Free resource · Word Kit
AI Marketing Kit: 30 everyday use cases and their solutions, in 3 levels
The step-by-step path for your marketing teams to develop their skills in generative AI independently: ten sheets to start, twelve to produce business-quality studies and campaigns, eight to industrialize with assistants and agents.
The AI Marketing Kit, in brief
- What it is
- A generative AI upskilling kit for marketing teams: 30 everyday use cases, each with its ready-to-copy prompt, organized into three progressive levels.
- For who
- Marketing managers, product managers, research managers, content managers, campaign and communication managers in large organizations, and managers who want to develop their team without waiting for training.
- Format
- 46-page Word document: 30 one-page sheets, a placement test, a level passage test and a progress dashboard.
- What it contains
- The four steps of good prompting, the color code of confidentiality in large organizations, 10 Start sheets, 12 Deepen sheets (studies and campaigns), 8 Industrialize sheets (brand foundation, assistants, prompt chains, agents), and on each sheet the points to check and an extract of results.
- With what tools
- Microsoft 365 Copilot, ChatGPT, Claude, Gemini and Vibe de Mistral (ex Le Chat): the prompts are written to work in each one.
- What is it based on?
- Our training with marketing teams of large French organizations, the AI Fluency framework from Anthropic and the 2026 surveys on the uses of AI in marketing (Jasper, Typeform, McKinsey).
- Price
- Free, in exchange for a professional email address. No unsolicited emails.
- Update
- September 2026.
- What it doesn't replace
- Neither your organization's AI usage charter nor human review: each sheet tells you what to check before using a result.
- Go Further
- AI training for marketing teams, AI charter template, Measurement and ROI kit, AI Ambassadors Kit.
Nine out of ten marketing teams are already using generative AI. Yet, according to Jasper's 2026 survey of 1,400 professionals, 61% of marketing directors are confident in the return on investment, compared to only 12% of contributors who use it daily. The gap doesn't stem from the tools themselves; it stems from the methodology, and the lack of time for anyone to learn it.
This kit offers a three-level self-training program based on the real-world tasks of a marketing team. Each sheet starts with a typical everyday situation, provides a tested quick fix, outlines what to check, and demonstrates the results. This page presents the method and a complete sheet; the Word document contains all thirty sheets.
1. A progression rather than a list: what surveys show
There's no shortage of marketing prompt libraries: HubSpot offers over a thousand, categorized by function. They all share the same flaw: they provide prompts without indicating the learning order or how to evaluate the results. A beginner might pick prompts that are too ambitious, an experienced user might find what they already know, and no one truly progresses.
A tiered approach solves this problem. First, you learn to delegate a simple task and achieve a usable result the first time; then, to provide context, iterate, and critique; and finally, to leverage this knowledge in assistants that the entire team can share. This is also what organizations that truly derive value from AI do: according to McKinsey, 73% of the highest-performing organizations have redesigned their ways of working instead of simply adding a new tool to their existing processes.
| What the 2026 surveys say | What the kit actually does |
|---|---|
| 91% of marketing teams use AI; the most widespread uses are idea generation (56%), multi-channel campaigns (51%), personalization (50%) and search optimization (43%) (Jasper). | Level 1 starts from these already established uses and makes them reliable: summaries, angles, variations, variants, localization. |
| 79% of marketers use AI to write, and 91% edit the generated content (Typeform). | Each sheet contains the points to check before using the result, and an extract of what it gives to calibrate the expectation. |
| Legal, compliance and trademark obstacles have increased 3,4 times in one year (Jasper). | A color-coded confidentiality system opens the kit, and level 3 installs a compliance review agent. |
| Marketing and the sales are the first function cited for AI-related revenue gains; 40% of large organizations deploy agents at scale (McKinsey). | Level 3 moves from individual prompts to shared assistants, prompt chains, and scheduled tasks. |
2. The three levels
Each level corresponds to a way of working with AI and a type of prompt. A five-question placement test at the beginning of the kit indicates where to start; a progress test at the end of each level indicates when to move on to the next.
| Level | What you learn | Form of the prompt | Outline |
|---|---|---|---|
| 1. Getting Started: Your Everyday Assistant Weeks 1 and 2 | Delegate a simple task and obtain a usable result the first time, on tasks you already do. | Four lines: the role, the task, the subject matter, the format | 10 cards, 10 to 25 minutes each |
| 2. Deepening: studies and campaigns on professional quality Weeks 3 to 5 | Produce deliverables suitable for committee presentation: provide context, attach documents, iterate, critique the result. | Six blocks: role, context, objective, subject matter, constraints, format, plus two prompts in succession | 12 cards, 30 to 45 minutes each |
| 3. Industrialize: assistants, workflows and agents Week 6 and beyond | No more starting from scratch: brand foundation file, shared assistants, prompt chains with checkpoints, scheduled tasks. | Instructions for assistant or agent, and operating procedure | 8 fact sheets, 1 to 2 hours of setup time |
Recommended path: one sheet per working day, for a total of six weeks. Level 1 is completed in two weeks and generates gains from the very first sheet. The pivot between levels 2 and 3 is the brand foundation file: one or two pages on voice, target audience, promises, and no-gos, which is pasted at the top of prompts and then uploaded to assistants.
3. The six use case families and the kit map
The fact sheets cover the two roles within the marketing team, research and campaigns, in six categories. Each category spans all three levels: you can therefore also follow only the column for your specific role.
The details of each case follow: the situation, what you get, and the point to check before using the result. The complete prompt, or instructions from the assistant or agent, are in the kit.
Level 1 · Getting Started: The Everyday Assistant
1.1 Summarize a long study or report into a one-page decision
Market and competition, 20 min. Savings: two hours of reading reduced to twenty minutes of rereading. You receive a 60-page study (from an institute, firm, or observatory) the day before a team meeting. You have to extract what matters to your market without reading it in its entirety.
What you get: One page: five lessons, three key figures, three implications for your organization, two limitations of the study. To check : The cited pages exist and clearly state what the summary makes them say.
1.2 Extracting themes and pain points from customer feedback
Segmentation, 20 min. Savings: half a day of analysis reduced to twenty minutes. You have 50 online reviews, 80 open survey responses, or a week's worth of customer service tickets. Nobody has time to read them all, and everyone has an opinion on what customers think.
What you get: A table of five to eight themes with the number of verbatim comments, two representative quotes and a level of irritation, then the three priority irritants. To check : The quotes exist word for word in the verbatim transcripts (AI sometimes rephrases). The full details are reproduced below.
1.3 Create a quick competitor profile using public content
Market and competition, 25 min. Gain: a coherent sheet in twenty-five minutes instead of a morning of copy-pasting. A competitor has just announced an offer, or your management has asked you for a "quick update" on them by tomorrow. You have their offer pages, public pricing, press releases, and job postings.
What you get: A fact sheet in eight sections: promise, targets, visible offers and prices, evidence, tone, new features, blind spots, questions to explore. To check : Each price and each figure mentioned is clearly shown in the glued pages.
1.4 Read a results table and draw three conclusions
Analysis, 15 min. Gain: a first level of analysis in fifteen minutes, before involving the data team. Your email marketing tool, advertising platform, or website is giving you a table of figures. You need to say something intelligent about it in a meeting in an hour.
What you get: Three quantified observations, two explanatory hypotheses, two questions to explore further, and one simple action to test. To check : Redo at least one calculation by hand (a rate, a difference).
1.5 Generate twenty content angles on a topic
Content, 15 min. Gain: a reserve of ideas for a quarter, produced in fifteen minutes and sorted into ten. The editorial committee meeting is in two days, and you have to suggest topics on a theme the team has already covered extensively. Writer's block looms.
What you get: Twenty angles classified by reader intention (understand, compare, decide), each with the question it answers and the most suitable format. To check : Do the angles overlap with what you have already published or with what a competitor is doing?
1.6 Drafting a first draft of a newsletter or email from raw data
Content, 15 min. Gain: the first draft in three minutes; your time goes to proofreading and personalization. You have three pieces of information (an event, a new development, a figure) and an email to send before tonight. The first draft is what takes the most time.
What you get: A structured email: two subject line options, a hook, two short paragraphs, a single call to action, a signature. To check : Dates, names, places and links: everything comes from your information, nothing has been "filled in".
1.7 Adapting long-form content into social media posts
Content, 20 min. Reward: a series of posts for two weeks in twenty minutes. An article, study, or webinar has just been published. The goal is to create a series of LinkedIn posts that aren't simply five identical summaries.
What you get: Five publications with distinct angles, ready to be reread, with a call to read the full content. To check : Every figure and every quote is included in the source content.
1.8 Adapt the tone and language level of a text to the guidelines
Campaigns, 10 min. Gain: a rewrite in ten minutes, without sending the text back to its author. A text arrives from an expert, a subsidiary or an agency: correct in substance, but in a tone that is not that of the brand, with jargon and sentences of forty words.
What you get: The text rewritten in the expected tone, at the same length, with a list of changes. To check : No information has disappeared, none has been added: compare the two versions.
1.9 Produce ten variations of email subject lines, titles, or taglines
Campaigns, 10 min. Gain: an A/B test to choose in ten minutes rather than an hour of brainstorming with three people. The message is ready, but the email subject line, article title, or banner headline isn't displaying. And you need to provide two versions for A/B testing.
What you get: Ten variants in different registers, with their length and their weak point, in a table. To check : Recount the length of the selected variant: the AI is mistaken about the characters.
1.10 Translating and localizing content for another market
Campaigns, 20 min. Benefit: a solid working basis for the local proofreader, who no longer has to start from scratch. The German subsidiary or the British market expects the local version of your campaign. A word-for-word translation is not enough: units, dates, references, and calls to action must be changed.
What you get: A version adapted to the market, with a list of adaptations made and points to be validated by a local colleague. To check : A native speaker reviews before publication: the translation is good, the local nuance not guaranteed.
Level 2 · In-depth study: professional quality studies and campaigns
2.1 Building data-driven personas
Segmentation, 45 min. Gain: a day of workshop replaced by forty-five minutes of work and one hour of validation. Your personas are three years old and look like casting scripts. Yet you have valuable data: interview transcripts, verbatim comments, reviews, and a CRM export. You need personas that the sales team recognizes.
What you get: Three to four personas with needs, irritants, triggers, objections and sources of information, each trait marked proven, probable or hypothesis, plus a comparative table. To check : Each "proven" trait can be found in at least two documents of the subject matter.
2.2 Mapping a customer journey
Segmentation, 40 min. Gain: a route map ready for a one-hour workshop, instead of a week of preparation. Everyone talks about the "customer journey," but no one has actually mapped it out since the website redesign. You have tickets, analytics, interviews, and a validated persona.
What you get: A journey in six steps at most, with what the persona does, wonders and feels, the points of contact, the irritants observed and the moments of truth, then three opportunities ranked by impact and effort. To check : Each "observed" irritant is linked to a quote or figure from the material.
2.3 Conduct a structured competitive analysis
Market and competition, 45 min. Gain: a defensible analysis in committee, produced in forty-five minutes from existing fact sheets. You have three to five competitor profiles (profile 1.3) and one profile of your own offering. Management wants to know "where we stand," summarized in one page and a diagram.
What you get: A comparison grid based on sourced criteria, a positioning map on two justified axes, three free spaces and three threats. To check : Each completed cell links to an existing record and section.
2.4 Analyze the results of a survey
Analysis, 45 min. Gain: a first level of analysis in one hour, which guides the order given to the data team. Your survey collected 300 responses in an Excel file. The research institute or data team won't deliver their analysis for another three weeks. You want reliable initial insights, not misleading averages.
What you get: Five insights with their figures and numbers, flat and cross-tabulations, limitations, and three questions for the next wave. To check : Recalculate a flat sort in your spreadsheet: the totals should match.
2.5 Synthesize market trends from multiple sources
Market and competition, 40 min. Gain: a sourced trend report in forty minutes, reusable in presentations throughout the year. You have compiled four to six studies, articles, and reports on your market. The marketing committee wants a trend report that isn't just a compilation of press releases from publishers.
What you get: At most five trends, each supported by at least two cited sources, with contradictions, a level of confidence and what this changes for you, then what the sources do not say. To check : Open at least one source per trend and check that the figure is there, on the indicated page.
2.6 Questioning a study: the questions it does not ask
Analysis, 30 min. Benefit: a critical reading in thirty minutes, which avoids a decision based on a sample of 40 people. An institute, agency, or department presents you with a study to support a decision. The figures look impressive. Before committing, you want to know its validity.
What you get: What is solid, what is fragile (with the pages), five questions to ask the author and a recommendation: decide, complete or redo. To check : Each weakness mentioned corresponds well to what the indicated page says.
2.7 Write a complete agency brief or creative brief
Campaigns, 40 min. Gain: a complete brief in forty minutes, and one less back-and-forth with the agency. The agency has been waiting for the brief for a week. You have the objectives in mind, a budget, a timeline and personas, but everything is scattered across emails and meetings.
What you get: A twelve-section brief ready to send, with missing information marked rather than invented, and a list of open questions. To check : No [TO BE COMPLETED] remains in the brief sent.
2.8 Designing a campaign concept
Campaigns, 45 min. Benefit: three well-reasoned ideas to discuss in one hour, instead of a brainstorming session with no substance. The brief is approved. Before receiving proposals from the agency, or for an in-house campaign, you want a messaging platform and creative territories to discuss as a team.
What you get: A messaging platform (insight, promise, proof, tone, signature) and three truly different creative territories, each with its variations and risks, then a recommendation. To check : The promise is achievable with the product as it exists today.
2.9 Adapting a message by segment and by channel
Campaigns, 40 min. Gain: twelve coherent texts in forty minutes, ready for proofreading. The concept has been validated. Now it needs to be adapted for three segments and four channels, without distorting the promise and without writing the same thing twelve times.
What you get: A segments × channels matrix with one text per cell, respecting the lengths of each channel, and a "why it's different" line for each cell. To check : The promise is identical in all 12 texts: read them again in succession.
2.10 Optimize content for traditional search and answer engines
Content, 40 min. Benefit: a complete SEO and GEO brief in forty minutes, without an agency. An important page or article is no longer appearing in search results and is never cited by the AI assistants your clients use. You want an optimization brief, not a blind rewrite.
What you get: A brief in six sections: structure, quoteable questions and answers, terms and entities to add, data to source, internal linking, rewriting of the introduction. To check : Each [SOURCE TO FIND] has been replaced by a real source or removed.
2.11 Building a launch plan
Strategy, 45 min. Gain: a structured launch plan in forty-five minutes, to be adjusted as a team rather than built as a team. An offer is launching in ten weeks. You have a product sheet, buyer personas, a budget, and a small team. You need a phased plan that everyone can follow.
What you get: A four-phase plan with dates relating to D-Day, messages, channels, deliverables, indicators, five risks and their countermeasures, and a reverse schedule. To check : Each deliverable has a real person responsible and a deadline compatible with its workload.
2.12 Transforming a data export into an annotated report
Analysis, 40 min. Gain: a structured report in forty minutes, with hypotheses to test instead of improvised explanations in a meeting. The campaign or quarter is over. You have the exports from your tools and a meeting with the marketing department. You need a report that explains, not a table that apologizes.
What you get: A report in four parts: facts, explanations with verifiable hypotheses, three recommendations with effort and impact, next actions. To check : Redo two key calculations (conversion rate, cost per result) in your spreadsheet.
Level 3 · Industrialize: assistants, workflows and agents
3.1 Write the brand foundation file
Strategy, 45 min + validation. Benefit: written context once for the entire team, and results that finally look the same. Each team member provides the AI with the same context at each prompt: who we are, who we are talking to, how we are talking. With variations, therefore, the results are not identical.
What you get: A one-to-two-page document (voice, targets, promises, prohibitions, examples) that all level 2 prompts reuse at the beginning, and that all level 3 assistants load with knowledge. To check : Every statement in the file is true today, but not in the strategic plan.
3.2 Create the "home editor" assistant
Content, 1 hour + test. Benefit: consistent first drafts for the whole team, and revisions that focus on substance. Five people write for the brand, each with their own style, and each reconfigures the AI in their own way. The revisions always correct the same things.
What you get: A shared assistant to the team who writes in the voice of the brand, asks the right questions before writing and signals what he is not allowed to state. To check : The assistant does ask the three questions when the request is incomplete, and only those three.
3.3 Create the "study analyst" assistant
Analysis, 1 hour + test. Gain: the team's study memory can be queried again in one minute. The team has amassed dozens of studies, barometers, and reports in a shared folder. No one knows what they contain anymore, and each new question triggers a two-hour search.
What you get: An assistant who answers the team's questions by citing the study and the page, who says when he cannot find it, and who does not mix up the years or the scopes. To check : Of the 10 test questions, each cited page does indeed contain the number.
3.4 Linking a field study in five steps
Campaigns, 2 hours, then 1 hour. Gain: from a raw study to a campaign developed in one morning, with human decisions in the right places. You know how to do each step separately: synthesis, insights, personas, concept, variations. But between each step, the context gets lost, and so does the coherence.
What you get: A chain of five prompts that follow one another in the same conversation or project, with a human checkpoint between each step, and a coherent final document. To check : Each step was validated by you, not automatically linked together.
3.5 Automate weekly competitive intelligence
Market and competition, 1 hour, then 10 minutes/week. Benefit: regular monitoring without spending more than ten minutes per week. Competitive intelligence is gathered in fits and starts, when someone stumbles upon something new. The competitor profiles (1.3) are not updated, and the analysis (2.3) dates back to the last seminar.
What you get: A monitoring note every Monday morning, produced by in-depth research or a planned task, which you reread in ten minutes and whose new facts feed into your notes. To check : Each fact noted has a link that opens and explains what the note makes it say.
3.6 Implement the compliance review agent
Campaigns, 1 hour 30 minutes. Benefit: human reviews that focus on real difficult cases, and content that arrives clean. Legal, compliance, and branding constraints are the primary obstacle to content production in large organizations. Every piece of content requires review, and reviewers consistently correct the same errors.
What you get: An agent who reviews all content before human review and provides a structured opinion: blocking, to be corrected, acceptable, with the rule in question each time. To check : Of the 10 test pieces of content, the agent identified the points that the human reviewer had corrected.
3.7 Produce all campaign assets from a brief
Campaigns, 1 hour 30 minutes. Gain: a complete campaign in the first draft in an hour and a half, instead of a week of scattered production. The brief and concept have been approved. Now we need to produce the email, the page, the posts, the video script and the banner texts, all consistent with each other, and quickly.
What you get: A single document with all campaign assets, produced by the in-house copywriting assistant from the brief, with a consistency table and points to check. To check : The consistency chart shows a single promise and a single main call to action.
3.8 Automate the monthly report
Analysis, 1 hour 30 minutes, then 20 minutes/month. Savings: the monthly half-day becomes twenty minutes of analysis. At the beginning of each month, the same half-day is spent on exporting, copying and pasting, and formatting the report. The analysis itself takes place in the last ten minutes.
What you get: A commented report is produced each month using the same mechanism based on your exports, which you review, correct and enrich in twenty minutes. To check : Two calculations redone by hand each month, different from one month to the next.
4. A comprehensive guide: extracting irritants from customer feedback
Here is worksheet 1.2, as it appears in the kit. It's the quickest to try: take 30 to 100 customer reviews or open-ended survey responses, remove the names, emails and contract numbers, and paste them below the prompt.
The situation
You have 50 online reviews, 80 open survey responses, or a week's worth of customer service tickets. Nobody has time to read them all, and everyone has an opinion on what customers think.
What you get
A table of five to eight themes with the number of verbatim comments, two representative quotes, and a level of irritation, followed by the three priority irritants. Half a day of analysis reduced to twenty minutes.
The copy prompt
You are a marketing research analyst. Here are [number] customer comments from [source: online reviews, surveys, customer service] about [product or service]. Group them into 5 to 8 themes. For each theme: a title, the number of comments, two representative quotes, and the level of irritation (low, medium, high). Conclude with the 3 most important pain points and a question I should investigate further. Format: a table, followed by a five-line summary. Don't make anything up: if a theme is based on two comments, state so. [paste the comments]
Check before using the result
- The quotes exist word for word in the verbatim transcripts (the AI sometimes rephrases them).
- Staff numbers by theme: recount at least one theme
- The level of irritation reflects the customers' words, not the supposed seriousness of the issue.
The rest of the journey
At level 2, worksheet 2.1 transforms these themes into data-driven personas, with each trait marked as proven, probable, or a hypothesis. At level 3, worksheet 3.3 creates a research analyst assistant who responds to the team by citing the study and page, and who states when they don't know.
5. Safeguards: four actions and a color code
The kit opens with the four gestures of the good prompt, taken from Anthropic's AI Fluency framework (delegation, description, discernment, diligence) and then applied to each card.
- Assign the right task. AI excels at repetitive tasks where the data is readily available and the outcome is verifiable. It struggles with tasks requiring information it lacks. Before submitting a prompt, ask yourself: Do I have the necessary data? Will I be able to accurately assess the result?
- Describe completely. A good prompt states the role (who is speaking), the context (for whom, in what setting), the objective (what is being achieved), the material (what is being pasted or attached), the constraints (length, tone, prohibitions, what should not be invented), and the format (table, list, document). Level 1 uses four of these, Level 2 uses all six.
- Judge the result. Read like a proofreader, not a reader: look for errors. Figures, names, dates, quotes, and sources are the first things to check. Ask the AI what is proven, probable, or a hypothesis; ask it what it hasn't found.
- Take responsibility for what you publish. You sign off on the result, not the AI. You verify what binds the brand, you respect the confidentiality and disclosure rules in force in your organization, and you state that the AI was used when asked.
He then establishes a simple color code for what can be pasted into an AI assistant, and where. In case of doubt, your organization's usage guidelines and your IT department will decide; our AI charter template It provides a basis if you don't have one.
| Areas | What it covers |
|---|---|
| Green: always possible | Public data, previously published content, non-sensitive internal drafts, purchased studies whose license allows internal use, anonymized verbatim transcripts. |
| Orange: only within the authorized company tool | Non-public internal data: campaign results, plans, briefs, internal studies, anonymized exports. Enterprise offerings (Copilot in Microsoft 365, ChatGPT Enterprise or via Azure, Claude Team or Enterprise, Gemini in Workspace, Mistral in dedicated deployment) contractually commit not to train their models on your data; consumer versions do not. |
| Red: never without a validated frame | Personal data of clients or employees, individual health or financial data, trade secrets, documents under confidentiality agreements, passwords and access keys. Anonymize beforehand (names, emails, contract numbers), or do not use them. |
6. What the kit contains
The kit is a 46-page Word document, editable and printable. It contains, in order: the user manual and the map of levels and families; the placement test; the four gestures of good prompting; the color code for confidentiality and the rules applied by the sheets; the ten sheets of level 1 and their pass test; the twelve sheets of level 2 and their pass test; the eight sheets of level 3, with the table of tools where assistants and agents are built, and their pass test; the printable progress dashboard; and the sources.
Download the AI Marketing Kit (Word, free)
46 pages, 30 one-page fact sheets, one tested prompt per fact sheet. Written by NextStart.AI based on training sessions we conduct for marketing teams in large French organizations.
- Placement test and six-week course
- 30 ready-to-copy prompts for Copilot, ChatGPT, Claude, Gemini and Vibe (formerly Le Chat)
- Points to check and extract of results on each sheet
- Passage tests and progress dashboard
7. Frequently Asked Questions
Is a specific AI tool required to use this kit?
No. Prompts work as is in Microsoft 365 Copilot, ChatGPT, Claude, Gemini, and Mistral's Vibe (formerly Le Chat). Use the tool your organization has authorized. Only the Level 3 documentation indicates where assistants and agents are built in each tool, as this is where the tools differ.
Where to start ?
The placement test consists of five questions and takes two minutes. From 0 to 3 points, start with sheet 1.1; from 4 to 7 points, start with sheet 2.1 after three sheets of level 1 in accelerated mode; from 8 to 10 points, go directly to the brand's foundation file (sheet 3.1).
Can customer data be pasted into an AI assistant?
Not just any data, nor anywhere. Public data and anonymized transcripts can go in any tool; internal, non-public data only in the authorized company tool; personally identifiable data never without a framework validated by your organization. Each sheet indicates what must be anonymized before pasting.
How long does it take to complete the route?
Six weeks, at a rate of one worksheet per working day, with each worksheet taking between 10 and 45 minutes. Level 1 is completed in two weeks. Level 3 worksheets require one to two hours of preparation, then a few minutes per week or month.
Is the kit suitable for an entire team?
Yes, it's designed to be shared within your organization. The process is best completed in pairs, with a colleague reviewing your progress tests. The Level 3 worksheets generate a brand foundation file and assistants that can be used by the entire team.
How does this differ from training?
The kit allows you to progress independently, at your own pace. Training saves time on levels 2 and 3: you work directly on the team's studies, briefs and campaigns, with a trainer who corrects prompts and establishes common usage rules.
8. Going further
To complete this journey as a team, using your own studies and campaigns, each level of the kit has its own format: a Prompt Party one hour for level 1, our AI training for marketing teams in half a day for level 2, and the workshops Agent Party et Cowork Party to build level 3 assistants and agents.
Read also : AI Charter in Business: The 2026 Model (Downloadable), AI Measurement and ROI Kit, AI Ambassador Network: The 2026 Toolkit et AI adoption roadmap in 12 months.
The entire series of free kits. To drive adoption: AI charter template, AI Ambassadors Kit, Manager's toolkit for dealing with AI, Adoption Roadmap Kit, AI maturity grid, AI Measurement and ROI Kit et AI Governance KitFor the AI project sponsor: AI Act kit, GDPR kit, Security and data kit, AI and Social Dialogue Kit, frugal AI kit, AI Sovereignty Kit et responsible AI kitBy profession: AI HR kit, AI finance kit, AI kit for lawyers, AI kit for salespeople, AI toolkit for project managers, AI innovator kit et communicator's AI kit.
9. sources
The figures and frameworks cited on this page and in the kit come from the following references, consulted in September 2026.
- jaspers The State of AI in Marketing 2026, survey of 1,400 marketing professionals, January 2026.
- Typeform, 2026 Trends in Generative AI and the Marketer, 1,191 marketers and 1,065 consumers, 2026.
- McKinsey, The state of AI 2026, 1,719 respondents in 97 countries, May and June 2026.
- Anthropic, Claude Academy, AI Fluency: Framework and foundations, framework of the four skills (delegation, description, discernment, diligence).
- Shann Holmberg, 5 Levels of AI Marketing: From Prompts to Agent Teams, progression of generic prompts to agent teams.
- Microsoft Adoption, Copilot Scenario Library: marketingmarketing scenarios in stages, applications and indicators.
- HubSpot, AI Marketing Toolkit and 1000+ AI-Driven Prompts, a library of prompts categorized by marketing function.
- Jaydai, Free AI prompt library by profession, prompt structure in six blocks.
- European Union, Regulation (EU) 2024/1689 on artificial intelligenceArticle 4, AI control, applicable from February 2, 2025.
- CNIL, The CNIL's Q&A on the use of a generative AI system, personal data and generative AI.