When it comes to training your teams in AI in 2026, almost everyone agrees on the principle. The real question lies elsewhere: how do you go from a few curious people testing ChatGPT on their own to an entire team that saves time every week, without putting company data at risk? This guide answers in order: why train, who to train first, which curriculum, which format, how to fund it and how to measure the result.
At diagonall, we train teams at SMEs and large groups on ChatGPT, Claude and Copilot. What follows comes from that fieldwork. If you are looking directly for a program, our AI training for companies follows this method, with workshops built on the real files of each department.
Why train your teams in AI now
AI is already in your teams, whether the company decided it or not. Employees write emails, summarize documents or prepare quotes with consumer tools, often using their personal accounts. Training is first about taking back control of these uses: choosing the approved tools, setting what can be entered into them and what must never be entered, then showing how to use them properly.
It is also a question of results. A trained team knows how to spot the tasks where AI really saves time, checks what the tool produces and avoids costly mistakes. Trained employees also become the ones who suggest the next automations, because they know the processes better than anyone.
Finally, tools change every quarter. A team that has understood the principles adapts to new versions on its own. A team that has only learned tricks has to relearn everything with each update.
What the AI Act says about training
Since February 2, 2025, Article 4 of the European AI regulation (AI Act) requires companies that use AI systems to address the "AI literacy" of their staff. This article was rewritten by Regulation (EU) 2026/1744, known as the Digital Omnibus, which entered into force on July 27, 2026. AI providers and professional users must now take measures to support this literacy, with no obligation to reach a specific level for each employee. Article 4 has no penalty of its own.
The Digital Omnibus postponement mainly concerns the high-risk systems of Annex III, such as those used for recruitment, now expected on December 2, 2027. It does not remove what already applies. In the event of an incident or an inspection, being able to show who was trained, on what and when remains the best protection. For details on the timeline, read our guide to the AI Act for companies.
Assess the current level of AI skills
Before launching a training program, it is crucial to assess your team's current level. This can be done with a short questionnaire, a few interviews or a more complete audit. Ask three simple questions: who already uses AI, for which tasks, and with which tools. The answers quickly reveal the gaps and the uses that need to be governed first. To go further on this diagnosis, see our article on the AI maturity audit.
Who to train first
Not everyone needs the same training, or at the same time. An order that works well:
- Management and the executive committee. They set the rules, choose the tools and lead by example. A short, concrete session is often enough, like our AI training for executive committees and management teams.
- Champions in each department. One or two motivated people per team, trained in greater depth, who answer everyday questions.
- Business teams. Sales, customer support, HR, finance, legal: each with their own use cases. Sales teams, for example, have their dedicated AI training.
- Technical profiles, for those who will go on to build automations or agents.
Design a tailor-made training program
An effective program must be adapted to the needs of each department. Technical teams may need to go deep on automations and APIs, while sales teams will benefit more from learning to prepare a meeting, write a follow-up or analyze an account with AI. Including practical modules, with workshops and concrete projects, clearly strengthens learning. To test a first use case before rolling it out more widely, it is useful to go through an AI POC.
A typical three-step program
- A common foundation, from half a day to one day: how a tool like ChatGPT or Claude works, what it does well, where it goes wrong, and the company's rules on data.
- Workshops by function, on the team's real documents: everyone leaves with two or three uses they apply the very next day.
- Follow-up a few weeks later, to correct bad habits, share what works and move to the next level.
Train on the tools your teams will use
There is no point training on a tool the company has not chosen. If you have chosen ChatGPT, a ChatGPT training for companies is more useful than a general course. The same logic applies to Claude, with a Claude training, or to Copilot if you are equipped with Microsoft 365. The substance stays the same: phrase a request well, give the right context, check the answer.
Understand the fundamentals of artificial intelligence
To train your teams effectively, you need to start with the basics, without jargon. How a language model produces an answer, why it can confidently make up information, what happens to the data you entrust to it, and how to check what it produces. A solid understanding of these notions helps everyone see how AI can change their daily tasks.
Emphasis must also be placed on responsible use: customer data, trade secrets, copyright and decisions that affect people. Investing time in this phase avoids future complications and lays a solid foundation for more advanced uses, such as advanced prompting.
Plan hands-on workshops and simulations
Once the basics are understood, hands-on workshops are the essential step. Participants work directly with the tools, on documents close to their daily work, and build small use cases. These sessions are ideal for identifying difficulties and refining skills in a supervised setting.
In addition, simulations that reproduce real company situations in a fictional context make it possible to test without risk: responding to a customer complaint, preparing an offer, summarizing a file. To boost engagement, plan challenges between teams that reward the most useful ideas.
In-person, remote or e-learning
In-person remains the most effective format for hands-on workshops: mistakes are corrected live and participants learn from one another. Online platforms such as Coursera or LinkedIn Learning can complement the internal program: they let people learn at their own pace and often offer discussion forums. The right choice depends on the size of the team, how geographically spread out it is and its level. We compare the two options in our article AI training: in-person or e-learning.
Implement pilot projects
Pilot projects are an excellent way to put the skills acquired into practice. They allow employees to test new ideas in a controlled setting and see the real impact of AI on their work. A successful pilot then serves as a model for other departments. To avoid common pitfalls, see our guide on mistakes to avoid in an AI project.
Encourage continuous learning
AI evolves quickly, and it is important to cultivate a culture of continuous learning. An internal channel where everyone shares their best prompts, a fifteen-minute monthly slot in a team meeting or a club of champions are often enough to keep the momentum going. Conferences, webinars and specialized publications also help people stay up to date. We detail this approach in our article on building an AI culture in your company.
Fund the training
Your employees' AI training can be covered by your OPCO, the skills operator for your industry branch. diagonall training courses are eligible for OPCO funding. Coverage rules vary depending on the OPCO, the branch and the size of the company: check with yours before signing, and plan the request ahead, as the application generally has to be filed before the training starts.
Measure the return on investment of AI training
It is essential to measure the effect of the training in order to decide what comes next. The right indicators are simple: time saved on a few specific tasks, the number of people who use the tools every week, the quality of the documents produced and team satisfaction. Measure before the training, then one to three months after. Learn more about the calculation in our article on AI ROI.
A typical scenario
Take a retail company that wants to manage its inventory better. It starts with a session for management, which chooses a tool and writes simple rules on data. It then trains one champion per department. The purchasing and logistics teams attend a workshop on their own files: analyzing past sales, spotting overstocked products, preparing orders. A few weeks later, a follow-up session brings up a heavier need, demand forecasting, which becomes a pilot project run with the teams already trained.
The gain does not come from a single workshop. It comes from the sequence: management setting the framework, champions answering questions, teams working on their real files, then a concrete project that extends the training.
Conclusion
Training your teams in AI is an investment whose payoff shows quickly when it starts from the real jobs. Start with management and a few champions, build workshops on real tasks, govern usage and track the results. Your teams will then be able to take advantage of today's tools and those arriving tomorrow.
Going further
- Building an AI culture in your company
- The best AI training courses for managers
- AI training: in-person or e-learning?
- Advanced prompting: mastering ChatGPT
Frequently asked questions
What is the ideal length of AI training for teams?
A common foundation takes half a day to one day. It should then be followed by workshops by function and a follow-up a few weeks later. What matters is that the training continues over time, because tools change quickly.
Should all employees be trained in AI?
Everyone benefits from knowing the basics and the company's rules, especially those who already use AI tools. In-depth training can be reserved for champions and the teams who will use AI the most.
Is AI training mandatory under the AI Act?
Article 4 of the AI Act requires companies that use AI to take measures to support the AI literacy of their staff. Since its rewrite in 2026, it does not impose a specific level to be reached and has no penalty of its own, but it remains in force.
How can AI training be funded?
Your OPCO can cover all or part of the cost, depending on your branch and the size of the company. diagonall training courses are eligible for OPCO funding. Submit the request before the training starts.
What are the common mistakes in AI training?
Training that is too general and disconnected from real tasks, the absence of rules on data and the lack of follow-up after the session. Successful training starts from the team's files and continues with a follow-up session.



