Dictionnaire de la Préparation d'Entreprise

A quick regular check of who has access to which business systems, so that ex-staff, old contractors and unused accounts don't quietly keep their keys to the door.

How often an AI tool produces correct, useful output.

A simple plan for introducing AI over time — which use cases first, who is trained when, what policies apply, and how progress is measured.

Software that can work toward a goal using instructions, tools and approved access — for example, drafting a customer reply, then waiting for a human to review and send it.

A chat-style tool that helps a person work faster — drafting, summarising, answering questions and organising notes — while the person stays in charge of the result.

The set of rules, roles and checks a business uses to keep artificial-intelligence tools safe, accurate and appropriate for the work it does.

How consistently and safely a business already uses AI — from occasional individual use, to team workflows, to organisation-wide processes with clear rules and review.

A short written document that tells staff which AI tools are approved, what information must never be pasted into them, and when a human review is required before sending or acting on AI output.

How prepared a business is to adopt AI tools responsibly — covering staff training, data hygiene, approved use cases and human review.

The set of practical risks that come with using AI — inaccurate output, data leakage, reputational damage, over-reliance and unclear accountability.

A defined set of steps a computer follows to produce a result.

A structured way for two software systems to talk to each other.

Software that performs tasks normally associated with human thinking — reading, writing, summarising, comparing and pattern-spotting — using models trained on large amounts of information.

Using software to handle a repetitive task from start to finish, sometimes with AI in the middle.

A separate copy of the business's important files, kept somewhere safe so they can be restored if the originals are lost, deleted or locked by an attacker.

When an AI system produces uneven or unfair results because of the data it learned from.

When an attacker either takes over a real business email account or convincingly impersonates one, then asks a customer or bookkeeper to send a payment to a new bank account.

A protected sum of cash a small business keeps aside so it can keep paying staff, rent and suppliers during a disruption.

An AI software development assistant that understands an existing codebase and helps developers build, improve and maintain software.

Files stored on a provider's servers over the internet rather than only on a local machine.

Business information — pricing, contracts, employee data, customer records — that should stay inside the business.

The background information an AI assistant is given alongside a request — company details, examples, tone — so its answer fits your business rather than being generic.

Protecting personal and confidential information from unnecessary exposure.

Using AI to prepare summaries, comparisons and options that help a person decide — without letting the AI make the decision itself.

A workplace safety document covering how staff evacuate, who is accounted for, and how emergencies like fire, medical events or severe weather are handled.

Turning information into a scrambled form that only the right person or device can read.

Any device that connects to the business's systems — a laptop, desktop, phone or tablet.

Modern security software that runs on laptops, desktops and phones.

Whether an AI tool can show why it produced a particular output.

The likelihood and business impact of water entering a workplace, whether from rivers, drains, storms or plumbing failures.

The European data protection law that governs how businesses collect, store and use personal information about customers and staff.

AI that produces new content — text, images, code or summaries — in response to a request.

When an AI system produces information that sounds confident but is inaccurate or invented.

A range of health problems caused by the body overheating — from cramps and heat exhaustion through to heat stroke, which is a medical emergency.

A person checking AI output before it is used or sent.

A workflow where AI does the first draft or analysis and a person makes the final decision.

Connecting an AI tool to systems the business already uses — email, calendar, CRM, documents — so it can help with real work instead of standalone chat.

An internationally recognised standard for how a business manages information security.

The type of AI model behind assistants like ChatGPT and Claude.

A Large Language Model — the type of AI behind tools like ChatGPT and Claude.

A safety procedure for physically switching off and clearly tagging machinery before maintenance, so it can't be turned back on while someone is working on it.

The branch of AI where software learns patterns from examples instead of being told every rule.

A general word for harmful software — viruses, ransomware, spyware and so on.

The underlying AI engine that produces answers.

A second sign-in step, usually a short code from an app or text message.

Tools that let non-developers build workflows, forms or small applications by clicking and configuring rather than writing code.

The U.S.

A secure app that remembers strong, unique passwords for every business system so staff don't have to.

The written rules a business sets for how staff choose, store and change passwords — for example, minimum length, no reuse across systems, and using a password manager.

A small update from the maker of the software that fixes a known weakness.

Any information that identifies a person — name, email, phone, ID, health, employment details.

A fake message — usually an email — that pretends to be from someone you trust to trick a staff member into clicking a link, sharing a password or paying an invoice.

A limited real-world rollout — one team, one workflow, one month — used to learn what works and what doesn't before adopting AI more widely.

A loss of electrical supply that can halt POS systems, refrigeration, lighting and access controls until power is restored or a backup source kicks in.

The practical measure of AI's business value: time saved on real tasks, at acceptable quality, after human review.

The instructions you give an AI tool.

The practical skill of writing clear instructions to an AI assistant so it produces useful output.

A type of attack where hidden instructions inside a document, email or webpage try to manipulate an AI assistant into doing something it shouldn't.

A small, time-boxed test to see whether an AI tool actually works for the business before committing money, training and process change to it.

A fake QR code — often stuck over a real one in a car park, on a menu or in an email — that sends people to a scam website designed to steal payment or login details.

A type of attack that locks a business out of its own files and demands payment to release them.

Using AI in a way that is fair, accurate, transparent, respectful of privacy and clearly accountable — with a person answerable for outcomes, not the tool.

Staff using personal or unapproved AI tools for work tasks without oversight — often with sensitive business data pasted into public chatbots.

Any event — weather, cyber, geopolitical, supplier failure — that interrupts the flow of goods or services a business depends on.

The information a model learned from.

Being open about where AI is used in the business — with staff, customers and, where required, regulators.

A specific task where AI is expected to help — for example, drafting quotations or summarising customer emails.

A private, encrypted connection between a staff member's device and the business's systems.

A specialist measurement combining heat, humidity, wind and sunlight to estimate how hard it is for the body to cool itself.

The sequence of steps a business follows to complete a task — quoting a customer, onboarding a new hire, closing the books.

A modern security approach that assumes no device or user is automatically trusted, even inside the business network.