Free resource · Word Kit

AI Kit for HR Teams: 28 Use Cases and Prompts in 4 Levels

The twelve-week program to upskill your HR team on generative AI, independently, one hour per week, with the safeguards of the profession.

📄 Word, 52 pages🧭 4 levels × 7 HR domains🗓️ 12-week program🆓 Free

Download the kitSee all four levels

Four out of five HR professionals already use a chatbot daily, yet most teams are still stuck in the same old routine: a question, an answer, a hastily reviewed copy-paste. Prompt libraries don't change this, because a list of fifty prompts categorized by process doesn't tell you where to start, when you're ready to move on, or what you're not allowed to do with employee data. What HR teams need isn't just another list; it's a progression.

This page explains why a level-based skills progression changes the results, describes the four levels and seven areas of the kit, summarizes the legal framework that each sheet includes, and gives you the complete kit in Word format, with its twelve-week program.

1. A progression rather than a list: what surveys show

The 2026 AI and HR barometer from Parlons RH (Let's Talk About HR) reveals that 80% of French HR professionals use a chatbot daily, primarily to summarize documents (71%) and generate reports (62%), but that 52% cite the lack of a clear strategy as the main obstacle. In the United States, SHRM reports that 39% of organizations use AI in HR, rising to 60% for those with over 5,000 employees, and that 57% of HR professionals are unaware of the regulations governing its use. BCG, in its global survey conducted with the WFPMA, identifies data protection and compliance as the top obstacles for 51% of respondents.

These figures all point to the same thing: the ability to use the technology is there, but mastery isn't. And mastery isn't acquired by reading; it's acquired by practicing in a specific order. Research on adult learning shows that short, spaced-out practice sessions are significantly more effective than a single, intensive day; Copilot's trial in the British civil service, involving 20,000 employees, demonstrated that the time saved increases with familiarity with the tool; and studies on the "serrated edge" of AI remind us that a user who doesn't know how to evaluate a result performs worse with AI than without. Hence the approach taken with the training kit: four levels, each building upon the skills acquired in the previous level, with data verification and protection taught before advanced uses.

We reviewed around twenty HR prompt libraries published in 2025 and 2026, in both French and English, ranging from AIHR to Culture RH, including SHRM, Lattice, PayFit, and Microsoft's Copilot scenario library. All are categorized by HR process or tool; none are organized by skill level, none offer self-assessment or a timeline, very few provide verification criteria, and almost all French kits ignore Copilot, Gemini, Claude, and Mistral, even though these are the tools deployed by large organizations. This kit is designed to address these four shortcomings.

2. The four levels

The four levels follow the grammar of Prompt Party® that NextStart.AI facilitates in large organizations, where each wave of sessions corresponds to a level, and aligns with the reference frameworks published in 2025 and 2026: the four levels of DigComp 3.0 mastery, the four tiers of the Pix AI skills framework, the three modes of the AI ​​Fluency framework (automate, augment, delegate) and the three tiers described by the CIPD for the HR function. Three weeks per level, one hour per week per person.

Level	What the team knows how to do at the exit	Techniques practiced	Tool functions
1. Initial reactions	Obtain reliable and sourced answers, produce and improve your HR writing, apply data rules without thinking about it.	Four-part method: iteration, sourced web research, systematic review.	Professional account, web search, conversation can be deleted.
2. Processing information	Reduce large volumes of documents and leverage aggregated HR data to make faster decisions.	Attachments, examples, step-by-step reasoning, required formats, quality criteria, anonymization in four steps.	Customized instructions, memory, attachments.
3 · Create and progress	Produce comprehensive courses, policies and support materials, train through role-playing, build the team's first assistants.	Project instructions, work in passes, in-depth research, assistant test game.	Persistent projects and spaces, in-depth research, shared assistants.
4. Innovate and automate	Delegate recurring follow-ups to supervised agents, produce formulas and controls, measure gains, govern HR practices.	Agent contract, pre-production simulation, draft submission principle, thresholds and aggregates, register of uses.	Agents, planned tasks, connectors, delegation of tasks.

Each level opens with a page explaining what you will be able to do, the techniques and settings to activate, and demonstrating the same need addressed through both a starting prompt and a more refined prompt. It concludes with a landing section: five self-assessment questions, three corrected vigilance questions, and a challenge to present to the team. Four out of five boxes must be checked to advance to the next level.

3. The seven HR areas and the toolkit map

Each level contains seven sheets, one per HR area. The kit can therefore be read in two directions: by level, it's the program; by area, it's your job, and you see the same case grow from one level to the next, from responding to a leave request from the agreement to the HR Helpdesk assistant and the continuous analysis of requests.

Area	Level 1	Level 2	Level 3	Level 4
Recruitment and integration	Clear and inclusive job offer	Structured and fair interview grid	Integration process within a project	Application and Arrivals Tracking Officer
Employee Services	Responding to a request based on the reference document	FAQ for a device extracted from an agreement	The HR Help Desk Assistant on the team	Continuous analysis of requests, improvement of the assistant
Policies and Procedures	Procedure rewritten in plain language	Comparison of two versions	A comprehensive HR policy	Monitoring deadlines and social compliance
HR data and reporting	Explain and comment on an indicator	Analysis of an aggregated export	Note and support for the management committee	Social report and equality index with controls
Listening and social climate	Invitation to a survey for three audiences	Verbatim quotes categorized by theme	From the result to the action plan and its presentation	ongoing analysis of the social climate
Skills, training and careers	Launch of an interview campaign	Training needs extracted from a campaign	Getting coached: difficult interview, development plan	Talent review: compiling facts, without evaluating
Monitoring, social dialogue and transformation	Sourced research on a text	Monitoring note	Preparing for organizational change	Prepare for a mandatory annual negotiation

Each form fits on half a page and follows the same template: the situation in two lines, what you need and what needs to be anonymized, the copy prompt with its highlighted fields, the follow-up prompt, the three points to check before using the result, and the link to the prompt, assistant, agent, or cowork version of the same form in the library app.nextstart.ai.

The details for each case follow: the situation, what you need, 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: First reflexes

A1 · Write a clear and inclusive job offer

Recruitment and integration, 20 min. Orange light. A manager sends you a three-page job description and expects the announcement tomorrow. You want a clear, attractive, yet not exploitative, job offer that is completely compliant with non-discrimination regulations.

What you need : The job description (text, without the name of the previous incumbent or individual salary). Three contextual elements: the team, the location and the pace, what makes the position interesting. To check : The expected information is there: title, contract, location, missions, process; remuneration if your transparency policy provides for it.

B1 · Respond to a request based on the reference document

Employee services, 15 min. Green light. An employee asks if he can split his paternity leave, postpone days off, or receive a bonus. The answer is in the agreement, but it's difficult to find and sometimes negative, making it tricky to write.

What you need : The question, reformulated without a name or identifying details. The excerpt from the agreement, policy, or memo that deals with the subject (copied into the prompt). To check : The cited article exists and clearly states what the response claims: open the document and reread the passage.

C1 · Rewrite a procedure in plain language

Policies and procedures, 20 min. Green light. The procedure for requesting teleworking, expense reports or internal mobility is four pages long, nobody reads it, and the HR department answers the same questions every week.

What you need : The text of the current procedure. The target audience and their reading channel (intranet, mobile, display). What raises the most questions today, if you know. To check : Compare rule by rule with the original: deadlines, exceptions, authorized persons, special cases.

D1 · Explain and comment on an HR indicator

HR data and reporting, 15 min. Green light. The monthly report shows a turnover rate of 14,2%, compared to 11,8% last year. A manager asks you what that means, and the executive committee expects a comment that says neither too much nor too little.

What you need : The internal definition of the indicator (calculation method, scope, period). The current value and the comparison value, aggregated, without any individual data. The known context: reorganization, seasonality, ongoing recruitment. To check : The figures copied in the comment are exactly yours.

E1 · Write the invitation to a survey for three audiences

Listening and social climate, 15 min. Green light. You are launching an engagement survey or a quarterly barometer. The response rate depends on the first email, and the same message does not speak in the same way to an employee, a manager, and a staff representative.

What you need : The purpose of the survey, the opening and closing dates, the response time. To check : The written guarantees of anonymity are exactly those of the system (threshold of five respondents, access to data): an inaccurate promise will cost you dearly.

F1 · Launch an interview campaign

Skills, training and careers, 15 min. Green light. The annual performance review campaign opens in two weeks. We need a launch message, a version for managers, and ready-made answers to the questions that come up every year.

What you need : The campaign schedule and the tool used. What's changing this year (framework, objectives, new module). Available resources: guide, tutorial, contact person. To check : The dates, the tool name, and the links are correct.

G1 · Sourced research: what changes with a text

Monitoring, social dialogue and transformation, 20 min. Green light. A decree, a law, a CNIL position, or a ruling has just been issued. You need to know in twenty minutes what it changes for your organization, with sources you can cite.

What you need : The exact name of the text or decision and its date. Web search enabled in the tool. Your scope: workforce, sector, collective agreement, public or private. To check : Open each source: the link exists, the date is correct, the text says what is claimed.

Level 2: Processing information

A2 · Construct a structured and fair interview grid

Recruitment and integration, 30 min. Red light. Three managers will interview the same candidates, and each has their favorite questions. You want an identical assessment grid for everyone, based on the skills required for the position, that provides the tools for decision-making without making the decision yourself.

What you need : The job offer or job description (A1) and the four to six truly decisive skills. The interview format (duration, number of interviewers, sequence). To check : Each question has a direct link to the position; questions that are excluded are explicitly excluded.

B2 · Extract the FAQ from a device based on an agreement

Employee services, 25 min. Green light. A new agreement (remote work, working hours, profit-sharing) is coming into effect. Questions are going to pour in, and you want a solid FAQ before the first day, with each answer linked to an article.

What you need : The text of the agreement or policy, as an attachment. The ten questions already received or anticipated, without names. The FAQ format of your intranet. To check : Each reference correctly refers to the article that says this: check all fifteen, that's the heart of the document.

C2 · Compare two versions of a procedure or agreement

Policies and procedures, 25 min. Green light. The legal department sends back version 4 of a draft agreement, or headquarters distributes an updated group procedure. You must state in one page what is changing, for whom, and what needs to be decided.

What you need : The two versions of the text, in two clearly named attached files (V1, V2). The target audiences. Your "what's changing" note template. To check : Open three random gaps in the two files: is the text before and after exactly the one quoted?

D2 · Analyze an export of aggregated indicators

HR data and reporting, 30 min. Orange light. You have a 24-month export of indicators per entity (staff numbers, new hires, departures, absenteeism, training hours). You want the trends, the anomalies, and three findings that you can support.

What you need : The export is aggregated by entity and by month, without any individual rows, with at least five people per group (2.2). The column dictionary includes: name, definition, unit, scope. Known events from the period (reorganization, transfer, change in calculation) are also included. To check : Recalculate two numbers manually in the original file: a column read error occurs.

E2 · Classify the verbatim transcripts of a survey by theme

Listening and social climate, 30 min. Orange light. Four hundred open-ended responses to the question "What would make your job easier?". No one has time to read them all, and yet that's where the early warning signs are.

What you need : The verbatim transcripts are in the file, anonymized according to section 2.2 (names, departments of fewer than five people, and identifying details removed). The question asked, the number of respondents, and the participation rate are also included. To check : Read ten random verbatim quotes and their theme: is the ranking the one you would have made?

F2 · Extract training needs from an interview campaign

Skills, training and careers, 30 min. Orange light. The campaign is closed. The "training preferences" field contains six hundred lines of free text. The skills development plan must be approved within one month.

What you need : Export of only the "training preferences" field, anonymized (code, department, category; no name, no rating). Your skills repository or training catalog. The known strategic priorities for the year. To check : The connections are plausible: check twenty lines at random.

G2 · Write a monitoring report

Monitoring, social dialogue and transformation, 20 min. Green light. You have identified an important text (G1) or received a court decision commented on extensively. The HR director wants a one-page memo, the executive committee five bullet points, the managers three lines.

What you need : The source text or the decision, as an attachment or link, rather than a comment. Your monitoring note template. What you already know about the impact on your organization. To check : The articles and dates are those of the official text, not those of a commentary.

Level 3: Create and progress

A3 · Building an integration pathway within an "Onboarding" project

Recruitment and integration: 45 minutes the first time, 10 minutes thereafter. Orange light. Every new arrival is prepared in a rush, with an Excel checklist that no one ever updates. You want a personalized onboarding process, tailored to each position and location, produced in ten minutes, and consistent with your existing documentation.

What you need : A project or a persistent space in the company tool, with your documents: welcome booklet, organizational chart of functions (without personal data), arrival checklist, calendar of mandatory training, AI charter. To check : The obligations are in place: information and prevention visit, security and cyber training, delivery of mandatory documents, follow-up interview.

B3 · Create the team's HR Helpdesk assistant

Employee services, 1 hour 30 minutes, then 15 minutes per month. Orange light. The same fifty questions come up every month. You want an assistant that the HR team uses first for themselves, that answers based on your documents, cites its source, and refers to a human as soon as the question becomes individual.

What you need : Ten to twenty up-to-date and non-personal reference documents: agreements, policies, FAQs (B2), procedures (C1), manager's guide. To check : No inventions were found in the test game; the five traps are correctly addressed by cross-references. Rerun the test after each document update.

C3 · Write an HR policy from A to Z

Policies and procedures, 1 hour in four passes. Green light. You are being asked to develop a policy on the right to disconnect, the use of generative AI, or internal mobility. You want to start with an assessment of current practices, propose options, and then draft the text, its implementation guide, and its presentation.

What you need : The objective and constraints (existing agreements, collective agreement, culture, schedule). In-depth research activated for the first pass. To check : The sources for pass 1 are open and dated; examples from other organizations are not copied verbatim.

D3 · Prepare the memo and supporting materials for the management committee

HR data and reporting, 40 min. Orange light. The management committee dedicates fifteen minutes to social indicators. You have the findings (D2); you need three key messages, one figure per message, clear visuals, and the decisions to be requested.

What you need : The findings and aggregated indicators (D2), without individual data. The decisions or arbitrations expected from the committee. The usual format of the committee (duration, number of slides, style). To check : Each figure of the support is found in your data; the graphs are produced in your tools, not copied.

E3 · From survey results to action plan and its presentation

Listening and social climate, 45 min. Orange light. The results are analyzed (E2). The hardest part begins: a credible action plan, and three presentations that say the same thing to audiences who do not expect the same thing.

What you need : The summary of themes and irritants (E2), aggregated. Constraints: budget, schedule, what has already been decided or is impossible. Target audiences: employees, managers, staff representatives. To check : Each commitment has a responsible party who has accepted it and an existing budget.

F3 · Getting coached: difficult interview, development plan

Skills, training and careers, 30 min. Red light. You need to conduct a performance review, announce a refusal of a transfer request, or help a manager transform an annual review into a development plan. The AI ​​never meets the person; it trains you and structures your preparation.

What you need : The situation is described by the function and the context, never by the name or by identifying details. To check : The decision and assessment remain human and explainable without reference to AI; no trace of AI in the person's file other than the preparation.

G3 · Preparing for organizational change

Monitoring, social dialogue and transformation, 1 hour. Orange light. A consolidation of services, a new tool for everyone, a relocation. Management has decided on the what, but not yet the how. You need the sequence, the messages, and an FAQ that doesn't lie.

What you need : The nature of the change, the timeline, the target audiences, what has been decided and what has not. The obligations: information and consultation with the Works Council, support, training. To check : The order respects the prior consultation of the CSE; the legal department has validated the sequence and the messages.

Level 4: Innovate and automate

A4 · An agent who tracks applications and prepares for arrivals

Recruitment and integration, 2 hours of setup. Red light. Candidates have been waiting for a response for three weeks, and arrivals are being prepared the day before. You want an agent who monitors the tracking spreadsheet, prepares reminders and checklists, and reports back to you weekly, without ever judging a candidate.

What you need : The tracking table (ATS export or shared file) is reduced to the bare minimum: application ID, position, stage, date of last action, expected arrival date. To check : The agent has no access to any CVs, evaluations, or data beyond the minimum table.

B4 · Continuously analyze requests and improve the assistant

Employee services, 1 hour per month. Orange light. The HR Helpdesk (B3) has been running for three months. The requests received by managers indicate where documentation is lacking, where a procedure is incomprehensible, and where a fundamental problem lies behind a recurring question.

What you need : Monthly export of anonymized requests: subject, category, date, response time, channel; no names, no identifying text. The assistant's log of unanswered questions and current instructions. To check : The export is anonymized and aggregated; no one can be tracked over the months.

C4 · A watchdog for deadlines and social compliance

Policies and procedures, 1 hour 30 minutes setup. Green light. Gender equality index as of March 1st, BDESE (Economic, Social and Environmental Council), professional development interviews and six-year review, mandatory negotiations, updating of the DUERP (Single Document for Risk Assessment), CSE (Works Council) elections, mandatory training: the deadlines are known and yet forgotten. You want a monitoring system that provides alerts 60, 30, and 7 days in advance, and a monthly sourced update.

What you need : The list of your obligations with their frequency, their responsible party and their legal basis, validated by the legal department. To check : The initial schedule is approved by the legal department; the agent reads, advises and proposes, he does not change anything.

D4 · Social report and equality index with controls

HR data and reporting, 2 hours. Orange light. Every year, the same indicators, the same formulas rewritten by hand, and an error discovered after publication. You want AI to write the formulas and controls, on a dummy column structure, and then for you to apply them to your real data in your HRIS or spreadsheet.

What you need : The official definitions of the indicators (decree and guide for calculating the index, framework of the social report or the BDESE). To check : Independent recalculation of at least one indicator by another person, either manually or using last year's tool.

E4 · A continuous analysis of the social climate

Listening and social climate, 2 hours of setup, 30 minutes per month. Orange light. Between annual surveys, indicators exist: short barometers, in-person inquiries, verbatim comments, absenteeism by entity. You want a monthly report that cross-references these aggregated and anonymized sources and asks managers questions rather than delivering verdicts.

What you need : Sources aggregated by entity of at least ten people: pulse results, themes of requests (B4), anonymous verbatim, monthly indicators. To check : The DPO's opinion is written, the information-consultation of the CSE has taken place, the system is described in the processing register and in your charter (L2312-38, L1121-1).

F4 · Talent Review: Compiling facts, without evaluating

Skills, training and careers, 2 hours. Red light. The talent review is approaching, and the applications are diverse. You want the AI ​​to compile, for each key position, the available facts and questions to be investigated, so that the committee can make a decision based on comparable elements. It doesn't rate, rank, or recommend anyone.

What you need : The key positions and criteria decided by management, in writing. To check : The decision is collective, human, fact-based and explained to the person; no AI-produced ranking exists, not even an "indicative" one.

G4 · Prepare for a mandatory annual negotiation

Monitoring, social dialogue and transformation, 3 hours in several stages. Orange light. The annual mandatory negotiations (NAO) begin in a month. We need a detailed assessment, scenarios, a file for each session, and a compelling argument for well-prepared delegations. AI compiles, simulates, and trains you; the strategy and integrity of the negotiation remain human.

What you need : The data aggregated by category and by theme (average and median remuneration, gaps, numbers), without any individual data, in the company tool. To check : The figures are recalculated by the social management control; the scenarios are validated by management before any session.

4. The three lights: the HR legal framework embedded in each sheet

Human resources is the field where generative AI most directly impacts people, and the law clearly states this. Annex III of the AI Act classifies systems used for recruitment and selection, evaluation, promotion, termination, task allocation, and performance monitoring as high risk; the corresponding obligations for deployers will apply from December 2, 2027, following the postponement decided by the Omnibus AI Regulation of July 2026, but Article 4 on AI governance already applies, and Article 22 of the GDPR has long prohibited fully automated decision-making. The CNIL has included recruitment in its 2026 priority controls. The Labor Code requires informing candidates and employees of recruitment and evaluation methods, and the judges reminded in 2025 and 2026 that the CSE must be consulted before any deployment of an AI tool, including when it is authorized by a simple charter: a pilot that equips an entire department, is installed over time or affects working conditions is already a deployment (TJ Nanterre, February 14, 2025), and only a truly voluntary test, limited in time and scope, can precede it (TJ Paris, February 10, 2026).

Rather than a legal chapter that no one rereads, the kit attaches a light to each card, indicating what precautions are necessary:

Feu	What it covers	Conditions
Vert	Internal non-personalized documents, public texts, content that you write.	Standard proofreading, recommended business tool.
Orange	Anonymized or aggregated personal data, confidential documents.	Enterprise tool only, anonymization verified in four steps, no groups of fewer than five people, peer validation.
Red	The high-risk areas of Annex III: recruitment, evaluation, promotion, termination, task allocation, performance monitoring.	AI prepares the plan, a human decides and can explain it. There is never any sorting, rating, or ranking of individuals. Prior notification is given to all individuals, the Works Council is consulted before any deployment, and the Data Protection Officer (DPO) receives their opinion.

The core of the kit includes a list of data that should never be pasted, a four-step anonymization method with a before/after example, a one-page legal guide dated September 27, 2026, and a twelve-question checklist to complete before sharing an assistant or deploying an agent to production. This checklist, along with the dated twelve-week calendar, serves as a record of the AI ​​control measures stipulated in Article 4.

5. The twelve-week program and the Prompt Club

Pacing is more important than volume. The kit offers three weeks per level: two or three solo worksheets per week, twenty to thirty minutes each, always based on a real-life case from the position with anonymized data, and thirty minutes of Prompt Club as a team, within an existing meeting. The Prompt Club consists of three parts: a member presents a case worked on during the week, the team reviews it live, and they decide what goes into the team's prompt library, under what name, with which fire, and to whom.

• Weeks 1 to 3: The base, self-positioning, the seven level 1 cards, the tier. The team displays its ten rules.
• Weeks 4 to 6: personalized instructions, anonymization of a real file together, the seven level 2 sheets, the team's red light usage list.
• Weeks 7 to 9: the first shared project, the HR Helpdesk assistant tested on twenty questions, the role-playing of a difficult interview, the pre-sharing checklist.
• Weeks 10 to 12: the contract of an agent reviewed by the DPO, the register of team usage, the before/after measurement, the plan for the following six months.

The "making it last" section of the kit then describes what sustains the practice after the program: the weekly ritual, the five rules of the team library, six indicators that measure the collective without monitoring individuals, and the three steps that follow, from assistant to agent and then to the delegation of complete missions.

6. What the kit contains

The kit is a 52-page Word document, with fields to be completed highlighted in yellow and "for the person in charge" boxes to be removed before distributing a sheet on its own. It includes the user guide (who it is for, the four levels, the seven areas, the program, the tool equivalency table), the foundation (the three fires, data that should never be pasted, the right in one page, the four-building block method, iterate and verify, the ten team rules), the four levels with their opening page, their seven sheets and their milestone, the "making it last" section, and six appendices: the five-skill self-assessment grid, the twelve-week calendar, the blank prompt sheet and the team library template, the checklist before any new HR use, the glossary, and the sources.

7. Frequently Asked Questions

Do I need to use ChatGPT or Copilot to get started?

No. Level 1 starts with the first prompt and establishes the data rules first. Members who are already comfortable with the system can use the grid in Appendix A and begin at Level 2, provided they complete the Level 1 hurdle, which contains the rules that prevent incidents.

Does the kit work with our internal tool?

Yes. Prompts can be pasted into Microsoft 365 Copilot, ChatGPT, Claude, Gemini, Mistral Vibe, or an internal assistant. An equivalency table indicates, for each level, the functions to activate in each tool (attachments, custom instructions, projects, assistants, agents). The kit recommends the enterprise version of the tool authorized by the IT department.

Can it be used with real employee data?

Never any personal data, health data, or sensitive individual situations. The kit teaches how to anonymize in four steps, to respect a threshold of five people per group, and highlights in red the uses in recruitment, evaluation, and career management where AI prepares and a human makes the decision.

How long does the program take?

One hour per week per person for twelve weeks: two or three worksheets to complete individually and thirty minutes of Prompt Club in teams. If twelve weeks seems long, the kit advises completing levels 1 and 2 on your own, then deciding how to proceed, rather than shortening the levels.

Does the kit constitute training within the meaning of Article 4 of the AI Act?

Since the omnibus regulation (EU) 2026/1744, Article 4 requires employers to take measures to promote the development of AI skills among those who use it on their behalf, taking into account their role: this is an obligation of means, which no longer requires guaranteeing a specific level for everyone. Following the program, dating the schedule in Annex B, and completing the checklist in Annex D constitute such a measure and keep track of it. The toolkit does not replace the advice of your legal department or your Data Protection Officer (DPO).

What's the difference with a Prompt Party® for HR?

The kit allows a team to progress independently, at its own pace, on its cases. A series of sessions, organized by level and facilitated using your tool and your HR scenarios, covers in a few hours what the kit offers over twelve weeks, and addresses questions the team might hesitate to ask on its own. The two are complementary: the kit prepares and extends the sessions.

8. Going further

The kit helps a team move forward on its own; it doesn't replace a facilitated session focused on your specific cases and tools. AI training for HR teams NextStart.AI takes place in 1 hour 30 minutes or half a day, on Copilot, ChatGPT, Claude, Gemini, Mistral or your internal solution, and the Prompt Party® They are deployed in waves, one per level, to develop an entire HR function. For recruiters, the AI training for recruiters deals in detail with the red-light practices of recruitment.

Read also : AI Charter in Business: The 2026 Model (Downloadable), Manager's toolkit for dealing with AI: getting your team on board, AI Ambassador Network: The 2026 toolkit for building your collective, AI systems governance: the 2026 toolkit et Measuring the uses, value and ROI of AI: the 2026 toolkit.

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 marketing 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 references cited on this page are based on the following sources; Appendix F of the kit provides the complete list.

• Let's talk HR, AI and HR Barometer 2026, 2026.
• SHRM The State of AI in HR 2026March 2026.
• BCG and WFPMA, Creating People Advantage 2026March 2026.
• European Commission, AI literacy, questions and answersUpdated July 27, 2026; AI Act, Annex III et 26 article White & Case entry into force of the Omnibus IA regulation, July 2026.
• CNIL, Questions and answers about using a generative AI systemJuly 2024, and Priority Controls 2026, April 3, 2026.
• Labor Code, item L1221-8 et item L1222-3 Village of Justice, Consultation with the Works Council and AI tools: what the judges decided, 2026.
• European Commission (JRC), DigComp 3.0November 2025; Pix, Digital skills in artificial intelligence, 2025; Anthropic, AI Fluency Framework, 2025; ICPD, AI skills planning, April 2026.
• Latimier, Peyre and Ramus, meta-analysis on spaced practice, 2021; British government, Microsoft 365 Copilot Experiment: Findings ReportJune 2025; Dell'Acqua et al., Navigating the Jagged Technological Frontier, 2023.