Reference
AI terms that actually matter.
Explained for leaders.
This glossary covers the AI vocabulary your leadership team will encounter in 2026 — defined in plain English, with a note on how each concept applies in real organisational contexts. No jargon. No PhD required.
Download the AI fluency starter guide →Want the full AI fluency starter guide?
A downloadable PDF covering these terms plus a self-assessment for your team.
A
- AI Agent
- An AI agent is a system that can take actions autonomously — not just answer a question, but complete a sequence of steps, use tools, and make decisions in order to achieve a goal. Unlike a simple chatbot that responds to a single prompt, an agent can browse the web, write and run code, call APIs, or coordinate with other agents to complete a complex task. The term covers a wide spectrum: from a simple script that reads emails and drafts replies, to a multi-agent pipeline that manages an entire research workflow. What defines an agent is the combination of a goal, a set of available actions, and the ability to adapt based on what happens at each step. For leadership teams, the key question is not "can we build an agent?" but "which processes in our organisation are stable enough, and high-value enough, to benefit from agent automation?" → Explored in depth during the AI Strategy & Roadmap engagement.
- AI Fluency
- AI fluency is the ability to work productively alongside AI tools — understanding what they can and cannot do, knowing how to direct them effectively, and applying sound judgement about when and how to use AI outputs. It does not mean being able to build AI systems; it means being able to use them confidently and critically in your daily work. A fluent AI user can craft an effective prompt, evaluate the quality of an AI-generated output, identify hallucinations or bias, and integrate AI into their workflow without abandoning their professional judgement. Fluency is role-dependent: what a product manager needs differs from what a finance director or Scrum Master needs. Building AI fluency across an organisation is not a one-time event — it is an ongoing capability that compounds over time. → The AI Skills Bootcamp is designed specifically to build role-specific AI fluency across teams.
- AI Governance
- AI governance is the set of policies, processes, and oversight mechanisms that determine how AI is used within an organisation — and how that use is monitored, audited, and controlled. It covers questions like: which data can AI systems access? Who is accountable when an AI-assisted decision goes wrong? How do we document AI use for regulatory purposes? Good AI governance is not about blocking AI adoption — it is about creating the conditions under which AI can be used confidently and responsibly at scale. In regulated industries, AI governance is increasingly tied to external compliance requirements including the EU AI Act. In most organisations, governance frameworks are lagging significantly behind the pace of AI tool adoption, which creates real operational and reputational risk. → AI governance frameworks are a core output of the AI Strategy & Roadmap programme.
- AI Maturity
- AI maturity refers to how far along an organisation is in its journey of integrating AI into its operations, culture, and decision-making. It is typically measured across several dimensions: data readiness, AI tooling adoption, workforce fluency, governance frameworks, and leadership alignment. A low-maturity organisation may have a few enthusiastic individuals experimenting with AI tools in isolation, with no shared standards or governance. A high-maturity organisation has AI embedded into core workflows, clear ownership structures, and a culture of continuous AI capability development. Understanding your organisation's current maturity level is the essential starting point for any AI strategy — because it determines what the right next step actually is. → Anteligen's free AI Maturity Assessment maps your organisation across five dimensions in 30 minutes.
- AI Operating Model
- An AI operating model describes how an organisation structures itself to adopt, deploy, and scale AI — covering roles and responsibilities, decision rights, tooling standards, governance processes, and how AI use is coordinated across teams and functions. Without a deliberate operating model, AI adoption tends to be fragmented: individual teams pick different tools, develop incompatible workflows, and create governance blind spots. A well-designed AI operating model does not require a dedicated AI team or a large budget — but it does require clear decisions about who owns what, how AI tools are evaluated and approved, and how capability is built and maintained over time. The operating model is the organisational layer that turns isolated AI experiments into sustainable, scalable capability. → Defining your AI operating model is a key deliverable in the AI Strategy & Roadmap engagement.
- AI-First
- AI-first is a design and decision-making philosophy where AI is considered from the outset of any product, process, or workflow design — rather than added as an afterthought. An AI-first approach asks: "how would we design this if we assumed AI capability was available and reliable?" rather than "where can we plug AI into what we already have?" This does not mean automating everything or removing human judgement. It means that the starting assumption is that AI is a productive member of the team — with specific strengths, specific limitations, and a role that needs to be deliberately designed. For most organisations, becoming genuinely AI-first requires a shift in mindset before it requires a shift in tooling. → The AI for Leadership programme helps managers build an AI-first perspective for their teams.
- Agile AI
- Agile AI refers to the integration of AI tools and practices into Agile frameworks — Scrum, Kanban, SAFe, and similar methodologies. It is also the philosophy that AI adoption itself should follow Agile principles: iterative, empirical, and focused on delivering value in short cycles rather than through large, multi-year transformation programmes. In practice, this means using AI to accelerate Agile ceremonies (planning, retrospectives, refinement), automating administrative overhead (velocity tracking, meeting summaries, backlog hygiene), and building AI use cases through sprint-based experimentation rather than waterfall-style projects. Anteligen was founded on the belief that Agile and AI are natural partners — each amplifies the other when applied well. → The AI for Scrum Masters and AI for POs & PMs programmes are built on Agile AI principles.
C
- COMEX (in the context of AI)
- COMEX — short for Comité Exécutif, or Executive Committee — refers to the senior leadership team that holds strategic and operational authority in a large organisation. In the context of AI, the COMEX is critical because AI transformation cannot succeed without explicit executive alignment on strategy, investment, and risk appetite. Many AI initiatives stall not because of technical failure but because the executive layer has not reached consensus on what role AI should play in the organisation, what resources it warrants, and who is accountable for outcomes. A COMEX that has been through a shared AI orientation — ideally a structured executive briefing — is far more likely to enable coherent, well-governed AI adoption than one where individual members are operating from different assumptions and different levels of AI literacy. → The AI Exec Briefing is designed specifically for COMEX-level alignment.
D
- Design Thinking AI
- Design Thinking AI is the application of design thinking methodology — empathy, problem framing, ideation, prototyping, and testing — to the identification and development of AI use cases. Rather than starting with a technology ("we have access to an LLM, what can we do with it?"), Design Thinking AI starts with a human problem ("what is the real obstacle our users or our team faces?") and works toward AI as a potential solution. This approach dramatically reduces the risk of building AI solutions that no one adopts, because it grounds development in genuine user needs and validates ideas quickly through low-cost prototypes before committing significant resources. It is the methodology underpinning Anteligen's Create programmes. → Practised hands-on in the AI Innovation Sprint and AI Prototype Sprint.
F
- Fine-Tuning
- Fine-tuning is the process of taking a pre-trained AI model — one that has already been trained on a large general dataset — and continuing to train it on a smaller, domain-specific dataset to improve its performance on a particular task or in a particular context. A general-purpose language model might know a lot about the world, but fine-tuning can make it significantly better at, for example, drafting legal contracts, responding to customer service queries in a specific brand voice, or classifying medical documents. Fine-tuning is a technical undertaking and generally requires labelled data, computing resources, and ML expertise. For most business teams, the more accessible alternative is RAG (Retrieval-Augmented Generation) or well-designed prompt engineering, which can achieve similar results without the infrastructure overhead. → The AI Strategy & Roadmap helps teams understand when fine-tuning is warranted versus lighter-weight alternatives.
G
- Generative AI
- Generative AI refers to AI systems that can produce new content — text, images, audio, video, code, or structured data — rather than simply classifying or analysing existing content. The term covers a wide range of tools: large language models that write text (Claude, GPT-4, Gemini), image generators (Midjourney, DALL-E), code generators (GitHub Copilot, Cursor), and multi-modal systems that work across content types simultaneously. What makes generative AI significant for business is not just that it can produce content, but that it can produce contextually appropriate, high-quality content at a speed and scale that would not be feasible for a human team alone. The practical implication is that many knowledge-work tasks that previously required significant human time — first drafts, research summaries, meeting notes, data analysis narratives — can now be substantially accelerated. Understanding the quality and reliability characteristics of these outputs is what separates productive AI use from problematic AI dependence. → Applied practically in every programme across the Learn pillar.
H
- Human in the Loop
- Human in the loop (HITL) describes an AI system design where a human reviews, validates, or approves AI outputs at one or more points in the process — before the output is acted upon or used downstream. It is the design principle that maintains human accountability and oversight in AI-assisted workflows. The degree of human involvement varies: at one end, a human reviews every single AI output before it proceeds; at the other end, humans only intervene when the AI flags uncertainty or an output exceeds defined risk thresholds. Choosing the right level of human involvement is a governance and risk management decision, not just a technical one. In high-stakes contexts — clinical decisions, financial advice, legal judgements — human in the loop is not optional. In lower-stakes contexts, reducing HITL friction is often what makes AI adoption genuinely efficient. → HITL design is a core topic in the AI Exec Briefing and AI Governance frameworks.
L
- LLM (Large Language Model)
- A Large Language Model (LLM) is a type of AI system trained on enormous quantities of text — books, websites, code, scientific papers, and more — to learn patterns in language and develop the ability to generate, summarise, translate, and reason about text. Models like Claude (Anthropic), GPT-4 (OpenAI), and Gemini (Google) are LLMs. They are called "large" because of the scale of both the training data and the model parameters — typically hundreds of billions of mathematical values that encode what the model has learned. LLMs are the foundation of most generative AI applications that organisations are deploying today, including chatbots, writing assistants, code generators, and document analysis tools. Their key limitations — including the tendency to hallucinate plausible-sounding but incorrect information — are important for any business user to understand before deploying them in production contexts. → Understanding LLM capabilities and limits is covered in depth in AI Learning Shots and the AI Skills Bootcamp.
M
- MVP (AI MVP)
- In AI contexts, an MVP (Minimum Viable Product) is the smallest functional version of an AI-powered product or workflow that delivers real value to real users and can be tested, measured, and iterated upon. An AI MVP is not a proof of concept or a demo — it is a production-grade (or near-production-grade) tool that actual users interact with as part of their real work. The AI MVP concept is important because it resists the tendency toward over-engineering: rather than building the perfect AI system in a long development cycle, you build the minimum version that tests the core value hypothesis quickly. In practice, a well-scoped AI MVP can often be built in two to four weeks using existing models and low-code infrastructure, without requiring a full engineering team. → Anteligen's AI MVP Full Build programme delivers a production-ready AI MVP in ten business days.
P
- Prompt Engineering
- Prompt engineering is the practice of designing and refining the inputs you give to an AI model in order to get better, more consistent, and more useful outputs. A prompt is not just a question — it is an instruction, a context, a constraint, and sometimes an example, all combined into a single input that shapes how the model responds. Good prompt engineering makes the difference between an AI that produces vague, generic output and one that produces accurate, specific, usable content. Techniques include role prompting (telling the model to act as a specific expert), chain-of-thought prompting (asking the model to reason step by step), few-shot prompting (providing examples in the prompt), and system prompting (setting persistent instructions that apply across a conversation). Prompt engineering is a skill that compounds quickly — an hour of structured practice produces noticeable improvement, and the investment pays off in every AI interaction afterwards. → Prompt engineering is a hands-on module in every AI Skills Bootcamp track.
R
- RAG (Retrieval-Augmented Generation)
- Retrieval-Augmented Generation (RAG) is a technique that combines a language model's generative capabilities with a retrieval system that pulls in relevant, specific information from an external knowledge base before generating a response. Rather than relying solely on what a model learned during training — which may be outdated, incomplete, or insufficiently specific — a RAG system first searches a curated database of documents, policies, product information, or other source material, and then uses that retrieved content to ground the model's response in accurate, up-to-date information. RAG is how organisations build AI tools that can answer questions about their own internal knowledge — company policies, product documentation, client records — without exposing that data to third-party training processes. It is currently one of the most practical and widely adopted approaches for building enterprise AI applications. → RAG architecture is covered in the AI Prototype Sprint and AI MVP Full Build programmes.
V
- Vibe Coding
- Vibe coding is a term — popularised in early 2025 — for a style of software development where the programmer describes what they want in natural language, and an AI code generator writes most or all of the actual code. The developer's role shifts from writing syntax to directing, reviewing, and iterating on AI-generated output. "Vibe" captures the intuitive, conversational, intent-driven nature of the process. For non-technical business professionals, vibe coding is genuinely significant: it dramatically lowers the barrier to building functional tools, internal dashboards, automation scripts, and simple applications. It does not replace engineering expertise for complex, production-critical systems — but it makes a wide range of useful software accessible to people who would previously have needed to commission a developer. Knowing how to vibe code effectively is increasingly a practical business skill, not just a developer curiosity. → Vibe coding is a core skill in the AI Rapid Prototyping programme — no coding background required.
Missing a term? Want the full guide?
Download the AI fluency starter guide — the full glossary plus a team self-assessment.
Every transformation starts with a free AI Maturity Assessment.
30 minutes. Senior consultant. No commitment. Results visible within weeks.
Book my free AI Maturity Assessment