Glossary

Measures of whether people use AI and benefit from it, such as weekly active users, tasks per person and time saved, tracked after the rollout.

Claude Enterprise rollout checklist

An open protocol that lets AI agents built by different vendors discover each other and work together on a task. It complements MCP, which connects agents to tools and data.

AI systems that plan, use tools and act on goals, often with several agents working together.

Full definitionAI agents for business

Software that uses a model to decide its next step and acts through tools until a task is done.

Full definitionAI agents for businessCustom AI agent development

A small cross-functional team that sets standards, shares what works and supports business units as they adopt AI.

AI strategy consulting

The policies, roles and controls that decide which AI is used, for what and under which rules, and keep them current as tools and laws change.

AI governance

The skills and understanding that people need to use AI responsibly. Article 4 of the EU AI Act requires providers and deployers to take measures to ensure it among their staff.

AI literacy training

The policies, processes and controls that an organisation uses to develop, provide or use AI responsibly, as set out in ISO/IEC 42001.

ISO/IEC 42001 readiness

The body within the European Commission that supervises general-purpose AI models and supports a consistent application of the EU AI Act across member states.

EU AI Act checklist

A time-boxed rollout to a real team doing real work, with success criteria agreed before it starts. A Claude Enterprise pilot typically runs about 3 weeks with up to 50 people.

Claude Enterprise rollout checklist

A company's written rules for using AI: the approved tools, the data that may and may not be shared, review duties and how to report a problem.

AI governance

A short review of goals, data, systems, security and skills that ends with prioritised use cases and a plan with costs.

AI readiness assessment

A voice agent that answers a company's calls at any hour, handles common requests, books appointments and passes the call to a person when needed.

AI receptionist

An inventory of every AI system a company uses or builds, with its purpose, owner, risk level and the obligations that apply. It is the first step of EU AI Act compliance.

EU AI Act checklist

Under the EU AI Act, a machine-based system that operates with some autonomy and infers from its input how to generate outputs, such as predictions, content, recommendations or decisions, that can influence physical or virtual environments.

EU AI Act checklist

Automating a business process end to end with AI agents that read, decide and act across systems, with people approving the steps that matter.

AI workflow automation

A defined way for software to talk to other software. AI models are offered through APIs so that companies can build them into their own applications and workflows.

A record of who did what and when in a system, including the use of AI, kept for security reviews, investigations and compliance.

Trust Center

The training, communication and support that help people change how they work, so that a new AI tool is used rather than ignored.

AI training for employees

OpenAI's plan for organisations, with admin controls, single sign-on and business data excluded from model training by default.

Claude Enterprise vs ChatGPT Enterprise

The family of AI models and the AI assistant made by Anthropic, used for writing, analysis, research and coding.

Claude Enterprise

Anthropic's agentic coding tool, included in Claude Team and Enterprise seats. It reads a codebase, makes changes and runs commands under the developer's control.

Claude Code for teams

The way of working in Claude where you hand over a task instead of asking a question: Claude plans it, works across your files and connected tools and delivers the result.

Claude Cowork

Anthropic's plan for organisations, with enterprise security, admin controls and integrations for company-wide use.

Claude EnterpriseClaude Enterprise rollout checklist

Anthropic's plan for groups of 2 to 150 people, with Standard and Premium seats, single sign-on, usage analytics and spend controls.

Claude Team or Enterprise

The ability of an agent to operate a computer as a person does, by reading the screen, clicking and typing, for tasks in applications that have no API.

The process that shows a high-risk AI system meets the requirements of the EU AI Act before it is placed on the market, followed by the CE marking and registration in the EU database.

EU AI Act checklist

A ready-made link between an AI assistant and a business tool, such as a document library, a chat platform or a CRM, so that it can read and act on that tool's data with each user's permissions.

MCP integrations

The amount of text, counted in tokens, that a model can consider at once: the conversation, the instructions and any attached documents. A larger window lets a model work through a long contract or codebase in one pass.

Microsoft's low-code tool for building agents and extending Microsoft 365 Copilot with company data and actions.

Claude vs Microsoft Copilot

An assessment that the GDPR requires before processing personal data in a way likely to create a high risk to people, which many AI projects do.

Where data is stored and processed geographically, for example in the EU. It matters for regulated sectors, public bodies and contracts with data location clauses.

Sovereign AI

Under the EU AI Act, an organisation that uses an AI system under its authority in a professional activity. Most companies that buy AI tools are deployers, with duties such as AI literacy, human oversight and transparency.

EU AI Act checklist

A package of changes to EU digital rules proposed by the European Commission in November 2025, which also changes when parts of the AI Act apply. Every current date is in our EU AI Act checklist.

EU AI Act checklist

Numbers that represent the meaning of a piece of text, so that texts with similar meaning sit close together. They power semantic search and retrieval-augmented generation.

Regulation (EU) 2024/1689, the European Union's law on artificial intelligence. It entered into force on 1 August 2024 and sets obligations by level of risk for the providers and deployers of AI systems used in the EU.

EU AI Act checklistEU AI Act compliance

Tests that measure how well an AI system does its task, run on a fixed set of real examples before launch and after every change. They show whether an agent is ready for production.

Custom AI agent development

Training an existing model further on examples for a specific task or style. For most company use cases, good prompts and retrieval from your own documents deliver more, sooner and at lower cost.

A large model trained on broad data that can be adapted to many tasks, from writing to coding to analysis. Most enterprise AI products are built on a few foundation models from a handful of providers.

An experienced AI leader who works part time for a company, setting its AI strategy, governance and roadmap without a full-time hire.

Fractional Chief AI Officer

An assessment that Article 27 of the EU AI Act requires before first use from certain deployers of high-risk systems, such as public bodies and companies that score credit or price life and health insurance.

EU AI Act checklist

Regulation (EU) 2016/679, the EU's data protection law. It applies to every use of AI that processes personal data, alongside the AI Act.

Trust Center

Google's platform for using Gemini models and agents at work, connected to company data in Google Workspace and other systems.

Claude vs Gemini Enterprise

A voluntary code, published in July 2025, that helps providers of general-purpose AI models show that they meet the obligations of the EU AI Act on transparency, copyright, and safety and security.

EU AI Act checklist

A model trained on broad data that can perform a wide range of tasks, such as the models behind Claude, ChatGPT and Gemini. Its providers have their own obligations under the EU AI Act, including technical documentation and a copyright policy.

EU AI Act checklist

Tying an answer to sources that can be checked, such as company documents or databases. Retrieval-augmented generation is the most common way to ground answers.

Retrieval-augmented generation

Controls that keep an AI system within agreed limits, such as the tools it may use, spending caps, content filters and the steps that need human approval.

An answer that sounds plausible but is false or unsupported, such as an invented citation or figure. Grounding answers in your own documents, asking for sources and human review reduce the risk.

An AI system that the EU AI Act classifies as high-risk, either as a safety component of regulated products (Annex I) or in uses such as recruitment, credit scoring, education and access to essential services (Annex III). It must meet strict requirements before and during use.

EU AI Act checklist

A design in which a person reviews or approves an AI decision before it takes effect, such as paying an invoice or onboarding a supplier above a set amount.

Custom AI agent development

Running a trained model to produce an answer. Inference is what you pay for when you use an AI product or API, and where response time and cost are decided.

The international standard for an AI management system, published in December 2023. Certification shows that a company governs its AI in a structured, auditable way.

ISO/IEC 42001 readiness

A prompt designed to make a model ignore its safety rules. Providers train their models against them; companies add their own guardrails and monitoring.

The documents, policies and data an AI system draws on to answer, such as contracts, manuals and procedures. Their quality and permissions decide the quality of the answers.

A model trained on large amounts of text to predict the next token, which lets it write, summarise, translate, analyse documents and follow instructions. Claude, GPT and Gemini are large language models.

Claude Enterprise

A program that connects AI applications to a tool or data source through the Model Context Protocol.

Full definitionMCP integrations

Microsoft's AI assistant inside the Microsoft 365 apps, such as Word, Excel, Outlook and Teams, working with the organisation's Microsoft 365 data.

Claude vs Microsoft Copilot

An open standard, introduced by Anthropic in November 2024, for connecting AI applications to tools and data sources in one consistent way.

MCP integrations

Several AI agents that each handle part of a task and pass work between them, for example one that researches, one that drafts and one that checks.

Agentic AI consulting

AI that works with more than one kind of input or output in the same task, such as text, images, documents, audio and code.

A voluntary framework from the US National Institute of Standards and Technology, released in January 2023, for identifying and managing the risks of AI systems.

Uses banned outright by Article 5 of the EU AI Act, such as social scoring, manipulative techniques that cause significant harm and emotion recognition at work or in education, with narrow exceptions.

EU AI Act checklist

The instructions and material given to a model for a task. A clear goal, the right context, examples and the expected format make the answer more reliable.

Prompt engineering training

The practice of writing and testing prompts so that a model does a task well and consistently. In a company it is a skill for everyone who uses AI, not only for developers.

Prompt engineering training

An attack that hides instructions in content an AI system reads, such as a web page, an email or a document, to make it ignore its rules or leak data. Agents that browse or read inboxes need defences against it.

Under the EU AI Act, the organisation that develops an AI system or a general-purpose AI model, or has one developed, and places it on the market or puts it into service under its own name. Providers carry most of the obligations for high-risk systems.

EU AI Act checklist

A model that works through a problem step by step before it answers, trading some speed for accuracy on complex tasks such as analysis, planning and code.

Testing an AI system by deliberately trying to make it fail or misbehave, before attackers or users do, and fixing what is found.

AI that answers from your own documents, and cites them.

Full definition

Software robots that repeat fixed steps in user interfaces, such as copying data between systems. AI agents add judgement: they read unstructured documents and handle cases that a fixed script cannot.

AI workflow automation

Permissions given by role, so that people and agents can only see and do what their job requires.

A standard that creates, updates and removes user accounts automatically from the identity provider, so that access to AI tools follows joiners, movers and leavers.

Claude Enterprise rollout checklist

Search by meaning rather than by exact words, so that a question about notice periods also finds the clauses on termination.

AI tools used at work without the company's approval or controls, often personal accounts of public chatbots. Giving people an approved, governed tool is the most effective way to reduce it.

Claude Enterprise

Signing in to many applications with one company account, managed in an identity provider. It is the first setting of any enterprise AI rollout.

Claude Enterprise rollout checklist

Folders of instructions, scripts and resources that Claude loads when a task needs them, so that it follows a company's own procedures and formats.

Claude Enterprise

Control over where AI runs, who can access it and which law applies.

Full definition

Standing instructions that set how a model behaves across a conversation or an application, such as its role, tone, rules and the tools it may use. Users usually do not see it.

Under the EU AI Act, the category of the most capable general-purpose AI models, presumed when the compute used for training exceeds 10 to the power of 25 floating-point operations. Their providers must assess, mitigate and report serious risks.

EU AI Act checklist

The unit a language model reads and writes: a word, part of a word or a punctuation mark. Usage limits, context windows and API prices are counted in tokens.

The ability of a model to call external functions, such as searching a database, sending an email or creating a record, and to use the result in its answer. It is what turns a chat model into an agent.

Custom AI agent development

The duties in Article 50 of the EU AI Act to tell people when they interact with an AI system, and to mark AI-generated or manipulated content such as deepfakes.

EU AI Act checklist

Ranking possible AI use cases by value and feasibility, so that the first projects are the ones most likely to pay off.

AI strategy consulting

A database that stores embeddings and finds the passages closest in meaning to a question. It is a common building block of retrieval-augmented generation.

An AI agent that speaks and listens, on the phone or in an app, to answer questions, book appointments or route calls in natural conversation.

AI voice agents

An arrangement in which a provider does not store prompts or outputs once a request is processed, except where the agreement or the law requires it.