AI Kit for Communication

Free resource in French · Word + Excel Kit

AI Kit for Communication: 27 Use Cases and Their Prompts, in Three Levels

From tone guidelines to responding to hyper-manipulation: ready-to-use prompts so that your internal, external, and press relations teams can develop their skills on their own, step by step.

📄 Word, 48 pages🧭 27 cards, 3 levels📊 Excel tool workbook🆓 Free
123

The communicator's AI toolkit, in short

What it is
A self-training kit for communication teams: 27 everyday use cases with prompts, classified into three levels to progress without prior training, and an Excel workbook of tools.
For who
The internal and external communications teams, press relations, employer branding and speechwriters for leaders of large organizations, as well as the communications correspondents for the business lines and subsidiaries.
Format
48-page Word document: user guide, 27 one-page worksheets, a validation challenge for each level, a section for the team and communications management, six appendices. Eleven-tab Excel workbook.
What it contains
Self-positioning in twelve questions, prompt framework in four ingredients and five prompt reflexes 2026, five security rules specific to communication, a copyright and AI section, 27 prompts ready to copy, tone charter, editorial calendar, variation matrix, question and answer bank, transparency register of generated content, crisis plan, statement of presence in AI.
What is it based on?
Article 50 of the European AI regulation, the Commission's code of good practice and guidelines, the ARPP fact sheet, Google Search Central documentation, Anthropic's AI Fluency framework, and our workshops with communication teams from large organizations.
Price
Free, in exchange for a professional email address. No unsolicited emails.
Update
Version 2 of October 26, 2026.
What it doesn't replace
Your organization's AI charter, the opinion of your legal department, or a workshop on your own content with the team.
Go Further
AI training for communications professionals, AI and Social Dialogue Kit, Manager's toolkit for dealing with AI.

Most communication teams in large organizations have a generative AI tool at their fingertips: Copilot in Word and Teams, a corporate ChatGPT or Claude account, Gemini in Workspace, or Vibe from Mistral. What's lacking isn't the tool itself, but a process of improvement. Everyone tries it on their own, ends up with a polished text that doesn't reflect the organization's identity, a quote that was never uttered, or an image whose rights are unknown, and concludes that AI isn't suited for communication. Meanwhile, a minority integrates it into their routines: a tone guide built into an assistant, variations in a single pass, and monitoring that runs automatically.

This page explains what recent surveys show about the use of AI by communicators, what Article 50 of the European regulation on AI changes from August 2, 2026, why a progression by level is better than a list of prompts, what the 27 use cases of the kit contain, and gives you the complete kit, in Word format, with its Excel workbook.

1. AI in communication teams: what the numbers show

Professional surveys published since 2024 converge on three points: the use of generative AI has become predominant among communicators, primarily for proofreading, searching and writing; usage policies and training are progressing but remain behind practices; and the main fear in the profession is the publication of unverified content.

The context changed in 2026: since July 22nd, Google has been displaying generated answers (AI Overviews and AI Mode) above links in France, and journalists, job seekers, and customers are increasingly asking AI assistants about organizations before reading their websites. The risks, however, are well-known: fabricated quotes, rounded figures presented as facts, visuals that mimic real work, and the use of a leader's voice or face.

Since August 2, 2026, the framework has changed. Article 50 of the European AI regulation requires clear labeling of generated or modified images, sounds, and videos that appear authentic, and of generated texts published to inform the public on matters of general interest when no one assumes editorial responsibility after a human review. This labeling must appear no later than the first exposure. The Commission published a code of good practice on labeling and marking in June 2026, followed by its guidelines on July 20; the ARPP (French Advertising Standards Authority) published a practical guide for advertising on August 3; and in the area of ​​commercial influence, a face or silhouette produced by AI requires the label "Virtual Images." Violations of Article 50 are subject to penalties of up to €15 million or 3% of global turnover. For a communications team, human review and traceability are no longer just good practices.

2. Why a progression by level, and not just a list of more prompts?

There's no shortage of prompt libraries for communications: tool publishers, agencies, and PR platforms all publish them. They share three common flaws. They're bland: fifty prompts presented out of order, leaving beginners unsure where to start. They stop at the AI ​​output: nothing on what to prepare beforehand, nothing on what to check afterward, even though that's precisely where fabricated quotes, unsourced figures, and images with questionable copyrights come into play. And they ignore the context of a large organization: embargoes, approved talking points, executives, works councils, crises, and international subsidiaries.

The toolkit takes a different approach to the problem. It follows the three ways of working with AI described in the AI ​​Fluency framework published by Anthropic in 2025: automation, where AI performs a specific task on command; augmentation, where humans and AI build together in multiple rounds; and agency, where humans configure AI to then perform tasks on their behalf. These three modes correspond to three levels, and the four skills from the same framework (delegating, describing, discerning, and being diligent) provide the self-assessment grid. Each level covers nine real-world situations, from editorial guidelines to crisis management and communication about AI itself, making the toolkit also suitable for on-demand reading.

Each fact sheet finally explains what the lists don't: when to use the case, what to prepare, what you should achieve, the three points to check before publishing, the variations depending on the tool, and the next fact sheet. And the first fact sheet in the kit, the tone guide, serves as the foundation for all the others: it's what ensures the AI ​​writes like your organization, and not like everyone else.

3. The three levels and the six-week course

LevelWorking methodWhat you know how to do at the endIndicative duration
Level 1 · StartThe AI ​​performs one task at a time, based on what you give it.Get in minutes a tone guide, a rewritten text, five variations, an internal note, talking points, a summary of media coverage, a visual brief, a localized version or a record of what AI assistants say, which you publish after verification.2 weeks, 20 to 30 minutes per day
Level 2 · StructuringCo-constructing AI: you provide your documents, you engage in dialogue over several rounds, you make it work in stages.Produce complete deliverables: communication plan and calendar, article from an interview, social calendar, communication of a change and its FAQ, executive speech, press kit, video storyboard, crisis preparation, internal communication on the deployment of AI.2 weeks, 30 to 45 minutes per day
Level 3 · SystematizeYou configure the AI ​​to work for you and the team: standing instructions, projects and agents, scheduled tasks, prompt chains.Having an editorial line assistant, a newsletter channel, a variation matrix, a reaction barometer, a speechwriter assistant for the executive, a media watchdog, a transparency register, a crisis exercise and a prepared response to hyper-rigging.2 weeks, 45 minutes a day, then a few minutes a week

The proposed program takes six weeks: four worksheets per week, presented in the order of the situations, and a validation challenge at the end of each level, based on a real-world topic from your communication plan. At level 1, rewrite and adapt a real advertisement according to your tone guidelines. At level 2, have your client approve a complete deliverable produced using AI. At level 3, implement the editorial line assistant and a tool that a colleague uses independently. The twelve-question self-assessment indicates your entry level; the progress log and self-evaluation grid in the appendices measure your progress.

4. The 27 use cases, from editorial policy to crisis and AI

Nine situations, three levels, one sheet per box. The situations are based on those we work on in the AI training for communications professionalscomplemented by press relations, transparency of generated content and, new in version 2, communication in the age of AI.

SituationLevel 1 · StartLevel 2 · StructuringLevel 3 · Systematize
Strategy and editorial lineFormalize your tone and style guidelines based on five reference textsDevelop a communication plan and editorial calendar based on a briefCreate an assistant guardian of your editorial line
Content writingRewrite an existing text: make it clearer, shorter, and in your own style.Write an article or newsletter based on an interview and internal documentsThe newsletter production chain: collection, sorting, writing, proofreading
Multichannel versionAdapt an announcement to five formats: intranet, email, Teams, LinkedIn, home screenConstructing a social calendar and variations by audienceThe adaptation matrix: from core content to formats, channels and languages
Internal communication and changeTransforming a management decision into a clear internal memoDevelop a communication plan for a change, population by population, including a FAQThe reaction barometer: analyze comments, questions and surveys, and adjust
Speeches and leadersPrepare talking points and difficult questions for a topicWrite a speech or leadership address, then rehearse it with a simulated audience.The executive's speaking assistant: their voice, their positions, their bank of questions and answers
Press relations and monitoringSummarizing a press review and the impact of an announcementWrite a press release and press kit with a specific angle, quotes, and media briefs.Media and social media monitoring: mentions, signals, responses to prepare, every week
Visuals, video and multimediaWrite a visual brief for a graphic designer, studio, or image generator.From the message to the storyboard and video script, with voice-over and subtitlesThe visual library and the transparency register for generated content
Crisis, sensitivity and internationalTo translate and localize a message for another country, not just to translate itPreparing crisis communication: scenarios, statement of expectation, Q&AThe simulated crisis exercise: AI-powered crisis simulation for the cell, then lessons learned
Communicating in the Age of AICheck what AI assistants say about the organizationCommunicate internally about the deployment of AI, and make it a key element of employer branding.Preparing a response to a deepfake targeting a leader

The details of each case follow: the situation, what you get, and the point to check before using the result. The complete prompt is in the kit.

1

Start: AI executes

9 case

Cas 1.1Strategy and editorial line

Formalize your tone and style guidelines based on five reference texts

  • When the "voice" of the organization exists in people's heads but not on paper.
  • Time saved: 2 hours, once, then at each prompt
  • To check: each example cited exists word for word in your texts: do a search on two or three.

Deliverables: a two-page charter, illustrated with your own sentences; ten "yes/no" rules and a homemade glossary with replacements.

Cas 1.2Content writing

Rewrite an existing text: make it clearer, shorter, and in your own style.

  • When a text arrives from a profession, an expert or a management, it is correct in substance but unreadable.
  • Time saved: 30 minutes per text
  • To check: compare figures, dates and names with the original: they must be identical.

Deliverable: a rewritten, shorter version, in your tone, without loss of information; the list of deletions and ambiguities to be decided by the author.

Cas 1.3Multichannel version

Adapt an announcement to five formats: intranet, email, Teams, LinkedIn, home screen

  • When the same message needs to be sent to multiple channels at the same time and you rewrite the same thing five times.
  • Time saved: 1 hour per ad
  • To be verified: the external version does not contain any internal or embargoed information.

Deliverable: five publication-ready versions, consistent with each other; a checklist of facts included and omitted by format.

Cas 1.4Internal communication and change

Transforming a management decision into a clear internal memo

  • When management communicates a decision to you (reorganization, new tool, new rule) in the form of a report or presentation, and you have to announce it to the teams without jargon, without omitting what changes for them, and without creating unnecessary anxiety.
  • Time saved: 45 minutes per note
  • To be verified: nothing that "cannot yet be announced" appears, even by allusion.

Deliverable: a 300-word note that says what is changing, for whom and when; the ten questions that readers will ask themselves, and those left unanswered.

Cas 1.5Speeches and leaders

Prepare talking points and difficult questions for a topic

  • Before an executive, spokesperson or manager speaks on a sensitive or new topic.
  • Time saved: 1 hour per topic
  • To be verified: each answer remains in fact validated: an invented answer becomes a quote.

Deliverable: three key messages with their evidence and variants; ten difficult questions and their short answers.

Cas 1.6Press relations and monitoring

Summarizing a press review and the impact of an announcement

  • After an announcement, an event or a crisis, when management wants to know "what the media said" and you have thirty articles and hundreds of comments to read.
  • Time saved: 1 hour per summary
  • To check: open three articles at random: the tone and the attributed quotes match.

Deliverable: a summary that can be read by management in two minutes; a table of impact assessments with justified tone and repeated messages.

Cas 1.7Visuals

Write a visual brief for a graphic designer, studio, or image generator.

  • When a campaign, article or event needs a visual and the brief is one line long ("a nice visual about the transformation"), this guarantees back-and-forth with the graphic designer or generic images.
  • Video and multimedia. Time saved: 45 minutes per brief
  • To be verified: no track reproduces a work, a character, a style of living artist or a brand.

Deliverable: a complete one-page graphic designer brief; three contrasting creative directions.

Cas 1.8Crisis

To translate and localize a message for another country, not just to translate it

  • When an advertisement needs to be sent to multiple countries or languages ​​and a literal translation is not sufficient.
  • Sensitivity and international reach. Time saved: 30 minutes per language
  • To be checked: a native proofreader validates the text before publication: the apparent fluidity masks misinterpretations.

Deliverable: a localized version, faithful to the glossary; the list of adaptations and their reasons.

Cas 1.9Communicating in the Age of AI

Check what AI assistants say about the organization

  • Since July 22, 2026, Google has been displaying AI Overviews and an AI Mode above links in France, and journalists, job seekers, and customers are using ChatGPT, Copilot, or Perplexity to ask questions about the organization. An incorrect answer (former executive, outdated figure, closed controversy) is circulating unnoticed.
  • Time saved: 1 hour per month, and errors spotted before journalists
  • To verify: answers vary from session to session: redo the statement each month with the same questions before concluding.

Deliverable: a list of errors and approximations, ranked by severity; the sources that feed the assistants on your organization and a correction plan.

2

Structuring: Co-constructed AI

9 case

Cas 2.1Strategy and editorial line

Develop a communication plan and editorial calendar based on a brief

  • When management entrusts you with a project to promote (launch, transformation, event, employer branding) with a thin brief, and you have to come back with a well-reasoned plan.
  • Time saved: 3 hours per plan
  • To be verified: the numerical objectives are set with the client, not by AI.

Deliverable: a simulated interview that reveals the questions to ask the real client; a plan by audience with messages, channels and risks of reception.

Cas 2.2Content writing

Write an article or newsletter based on an interview and internal documents

  • When you have to produce an in-depth article (profile, project, testimony) from a recorded interview or from notes and documents, and the first draft takes half a day.
  • Time saved: 2 hours per article
  • To check: every quote in quotation marks exists in the transcript: this is the most serious and most frequent mistake.

Deliverable: three comparative angles, then an article in your tone; faithful quotes, listed for validation by the interviewee.

Cas 2.3Multichannel version

Constructing a social calendar and variations by audience

  • When the social networks of the organization (or its leaders) are fed haphazardly, without a guiding thread, and each publication goes out without testing or variation.
  • Time saved: 3 hours per month of publishing
  • To be verified: the conclusions about "what works" are based on enough publications to not be random.

Deliverables: an analysis of what works for you, based on your figures; a monthly calendar per account.

Cas 2.4Internal communication and change

Develop a communication plan for a change, population by population, including a FAQ

  • When a project changes the work of entire populations (new tool, reorganization, relocation, new policy) and communication must be consistent over several months, carried by managers who have not followed the project.
  • Time saved: 1 day per project
  • To be checked: the dissemination schedule respects the information or consultation of the CSE when the project is submitted to it.

Deliverables: an impact analysis by population, in tabular form; a dated communication sequence by population and a manager kit.

Cas 2.5Speeches and leaders

Write a speech or leadership address, then rehearse it with a simulated audience.

  • Before a convention, New Year's greetings, a plenary meeting, a round table.
  • Time saved: 3 hours per intervention
  • To be checked: the leader rereads and says the text aloud: what is not said well is rewritten.

Deliverable: three structures compared, then a timed written text for oral presentation; an intervention in the leader's voice, limited to validated facts.

Cas 2.6Press relations and monitoring

Write a press release and press kit with a specific angle, quotes, and media briefs.

  • Before an external announcement (results, partnership, launch, appointment, commitment).
  • Time saved: 4 hours per case
  • To be verified: the spokesperson's quote is validated by him, word for word: an unvalidated quote never goes through.

Deliverables: three journalistic angles, then a sourced 400-word press release; a complete press kit with figures, chronology and questions and answers.

Cas 2.7Visuals

From the message to the storyboard and video script, with voice-over and subtitles

  • When a video is planned (testimonial, motion design, internal announcement video, tutorial) and the brief to the studio or videographer is unclear.
  • Video and multimedia. Time saved: 3 hours per video
  • To be verified: the people filmed or quoted have given their written consent; no real voice or face is generated or cloned without authorization.

Deliverables: three concepts with their tagline, then a shot-by-shot storyboard; a timed voice-over script and a subtitle file.

Cas 2.8Crisis

Preparing crisis communication: scenarios, statement of expectation, Q&A

  • In a calm manner, to prepare for plausible crises in your organization (industrial incident, cyberattack, questioning of a leader, product recall, social conflict).
  • Sensitivity and international reach. Time saved: 1 day per scenario
  • To be verified: in a real crisis, no non-public information is pasted into an unauthorized tool, and the crisis cell validates every word.

Deliverable: three realistic scenarios in four stages; a statement of expectation, messages per audience and a validation circuit.

Cas 2.9Communicating in the Age of AI

Communicate internally about the deployment of AI, and make it a key element of employer branding.

  • The organization is deploying Copilot or another assistant, automating tasks, or reorganizing teams around AI. Employees are wondering how this will change their jobs and careers; candidates are looking at what the employer has to say.
  • Time saved: 1 day per plan
  • To be verified: nothing is announced before the information-consultation of the CSE when the project is submitted to it (article L. 2312-8 of the Labour Code).

Deliverables: concerns by population, and those that remain unanswered; a six-month plan, a FAQ to be validated and three employer branding pieces of content faithful to internal reality.

3

Systematize: AI works for you

9 case

Cas 3.1Strategy and editorial line

Create an assistant guardian of your editorial line

  • As soon as you re-paste the tone charter, the validated messages and the in-house lexicon to each prompt.
  • Time saved: 2 hours per week, then
  • Please check: the attached documents do not contain any confidential or embargoed information, nor personal data.

Deliverable: an assistant who writes and proofreads in your voice, for the whole team; a compliance check of any text against the charter and language elements.

Cas 3.2Content writing

The newsletter production chain: collection, sorting, writing, proofreading

  • When the internal or external newsletter requires a person to spend a day per issue, between follow-ups to obtain topics, sorting, writing and formatting, and the quality varies from one issue to another.
  • Time saved: 3 hours per issue
  • To check: the verification table is actually checked: a broken link or a wrong date is immediately visible.

Deliverables: classified topics, identified gaps and ready follow-up actions; a reasoned summary and formatted articles.

Cas 3.3Multichannel version

The adaptation matrix: from core content to formats, channels and languages

  • When each important announcement is broken down into ten different pieces of content (internal and external channels, multiple languages, manager versions) and the inconsistencies between versions end up creating questions or errors.
  • Time saved: 1 hour per ad, instead of half a day
  • To be verified: external versions do not contain any internal information.

Deliverable: all variations of an advertisement in one pass, consistent with each other; language versions with their adaptations indicated.

Cas 3.4Internal communication and change

The reaction barometer: analyze comments, questions and surveys, and adjust

  • During a change project or a sensitive period.
  • Time saved: 2 hours per month, and rumors dealt with sooner
  • To be verified: the reactions are anonymized before processing and the processing has been validated by HR and the data protection officer.

Deliverable: the themes that really occupy the employees, counted and illustrated; the rumors detected early and the misunderstood messages.

Cas 3.5Speeches and leaders

The executive's speaking assistant: their voice, their positions, their bank of questions and answers

  • When you are the speechwriter for a leader: editorials, LinkedIn posts, speeches, interview answers, internal messages, and each text starts from scratch to find its style and positions.
  • Time saved: 2 hours per week for the executive's writing
  • To be verified: the manager has given his approval for this assistant and validates each text published in his name.

Deliverable: first drafts in the leader's voice, limited to his validated positions; a lively and dated question-and-answer bank.

Cas 3.6Press relations and monitoring

Media and social media monitoring: mentions, signals, responses to prepare, every week

  • When monitoring is limited to a Google alert or a paid tool whose exports no one reads, and management discovers an article or controversy before the communication.
  • Time saved: 3 hours per week
  • To check: every Monday, open the links of the signals that you report to management.

Deliverable: a short, dated and sourced weekly summary; negative signals with a treatment recommendation.

Cas 3.7Visuals

The visual library and the transparency register for generated content

  • As soon as the team produces images, videos, or voice recordings using AI, it must be reported. Since August 2, 2026, Article 50 of the European AI regulation requires the reporting of "hyper-realistic manipulations" (generated images, sounds, or videos that appear authentic) and generated texts published to inform the public on topics of general interest, unless they have undergone genuine human review under editorial responsibility. The Commission's Code of Practice and the ARPP's practical guide outline the methodology; the register retains proof of compliance.
  • Video and multimedia. Time saved: 1 hour per week, and controlled legal risk
  • To be verified: the internal rules for mentioning have been validated by the legal department: the AI ​​applies your rules, it does not set them.

Deliverable: for each piece of content generated, the notice to be added and the rights to be verified; an up-to-date transparency register, available for use in case of questions.

Cas 3.8Crisis

The simulated crisis exercise: AI-powered crisis simulation for the cell, then lessons learned

  • Once or twice a year, to train the crisis team and spokespeople on a realistic scenario, without mobilizing an exercise provider.
  • Sensitivity and international reach. Time saved: Half a day per exercise, and a well-oiled machine.
  • To be verified: the exercise uses a fictitious organization and data: no real internal document is provided to the AI.

Deliverable: a realistic, paced crisis exercise adapted to the cell's responses; a structured debriefing with chronology and deadlines.

Cas 3.9Communicating in the Age of AI

Preparing a response to a deepfake targeting a leader

  • A generated video, audio, or image makes the executive say or do things they never said; a fake voice calls the finance department requesting an urgent transfer; a fake press release circulates. The response must be prepared in the first few hours: it must be planned in advance, with legal and cybersecurity expertise.
  • Time saved: hours saved on the big day
  • To be verified: the procedure was played in exercise (sheet 3.8), and the legal department validated the appeals (article 226-8 of the Penal Code).

Deliverable: a verification procedure with responsibilities and deadlines; three ready texts (expectation, denial, internal message) and a directive against fraud to the president.

5. What a record contains, with an example

Each sheet fits on one page and follows the same format: a heading (situation, level, compatible tools, approximate time saved), "When to use it," "What to prepare," the prompt to copy with its sections to replace highlighted, "What you should achieve," "Check before using," "Variations depending on the tool" (Copilot in Word, Outlook, Teams, or PowerPoint; ChatGPT, Claude, and Gemini with their Projects, Gems, and voice modes; Gemini Notebook; Vibe), and "Next step." Here, to set the tone, is the prompt for sheet 1.4, the internal memo based on a management decision.

Here is the decision made by [direction] : [paste the source document]. It relates to [populations]What changes for them: [what, when]What remains unchanged: [What]What cannot yet be announced: [elements].

Write an internal memo of no more than 300 words: 1) a title stating what is changing; 2) the decision in one sentence; 3) the reasons why, in two sentences, without mincing words; 4) what is changing for you, concretely and with a date; 5) what is not changing; 6) the next steps and who to contact with questions. Use a direct tone, no euphemisms, no superlatives.

Then list the ten questions that employees will ask themselves when reading this note, indicating those that the note does not answer.

And the three checkpoints that accompany it: nothing that cannot yet be announced appears, even by allusion; if the decision involves informing or consulting the Works Council, the dissemination schedule is validated with HR; the memo is validated by the decision-maker, because it is they who are making the commitment. It is this discipline, repeated across 27 checklists, that transforms a trial into competence.

6. The five rules of communication security

A communications professional handles information before anyone else: results, appointments, reorganizations, crises. What they publish commits the organization and its leaders. The toolkit's user guide sets out five rules, to be applied in addition to your organization's AI charter, which should never replace it.

  • Non-public information does not leave the authorized tool. Embargoed announcements, results, transactions, reorganizations, crisis information: only in the company tool approved by the IT department. In a listed company, information likely to influence the share price falls under market disclosure rules.
  • A quote is sacred. AI never creates a quote attributed to a real person; it suggests it, and the person validates it word for word. Any statement in quotation marks is checked against the source.
  • Anonymize and respect individuals. Verbatim quotes, comments, photos, names: labels are sufficient. Segmenting a newsletter is possible; individually profiling readers requires a legal basis and the approval of the data protection officer. One may not generate or modify the face or voice of a real person without their written consent (right to one's image, Article 9 of the Civil Code; a montage disseminated without consent is punishable under Article 226-8 of the Penal Code if it is not clearly indicated), and one may not imitate a brand, a work, or the style of a living artist.
  • AI proposes, the organization expresses. No text, image, or AI-generated response is published without review and validation by the person responsible for it.
  • Report what needs to be reported and keep track. Since August 2, 2026, Article 50 of the European AI Regulation requires the labeling of generated images, sounds, and videos that appear authentic, and generated texts published to inform the public on topics of general interest without editorial review. In commercial influence, a face or silhouette produced by AI requires the label "Virtual Images," and such a label never legitimizes misleading content. The toolkit provides a transparency register; Article 4 of the same regulation, rewritten by Regulation (EU) 2026/1744, also requires organizations to take measures to support the control of AI by those who use it on their behalf, and the toolkit contributes to this.

7. What the kit contains

Version 2, updated October 26, 2026: three new fact sheets focusing on a ninth scenario, "Communicating in the Age of AI" (what AI assistants say about the organization, internal communication on AI deployment, responding to management manipulation), updated transparency rules (Article 50, code of best practices, ARPP fact sheet, mention of "Virtual Images"), and a new section on copyright. The kit is a 48-page Word document, with the sections to be replaced highlighted in yellow in each prompt. It opens with a user guide: why this kit, the three levels and the six-week learning path, the twelve-question self-assessment, the prompt template with four ingredients (objective, context, sources, expectations), the verification reflex and five prompt reflexes for 2026, the five security rules, an update on copyright and AI in October 2026, and the key for the fact sheets. Next come the three levels, nine sheets each, each level beginning with what you will be able to do and ending with a validation challenge and its passing criteria. The fifth part is aimed at the communications team and management: the Monday AI quarter-hour, the shared prompt library, the eight rules of use, the five indicators for measuring gains, and the role of management. Six appendices complete the kit: the use case matrix and the "which tool for which" table, the self-assessment grid, the pre-publication checklist, the progress log, the glossary, and the sources.

The accompanying Excel workbook contains eleven tabs, including a readme: the tone charter and the house glossary, the editorial calendar with its calculated delays and the counter of generated content to be reported, the derivation matrix, the communication of change by population with its calculated priority, the question and answer bank with its review dates and its validation alert, the transparency register of generated content which calculates the required mention and its location (preparatory use, "Virtual Images", hyper-falsification), the crisis plan with the criticality of each scenario and a hyper-falsification scenario, the new "Presence in AI" tab (citation and accuracy rate), the indicators and the progress log.

Download the AI ​​communicator's kit (Word + Excel, free)

48 pages, 27 one-page fact sheets, three levels, six appendices, and an eleven-tab workbook. Written by NextStart.AI based on Article 50 of the European AI Regulation, the Commission's Code of Practice, the ARPP's practical guide, Anthropic's AI Fluency framework, and workshops conducted with communication teams from large organizations.

  • Self-positioning, prompt template, and five communication-specific security rules
  • 27 ready-to-copy prompts, from tone guidelines to responses to deepfakes, in three levels
  • A validation challenge by level and a six-week course
  • Excel spreadsheet: editorial calendar, variations, Q&A, transparency register, crisis plan, presence in AI

    Related trainingAI training for communications professionals →Internal communication, content, key messages, multi-channel and multilingual adaptations: your teams practice on their own topics.1 hour 30 minutes or half a day€1,300 to €1,800 excluding VATEligible for OPCO and ESF+ funding

    8. Frequently Asked Questions

    Who is this kit for?

    For internal and external communications teams, press relations, employer branding, and speechwriters in large organizations, regardless of their starting level: the twelve-question self-assessment indicates where to begin. Communications departments can distribute it as is to their counterparts in various departments and subsidiaries.

    Is training required before using the kit?

    No. The kit is designed for self-sufficiency: each sheet explains what to prepare, provides the prompt to copy, and lists the points to check. Training on your own content accelerates skills development and addresses the specific needs of your organization, but it is not a prerequisite.

    What tools does it use?

    With Copilot Chat and Microsoft 365 Copilot (often called "Copilot Premium"), ChatGPT, Claude, Gemini, and Vibe from Mistral (formerly Le Chat). The prompts are neutral; each card indicates useful alternatives when the tool changes the situation, for example, Teams transcription for an interview, Copilot in Word for rewriting, voice mode for rehearsing a speech, or scheduled agents for monitoring.

    Should we mention that content was produced using AI?

    Since August 2, 2026, Article 50 of the European AI Regulation requires the labeling of generated images, sounds, and videos that appear authentic, and generated texts published to inform the public on matters of general interest when they have not been reviewed under the editorial responsibility of a person; preparatory uses (ideas, scripts, storyboards) do not require labeling. The toolkit provides a transparency register that calculates the required labeling and its placement; your labeling rules remain determined by your legal department.

    Can AI be used for crisis communication?

    To prepare for it, yes: scenarios, statements of expectation, Q&As, simulated exercises, and responses to deepfakes targeting a leader are covered in three fact sheets. In a real crisis, only authorized tools are used, no non-public information is shared on public platforms, and every word is validated by the crisis management team and the legal department.

    How long does the journey take?

    Six weeks, 20 to 45 minutes per day, two weeks per level, with a validation challenge at the end of each level. The kit can also be used on demand: the matrix in Appendix A allows you to directly find the sheet that corresponds to your current situation.

    9. Going further

    The kit allows you to progress independently; it does not replace a workshop on your own content. AI training for communications professionals NextStart.AI's 1.5-hour or half-day training program gets your teams working on their press releases, internal announcements, and real-life public speaking engagements, with 70% hands-on practice. They leave with the prompts and agents they created during the session. The Level 3 worksheets also link to the resources of the library app.nextstart.ai, including the guardian of key messages and the media and fallout watch.

    Read also : AI toolkit and social dialogue of the AI ​​project sponsor, Manager's toolkit for dealing with AI: getting your team on board, AI Ambassador Network: The 2026 toolkit for building your collective, Sponsor's AI Act Kit et AI Marketing Kit.

    The entire series of free kits. To drive adoption: AI charter template, AI training plan template, AI Ambassadors Kit, Manager's toolkit for dealing with AI, AI-powered time-saving kit, Soft Skills Kit for AI, Adoption Roadmap Kit, AI maturity grid, AI Measurement and ROI Kit, AI Governance Kit, Cowork kit, 50 Copilot M365 prompts, Copilot Studio kit, Deploy Microsoft 365 Copilot kit, Copilot Chat Kit, Copilot Excel kit, Copilot Teams and Outlook kit, Copilot Word and PowerPoint kit, Configure Claude kit, Deploying Claude in the Workplace Kit, Claude Agents Kit, Claude Skills kit, Claude's Marketing Skills, Skills Claude HR, Claude Plugins Kit, Claude Code kit, Claude Design kit et Vibe coding kitFor the AI ​​project sponsor: AI Act kit, GDPR kit, Security and data kit, Shadow AI kit, AI and Social Dialogue Kit, frugal AI kit, AI Sovereignty Kit et responsible AI kitBy profession: AI HR kit, AI marketing kit, AI finance kit, AI kit for lawyers, AI kit for salespeople, AI toolkit for project managers, AI innovator kitAI kit for communicators, AI Procurement Kit, AI Kit for Researchers et AI kit Customer service.

    10. Sources

    The figures and references cited on this page and in the kit are based on the following sources, verified on October 26, 2026.

    Back to top
    All NextStart.AI offersEight consulting missions, seven pillars, around twenty formats, eighteen professions, five tools
    Around thirty major organizations were supported.Large groups, public agencies and networks: from acculturation to production

    Insurance and social protection

    Public sector and research

    Consumption and retail

    The NextStart.AI collectiveConsultants, trainers and industry experts: more than 5 350 trained professionals since 2023
    All free resourcesDiagnostics, Finders, prompt library, kits by profession, by tool, and for managing an AI program
    The NextStart.AI blogGuides, prices and comparisons of Copilot, Claude, ChatGPT, Gemini and Mistral, and field analyses
    Let's talk about your AI projectResponse within 24 business hours
    AI agents in businessUnderstanding, choosing the tool, training the teams, deploying
    Delegate entire tasksClaude Cowork, Copilot Cowork, ChatGPT Work, Vibe Work