Free resource · Word Kit
Manager's toolkit for dealing with AI: getting your team on board
What to say, what to expect, what to measure: the frameworks and rituals for bringing generative AI into your team's work.
The manager's toolkit for dealing with AI, in brief
- What it is
- An operational kit for the manager who has to bring generative AI into the work of their team: what to say, what to expect, what to measure.
- For who
- Team managers, and transformation directors, HR directors and training managers who need to equip them.
- Format
- 30-page Word document, 5 parts and 5 appendices, all templates ready to use.
- What it contains
- Self-diagnosis and five uses of the manager with their prompts, kick-off meeting, individual interview, fourteen objections and their answers, expectations by function, protected time, grid of three tasks, team rule on time saved, rituals, agents in the team, workload and social dialogue, dashboard, 90-day plan.
- What is it based on?
- Gallup, Microsoft Work Trend Index and BCG surveys on the role of managers in AI adoption, and the Managers Catalyst programs run in large organizations.
- Price
- Free, in exchange for a professional email address. No unsolicited emails.
- Update
- Version 2, updated on October 26, 2026: verified and dated figures (Gallup 2026, Work Trend Index 2026, BCG AI at Work 2026), new team rule on time saved and work intensification signals, new sections on agents in the team and on workload, the right to make mistakes and social dialogue.
- What it doesn't replace
- There is no collaboration between peer managers, nor a clear decision from management on what is expected.
- Go Further
- Catalyst Managers Program, AI Ambassadors Kit, AI charter template.
Licenses are deployed, a wave of training sessions has taken place, and yet usage has plateaued. In most organizations, it's neither the tool nor the training that's the bottleneck: it's the lack of support at the level where the work is actually done. An employee doesn't ask themselves if AI is powerful; they ask if it's right for them, if they're expected to use it, and what the risks are. Three questions that only their manager can answer. And most managers don't answer: they don't openly resist, they silently withdraw, and the team concludes that the issue isn't a priority.
This page explains what recent surveys show about the manager's role in AI adoption, what actually works in a team, the mistakes that cause the best intentions to fail, and gives you a complete toolkit, in Word format, to take action in 90 days.
1. The manager, the primary driver of adoption: what the figures show
In July 2026, Gallup published its clearest measurement to date, based on US data from the first half of 2026. Among employees whose managers actively support the use of AI, engagement reaches 48%; among others, it's 30%. This eighteen-point difference is greater than the effect of frequency of use (8 points) or the existence of a clear business plan (15 points). Gallup draws three recommendations from this: set explicit usage expectations by role (where to use AI, where not to use it, and how the manager will judge it), train managers as coaches in the AI conversation, and integrate AI into performance reviews instead of addressing it separately.
Microsoft's Work Trend Index 2026 (based on 20,000 AI users in ten countries, including France) supports this finding: organizational factors, including manager support, have twice the impact of individual behaviors on the measured effect of AI. Managers of the most advanced teams are far more likely to create space for experimentation (84% vs. 61%) and to set quality standards for work done with AI (83% vs. 57%).
The problem is that these answers are lacking. In the BCG AI at Work 2026 survey, only a third of non-managerial employees say their management communicates clearly about AI, half of respondents regret the lack of rules for organizing work between humans and AI, and France is among the countries with the lowest regular usage. Most managers aren't openly resisting: they're withdrawing silently. This isn't resistance; it's a silent withdrawal, which is reflected in the indicators as an absence.
2. What the teams are feeling, and how to respond to them
The concerns are measured. In the Work Trend Index 2026, 65% of AI users fear being left behind if they don't adapt quickly. Added to this are the fear of appearing replaceable, the fear of "cheating," and a frequently unspoken fear: having even more work. This explains a phenomenon that all managers encounter: employees who use AI secretly and don't disclose it.
Faced with this, there's one rule: respond honestly rather than offering false reassurances, and don't promise what you can't deliver, such as no job changes or AI automatically reducing workload. Involve teams before decisions are made, not after, and offer concrete skills development opportunities to those who are concerned. And there's a simple, often overlooked practice: explicitly state what won't change (the tasks, the expected quality, the way we communicate), because that's what provides the most reassurance during a period of uncertainty.
| object | Honest answer | To avoid |
|---|---|---|
| "This will eliminate my position." | I cannot guarantee that nothing will change in ten years. Here, today, the goal is to do better what we do, and those who develop their skills are the ones who have the most influence on the future. | "Don't worry, no one will be replaced." |
| "I don't have time." | That's true, and that's why I set aside an hour a week in my schedule. The first few weeks take time; the savings come later, on tasks that come up frequently. | "We need to find the time." |
| "It's not reliable, it's just making things up." | Yes, sometimes, with a lot of confidence. That's why we check everything that comes out. For tasks where errors are costly, we don't use AI alone. | "The new models no longer make mistakes." |
| "They're going to be watching us." | This is exactly what I look at and what I don't look at: I don't read your prompts, I don't count your queries, and the dashboard has no individual metrics. | Say nothing about what is measured. |
| "You don't use it yourself." | You're right to say that. Here's what I did with it this week, and what didn't work. | To justify oneself. |
3. What works in a team: six practices
Show before you ask. A manager's first action isn't a speech; it's a ten-minute demonstration of how they actually use their own work, including its limitations. This makes the topic concrete and the manager credible. A manager who recommends what they don't practice is exposed in a single meeting.
Listen to everyone before setting expectations. A twenty-minute one-on-one interview with each team member to understand their current situation, what's holding them back, and what they need. This is where we discover who's already using AI secretly, who's afraid, and who's waiting for permission. A request made before even listening sounds like an order.
Expectations based on function, in terms of behaviors. Three expected uses per team role, formulated as observable behaviors ("prepares meeting minutes with AI and reviews them"), never as tools or query quotas. Then a place in annual objectives, focusing on knowledge sharing and quality, not quantity.
Protected time. One hour per week, at the same time for the entire team, scheduled by the manager and respected by him, for at least three months. Each person works on the three tasks that take them the most time and repeats them with the AI until they become proficient. The primary reason for abandonment is lack of time; this is the only valid explanation.
A ten-minute ritual. At each team meeting, one or two volunteers present what they've done with AI that week, whether it was a success or a failure, using a split screen. The manager presents first once a month. The ritual works as long as it never becomes a performance review.
To recognize without monitoring. We celebrate shared use cases, errors corrected through verification, and the partnerships between a senior employee who assesses the result and a junior employee who uses the tool. We never measure the number of requests per person, we don't rank employees, and we don't read prompts. The teams verify this from the very first dashboard.
4. The five mistakes that ruin even the best intentions
Recommending without practicing. The manager relays the company's communication without having tried it himself. The team senses this and concludes that the subject is not serious.
Falsely reassuring people about employment. The phrase "No one will be replaced" came back to haunt them six months later, during the first reorganization. A team can forgive uncertainty, but not lies.
Asking without giving a timeframe. "Two hours a month, you'll manage." Without a protected time slot, the employee will choose their work at each arbitration, and they will be right.
Let the time gained disappear. An employee who gains two hours a week without being told what to do with them either loses them to the overall workload or hides them for fear of being given more. Individual gain only becomes a collective benefit if it is acknowledged and reinvested, both by the employee and the manager.
Carry the subject alone. A manager who hasn't negotiated time, structure, and support with their superiors will burn out within six months. The 90-day action plan must include a section on "what I expect from my organization to succeed"; this is the organization's responsibility, not just the manager's.
5. The manager, the AI ambassador, and the charter: who does what
Three systems complement each other without overlapping. The manager He brings about change within his team: he demonstrates, listens, sets expectations, manages time effectively, and maintains established routines.AI ambassador is a volunteer peer who leads practical workshops and answers questions throughout the organization; the manager relies on them for a one-hour workshop or demonstration, but does not replace them. AI usage charter sets the common framework: the validated tool, the prohibited data, the verification, the transparency; the manager must be able to explain it in two minutes and answer "am I allowed to…".
One point deserves to be explicitly stated to the team, because it's often considered but never explicitly stated: what the manager entrusts to AI and what they never do. Preparing for an interview, summarizing a flow of information, drafting a first draft, structuring a decision: yes. Evaluating someone, deciding on a promotion or contract termination, sending a difficult message without rewriting it, analyzing an employee's individual messages: never.
6. What the kit contains
The kit is a 30-page Word document, with fields to be completed highlighted in yellow and advice boxes for the manager, to be removed before sharing a document with the team. It follows the 90-day calendar: the manager's self-diagnosis in twelve questions, the five uses to be installed first with their starting prompts, the one-page manager's sheet, the timed sequence of the launch meeting, the framework for the individual interview, the answers to the fourteen most frequent objections, the message "what does not change", the usage expectations by function, five formulations of annual objective, the rules of protected experimentation time, the grid of three tasks, the team rule on time saved, the team rituals, the support of five profiles, the table of recognition without monitoring, the handling of three incidents, the line between what is entrusted to AI and what is not entrusted, the five decisions of the manager when agents arrive, the rules on workload and the right to error, the team dashboard with nine indicators, the quarterly mini-survey, the 90-day action plan with its section "what I ask of my organization", and five appendices.
The kit, part by part: what each part contains and what it is used for. The templates to fill in are in the Word document.
Clarifying one's own position (week 1)
Manager's self-assessment
Twelve questions that the manager answers alone, in writing, in ten minutes: how many times they used generative AI last week, what they fear for themselves and their team, what the team sees in its use, and what they know about the framework. The reading guides the rest: fewer than three uses per week, start with the five uses for two weeks; nothing seen by the team, your first action is a ten-minute demonstration, not a presentation.
The five uses of management to implement first
Five use cases were chosen because they are used every week, are visible to the team, and do not require any sensitive data: preparing for an interview or feedback session, summarizing a flow of information, drafting team communications, preparing for and following up on a meeting, analyzing a problem, or structuring a decision. For each use case, the toolkit specifies what the AI does and what the manager retains: the assessment of the individual, the decision, and the final tone.
The framework, and the responsibility that remains yours
The manager must be able to explain the framework in two minutes and answer the question, "Am I allowed to…?". The "Framework in One Page" sheet is completed based on the usage guidelines: the authorized tool, what is never included in a prompt, what is always checked, what is said, what remains a human element, and who to contact. Three responsibilities cannot be delegated: verification, performance evaluation, and sound judgment.
Talking about AI with your team (weeks 1 to 2)
The kick-off meeting (45 min)
The sole objective is for everyone to leave understanding why AI is being introduced to the team, what changes, what remains the same, what is expected of them, and what they are given. The timed six-step process begins with a live demonstration of real-world use by the manager, not with slides. No specific numerical targets are announced: expectations are addressed after individual interviews.
Individual interview on AI (20 min)
An interview with each team member within two weeks of the meeting is conducted to listen, not to persuade. The four-step process (opening, listening with five questions, answering, concluding with a commitment from each member) reveals who is already using AI secretly, who is afraid, and who is waiting for permission. The notes focus on the profile, the chosen task, and what the manager should contribute, never on attitude: this is not an evaluation.
Addressing the fourteen most common objections
For each objection, the honest answer, the one best avoided, and the proposed action. The fourteen: "It will eliminate my job," "I don't have the time," "It's not reliable, it's just making things up," "It's not for my job," "It's cheating," "We'll be monitored," "I prefer to do it myself," "I don't know where to start," "Young people know, but I don't," "We change tools every six months," "What about confidentiality?", "Don't you use it yourself?", "AI will make us work more," "If an agent does my job, what's the point of me?". General rule: a frank answer that acknowledges the uncertainty is better than false reassurance.
To state what does not change
In a time of change, what is most reassuring is a clear list of what remains. The kit provides the written message to send to the team after the kick-off meeting, in four sections: what is changing, what is not changing (missions, quality, decisions about people made through conversations, confidentiality), what the manager doesn't know, and what they are doing themselves.
Set usage expectations (weeks 2 to 4)
Expectations by function
As recommended by Gallup, three expected uses per function are sufficient, formulated as behaviors ("prepares meeting minutes with AI and reviews them"), never as tools ("uses Copilot"). The table is completed after individual interviews, with a start date for each function; examples by profession initiate the discussion, from sales to development, including purchasing, HR, legal, and finance.
Integrating AI into annual objectives
A usage objective is based on knowledge sharing and quality, never on the quantity of requests. The toolkit offers five formulations to adapt: usage, knowledge sharing, quality, sharing, and guidance for the manager. Absolutely avoid: counting prompts, comparing employees, or penalizing non-use without providing time and support.
The protected experimentation time
The primary reason for dropping out is lack of time. The solution: one hour per week, at the same time for the entire team, for at least three months, allocated by the manager who adheres to this schedule. Each team member works on their own grid of three tasks, without any expected deliverables; after three months, the team collectively decides whether to keep, reduce, or replace the allocated time slot.
The three-task grid
This is the central tool in the kit: each employee identifies the three tasks that cost them the most time or energy, and redoes them with AI until they become proficient. The grid is completed during individual meetings or during protected time, and reviewed at 30 and 90 days. The time saved isn't automatically reinvested: the manager explicitly states where it will be used.
The time saved: deciding where it goes, and monitoring the intensification
In the BCG AI at Work 2026 survey, two out of three non-managerial employees reported receiving little to no guidance on how to use their time saved, and Berkeley researchers observed that AI can intensify work instead of lightening it. The toolkit offers five possible uses for time saved, a typical team rule, and five signs of intensification to watch for, including sloppy deliverables produced with AI; the AI-Gained Timekit delves deeper into the topic.
Promoting usage within the team (months 2 to 12)
Team rituals
Six rituals: "What I did with AI this week" (10 minutes in a team meeting), protected experimentation time, the monthly use case review, the team channel, the monthly check-in with the AI ambassador, and the 30-, 60-, and 90-day plan review. The weekly ritual is the most important and the most fragile: it only works if the manager demonstrates their own progress, including their failures, and if it never becomes a performance review.
Supporting each profile
Five profiles recur across all teams, each with its own stance and actions: the enthusiast who moves too fast, the competent skeptic, the silent one, the senior member who fears losing their value, and the junior member who knows better than the manager. The silent withdrawal deserves particular attention: it isn't reflected in standard performance indicators, and the response is never collective; it's a conversation without an audience.
Recognize without monitoring
The use of AI is acknowledged, not monitored: we celebrate results and sharing, we don't measure individuals. The table contrasts what we celebrate (a shared use case, an error corrected by verification, a "I can't do it" said aloud) with what we never measure: the number of requests per person, the time spent in the tool, the content of prompts, a ranking of collaborators.
Handling incidents
Three incidents will occur: an AI error in a deliverable, sensitive data entered into a tool, and a hidden use case discovered. For each, the toolkit outlines what the manager should do within the hour, what they should tell the team, and what they shouldn't do, such as assigning blame or sanctioning the person who reported the incident. Addressing these incidents quickly and calmly allows the team to continue trying.
What the manager entrusts to AI, and what he never entrusts to it
The question often comes up unasked: "Is my manager using AI to evaluate me?" The toolkit draws a clear line in two columns, to be stated before it's asked: preparing for an interview, summarizing, drafting a first version, analyzing anonymous and aggregated feedback, yes; evaluating, ranking, deciding on a promotion or contract termination, recording a meeting without mentioning it, never. The time freed up is meant to be more present, not less.
When the agents join the team
An agent follows a series of steps and can act within the applications. The manager doesn't have to build the agent, but they do make five decisions: what tasks are assigned, who builds and supervises the agent, the required level of quality and review, the data to be used, and the cost, including Copilot Credits for employees without a Microsoft 365 Copilot license (often called "Copilot Premium"). A worksheet in Appendix C helps prepare for delegating a task to an agent.
Responsibility, the right to make mistakes and social dialogue
Four rules: the right to make mistakes is written down; what is transmitted is proofread, so as not to shift the burden onto the recipient; the burden is measured, and the right to disconnect also applies to tasks launched with AI; social dialogue is prepared, because the introduction of a technology that modifies working conditions requires consultation with the CSE (article L. 2312-8 of the Labor Code), including in the pilot phase.
Measuring and reporting (from month 1)
The team dashboard
Nine indicators, all at the team level, none at the individual level, are updated monthly in ten minutes and shared with the team. They range from the percentage of the team using AI each week to the feeling of "I feel comfortable trying it," through perceived workload, to self-assessed managerial commitment across four behaviors: practicing, demonstrating, listening, and protecting time. This last indicator is the one managers often overlook, even though surveys place manager behavior among the factors most strongly linked to adoption.
The quarterly team mini-survey
Six anonymous questions, two minutes each, every three months: frequency of use, time saved and what is done with it, ease of experimentation and learning from mistakes, clarity of expectations and limitations, changes in workload, and what would be most helpful. The aggregated results are shared with the team and contribute to the dashboard.
The 90-day action plan
The plan fits on one page: a 90-day objective, milestones at 15, 30, 60, and 90 days, three personal commitments from the manager, and anticipated risks. Its last section is the most important: what the manager asks of their organization (time, a clear framework, support, priority, visibility). They review the plan at 30, 60, and 90 days with their own superiors or the AI program lead.
Attachments
Frequently Asked Questions from the Manager
Eleven questions, each with a short answer: not being convinced yourself, an already overloaded team, a colleague who refuses to use AI, preparing for an annual review with AI, whether the team is using AI without monitoring, a tool that changes without warning, what to do with the time saved, a team that says it's working more since the arrival of AI, an agent created by a team member, a Copilot Premium license for the whole team or not, a team member who knows more than the manager. Example: for the annual review, AI helps prepare the structure and questions, never evaluates the person or writes the assessment.
The five uses of the manager, briefing notes and departure prompts
A short fact sheet for each use case, with a starting prompt to adapt, a verification step, and a point of vigilance, without any personal or confidential data, plus a sixth fact sheet to prepare for delegating a task to an agent. Examples of points of vigilance: the AI does not know the person whose interview is being prepared; it sometimes confuses a proposal and a decision in a summary; it tends to validate your initial idea if you present it as a given.
Download the manager's toolkit for dealing with AI (Word, free)
30 pages, 5 parts and 5 appendices, all templates ready to use. Written by NextStart.AI based on Gallup, Microsoft Work Trend Index and BCG surveys on the role of managers in AI adoption, and on the Managers Catalyst programs conducted in large organizations.
- Self-diagnosis and the five uses of management, with prompts
- Kick-off meeting, individual interview, fourteen objections
- Expectations by function, protected time, three-task grid
- Rituals, profiles, dashboard, 90-day plan
7. Frequently Asked Questions
Should a manager master AI before their team?
He doesn't have to be the best user on his team, but he does need to practice. Adoption surveys consistently show that usage drops in teams whose managers don't use the tool, regardless of the training plan. The toolkit offers five ways managers can use it, with prompts, to get started in one hour.
How should you respond to an employee who fears for their job?
Don't promise anything the organization can't deliver, and clearly distinguish between what has been decided and what hasn't. The kit contains fourteen common objections with a response template, and a one-on-one interview template for addressing the issue in a face-to-face setting rather than in a meeting.
Should we set specific usage targets?
A target volume of output from the facade usage. The kit proposes setting expectations by function, protecting learning time and working on the grid of three tasks: what the AI does, what it prepares, what it does not touch.
How much time does this take the manager?
The 90-day plan relies on short rituals integrated into existing meetings: a use case review during team meetings, a case study sharing every two weeks, and dedicated time for testing. It's about regularity, not time commitment.
How does this relate to the AI charter and the ambassadors?
The charter outlines what is permitted, the ambassador demonstrates how to do it, and the manager states their expectations for the team and protects time. All three are complementary: section 5 details the division of roles and areas of overlap.
Is the kit linked to a particular tool?
No. It works with Microsoft 365 Copilot, ChatGPT, Claude, Gemini, Mistral, or an internal assistant. The provided prompts are designed to be pasted into any of these tools.
8. Going further
The toolkit equips the manager; it does not replace peer group work. The process Managers Catalyst NextStart.AI brings managers together for two half-days, each beginning with a two-hour practical workshop to gain practical experience, followed by work on the AI change management methodology, team communication, two weeks of practical implementation, and a 90-day action plan. For field representatives, the program AI Ambassadors trains the volunteers who then lead the workshops in each entity. Once the AI is adopted, the AI-powered time-saving kit helps the team decide what to do with the freed-up time, without increasing the workload.
Read also : AI Ambassador Network: The 2026 toolkit for building your collective, AI Charter in Business: The 2026 Model (Downloadable), AI Adoption Roadmap in 12 Months: The 2026 Toolkit, AI Maturity Grid: Assess your organization's readiness in 90 minutes, Measuring the uses, value and ROI of AI: the 2026 toolkit et AI Kit for HR Teams: 28 Use Cases and Prompts in 4 Levels.
The entire series of free kits. To drive adoption: AI charter template, AI training plan template, AI Ambassadors Kit, AI-powered time-saving kit, Adoption Roadmap Kit, AI maturity grid, AI Measurement and ROI Kit, AI Governance Kit, Cowork kit, 50 Copilot M365 prompts, Copilot Studio kit, Deploying Claude in the Workplace Kit, Claude Agents Kit, Claude Skills kit, 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 kit et communicator's AI kit.
9. sources
The figures and references cited on this page are based on the following sources.
- Gallup, Employee Engagement Remains Flat as AI Adoption Accelerates, July 21, 2026.
- BCG, AI at Work 2025: Momentum Builds, but Gaps Remain, June 2025.
- Microsoft, 2026 Work Trend Index: Agents, human agency, and the opportunity for every organization, May 2026.
- BCG, AI at Work 2026, June 2026.
- Ranganathan and Ye (Berkeley Haas), AI Doesn't Reduce Work, It Intensifies It, Harvard Business Review, February 2026.