI leadership is a model where an SME hires a part-time AI expert to steer its Artificial Intelligence strategy and projects — without…
- AI project manager An AI project manager is the operational lead who coordinates the development and deployment of an Artificial Intelligence project, from scoping…
- Machine Learning engineer A Machine Learning (ML) engineer is the specialist who designs, trains, deploys and maintains AI models in production.
- Data engineer A data engineer is the specialist who designs, builds and maintains the data infrastructure powering AI and analytics applications.
- MLOps engineer An MLOps engineer is the specialist who combines IT operations practices (deployment, monitoring, automation) with the AI model lifecycle to…
- AI architect An AI architect is the senior profile who designs the overall technical architecture of an AI project: component selection (LLM, vector database,…
- AI data analyst An AI data analyst is a professional who combines classical data analysis skills (statistics, visualization, business intelligence) with AI tools…
- AI compliance officer The AI compliance officer is the responsible party who ensures the company's Artificial Intelligence usage complies with current regulations —…
- AI product owner The AI product owner is the responsible party who defines what the AI product must do — from a business perspective — and steers the feature backlog.
- AI trainer An AI trainer is the professional who teaches company teams to use Artificial Intelligence tools daily: writing with an LLM, automating with N8N…
- SME AI training SME AI training is a structured learning program for Artificial Intelligence adapted to the constraints of small and medium businesses: short…
- AI ethics officer The AI ethics officer is the responsible party who ensures the company's Artificial Intelligence systems respect ethical principles: fairness (no…
- AI product manager An AI product manager is the responsible party who steers the complete lifecycle of an AI product or use case — from idea to deployment through…
- AI security engineer An AI security engineer is the specialist who designs and implements protection measures specific to Artificial Intelligence systems: prompt…
- AI researcher An AI researcher is a professional working at the forefront of Artificial Intelligence: they design new model architectures, develop innovative…
- AI CTO The AI CTO is the responsible party who leads the technical side of Artificial Intelligence adoption: they validate architectures, supervise…
- Chief Data Officer (CDO) The Chief Data Officer (CDO) is the responsible party of an organization's data strategy: data quality, governance, accessibility and value creation.
AI Methods & Processes
- AI agent specification document An AI agent specification document is the structured document that precisely defines what an AI agent must do, in what context, with what…
- AI audit (method) An AI audit is a structured methodology for evaluating a company's Artificial Intelligence maturity.
- Few-shot prompting Few-shot prompting is a prompt engineering technique that involves providing the language model with a few concrete examples of the expected task…
- AI production deployment AI production deployment is the step that turns a functional Artificial Intelligence prototype into a tool used daily by the team.
- LLM evaluation LLM evaluation is the method of objectively measuring the quality, reliability and relevance of a language model's responses on test cases…
- RAG routine (Qlari method) The RAG routine (Qlari method) is a structured, recurring approach that keeps a RAG system healthy — that is, an AI assistant connected to your…
- Prompt chaining Prompt chaining is a technique that chains multiple prompts sequentially, where the output of prompt n becomes the input of prompt n+1.
- AI A/B testing AI A/B testing is the method of comparing two versions of an AI system (two prompts, two models, two workflows) on a real data sample to measure…
- MLOps MLOps (Machine Learning Operations) are the practices, tools and processes that automate the lifecycle of AI models in production: training,…
- Advanced RAG Advanced RAG refers to techniques that improve a basic RAG system's quality beyond simple "search then inject." It includes: hybrid search…
- Responsible AI Responsible AI is a methodological framework aimed at developing and deploying Artificial Intelligence ethically, fairly, transparently and safely.
- GDPR & AI The intersection of GDPR and AI encompasses all legal obligations and best practices that apply when a company uses Artificial Intelligence on…
- AI retraining AI retraining is the process of updating an AI model with more recent data to maintain its performance.
- Data quality Data quality is the degree of reliability, completeness, accuracy and freshness of a company's data.
- Prompt injection A prompt injection is an attack that inserts malicious instructions into an AI's conversation to hijack its behavior: disclosing internal…
- AI agent evaluation AI agent evaluation is the method of measuring whether an AI agent — which chains several tasks through several tools — correctly achieves its…
- AI guardrails AI guardrails are the rules, controls and mechanisms that frame an AI system's behavior to prevent dangerous, false or non-compliant outputs.
- Chain-of-thought Chain-of-thought is a prompting technique that asks the model to break its reasoning into explicit steps before giving its final answer.
- Data augmentation Data augmentation is a technique that enriches a dataset by creating variants of existing examples: reformulations of the same text, translations,…
- AI benchmark An AI benchmark is a comparative evaluation of AI systems on a standardized test set, to measure their relative performance: accuracy, source…
- Prompt optimization Prompt optimization is the iterative process that improves the quality, reliability and cost of an AI system's prompts: reformulation, adding…