An AI workshop for your team: what should it deliver and how do you choose one?
An AI workshop only succeeds if your team works differently afterwards. What research says about training that sticks, which format fits and how to choose
Summary
- One in three Belgian workers uses generative AI for work: 32.7% in 2025, against 21.8% in the EU (Eurostat)
- According to an Acerta survey (2026), 21.4% of employees took AI training through their employer in the past year, while 59.1% want to
- Among Belgian companies that considered AI but do not use it, lack of expertise is the biggest barrier: 74.5% cite it (Eurostat, 2025, companies with 10 or more employees)
- In an experiment with 758 consultants, a short explanation of prompting (videos and documents) made people better where AI is strong, but worse where AI fails: a workshop has to teach people when AI can and cannot be trusted
- Training that sticks works with people's own tasks, lets them practise and make mistakes, and is followed up after the session
- A certificate is not mandatory, but even after the amendment of 27 July 2026, anyone using AI must still take measures to support the AI literacy of their people
You bought a few licences for ChatGPT or Claude, and a couple of colleagues use them every day. The rest tried it once and gave up. Sound familiar? Then the problem usually lies in what your team can and dares to do with the tool.
An AI workshop should change that, but not every workshop does. Below, you’ll find what a good AI workshop delivers, what Belgian figures and research say about AI at work, why a lesson in prompting alone is not enough, which format suits your team and how to choose a provider.
What is an AI workshop and what should it deliver?
An AI workshop is a practical session in which your team learns to work with tools such as ChatGPT, Claude or Gemini, applied to their own work. It has succeeded when your team then works with AI independently and responsibly: producing texts, analyses or emails faster, and knowing where AI makes mistakes. A workshop that only inspires is a keynote.
The difference lies in what happens the week after. An inspiration session makes a team want to start, a workshop gives the team the skills to actually do it. Both have their place, as long as you know what you are buying.
ClickForest delivers founder-led AI training to teams of SMEs in Mechelen, Antwerp and across Flanders: Frederiek Pascal runs the sessions himself and builds with AI every day. The content is tailored to the team’s sector and tasks, with the aim that the team then works independently, without lasting dependence on an outside party. More about the approach is on the page about AI training for teams.
Why does your team need an AI workshop if the tools are already there?
Because your team is probably already using AI, just without any agreements. In Belgium, one in three workers used generative AI for work in 2025. At the same time, 57% of employees worldwide say they hide their AI use, and 66% rely on the output without checking it. A workshop brings that use into the open and improves it.
According to Eurostat, 32.7% of Belgian workers used generative AI for their work in 2025, compared with 21.8% on average in the EU (High Council for Employment, 2026). Among Belgian managers, 47% admit to sometimes using AI secretly, and 18% use tools their employer has not approved (Inetum, 2026). According to an Acerta survey of 1,204 employees and 449 employers in the private sector, 62% of employees use AI tools, while 53.2% of employers were working on an AI policy (Acerta, 2026).
The international picture is the same. In the global study by KPMG and the University of Melbourne, 57% of employees hide their AI use, 66% rely on the output without checking it and 56% say they make mistakes because of AI (KPMG, 2025). According to the Work Trend Index by Microsoft and LinkedIn, 78% of AI users bring their own tools to work, and 80% at SMEs (Microsoft, 2024).
“They highlight the importance of effective governance and training, and creating a culture of responsible, open and accountable AI use.”
— Nicole Gillespie, Chair of Trust and Professor of Management, Melbourne Business School
That use without agreements comes at a price: company data ending up in a public tool, or a wrong figure in a quote. Which tool suits which task is compared in the article on ChatGPT, Claude and Perplexity in practice.
How many Belgian companies already use AI, and where does it get stuck?
In 2025, 34.5% of Belgian companies with at least 10 employees used one or more AI technologies, compared with about 20% in the EU. Smaller companies lag behind: at 10 to 49 employees, it is 28.8%. The biggest barrier is knowledge: 74.5% of Belgian companies that considered AI but do not use it cite a lack of expertise.
| Company size | Belgium (2025) | EU-27 (2025) |
|---|---|---|
| All companies with 10 or more employees | 34.5% | 20.0% |
| 10 to 49 employees | 28.8% | 17.0% |
| 50 to 249 employees | 54.5% | 30.4% |
| 250 employees or more | 76.4% | 55.0% |
Source: Eurostat and Statbel, companies with 10 or more employees; micro-businesses are not included. Belgium scores well, but the gap between small and large companies is wide.
Training does not keep up with use. According to Acerta (2026), 21.4% of Belgian employees took AI training through their employer in the past year, while 59.1% want to. 32% of companies offer it, and 38.1% have no plans to (Acerta, 2026). The Belgian High Council for Employment paints a similar picture: 39% of employees need extra AI knowledge, 14% took relevant training.
“Als werkgever neem je dus beter het zekere voor het onzekere door je medewerkers in het gebruik van AI op te leiden.”
— Hannelore Van Meldert, AI expert at Acerta
In other words: as an employer, the safe choice is to train your staff in using AI.
What does research say about what AI really delivers?
A lot, but not everywhere. In an experiment with 758 BCG consultants, people with AI worked 25.1% faster and delivered better quality on tasks within AI’s capabilities. On a task outside them, they reached the right answer on average 19 percentage points less often than colleagues without AI. AI makes your team stronger or weaker, depending on the task.
That study by Dell’Acqua and colleagues was published in Organization Science in 2026 (Harvard Business School). Within the frontier, the consultants completed 12.2% more tasks. On the task outside it, the control group was right 84.5% of the time, the AI groups 60% and 70.6%. The researchers call that uneven boundary the “jagged frontier”.
Two other studies point the same way. Among 453 professionals, the time for writing tasks fell by 40% and quality rose by 18% (Noy and Zhang, Science, 2023). Among 5,179 customer service agents, productivity rose by 14% on average, and by 34% for new and less experienced agents, while experienced people gained little (Brynjolfsson, Li and Raymond, 2023).
An important caveat: all these studies measure what access to AI delivers, not what a workshop delivers. We did not find a study that measures the effect of an AI workshop itself on productivity. Ethan Mollick, one of the authors, sums up the lesson like this: “The authoritativeness of AI can be deceptive if you don’t know where the frontier lies” (One Useful Thing, 2023).
Why is a lesson in prompting not enough?
Because a general explanation of prompting makes people better where AI is strong, but does not protect them where AI fails. In the BCG experiment, one group received videos and documents on good prompting beforehand. Within the frontier, that group did at least as well as the group with AI only. On the task outside it, it scored worse than the group without the explanation: 60% correct versus 70.6%.
That explanation was a package of videos and documents, so not a workshop. Still, the lesson is useful. One possible explanation, which the study itself does not test: people who learn how to get AI to do more also start relying on it more. The researchers write that the effect of AI depends on human judgement: knowing which tasks suit AI and where your own expertise should take priority. A good workshop teaches that judgement, with examples from the team’s own work where AI gets it wrong.
“Leaders also should consider training less an opportunity to learn prompting techniques (which are valuable but getting less important as models get better at figuring out intent), but as a chance to give people hands-on AI experience and practice communicating their needs to AI.”
— Ethan Mollick, Associate Professor of Management, Wharton
There is another reason. Danielle Li of MIT Sloan, co-author of the customer service study, warns: “If you’re just typing the recommendations, then you may not necessarily be learning” (MIT Sloan, 2023). Anyone who wants to learn to write, analyse or calculate has to keep doing the thinking. Prompting remains useful; the article on writing better prompts explains the basics.
What goes into a good AI workshop?
A good AI workshop works with the team’s own tasks, lets participants practise and make mistakes themselves, teaches where AI fails, sets agreements on data and tools, and is followed up after the session. Most of those elements come from research on workplace training. What happens before and after the session matters as much as the session itself.
| Element | What research says | What to look for |
|---|---|---|
| Knowing the needs in advance | What happens before and after training is as important as the training itself (Salas et al., 2012) | A conversation about tasks, tools and level before the session |
| Practising with mistakes | Active exploration and encouraging errors improve application to new tasks (Keith and Frese, 2008) | Participants work themselves, on real assignments |
| Own work as material | Transfer is linked to the opportunity to practise and to support at work (Blume et al., 2010) | Examples from the team's own sector and tools |
| Knowing where AI fails | Outside AI's frontier, people give the right answer less often, and even less often after an explanation of prompting (Dell'Acqua et al., 2026) | Explicit exercises on errors and made-up answers |
| Follow-up and spacing | People who received five or more hours of training use AI more often on a regular basis (BCG, 2025, correlation) | Several sessions or a follow-up moment after a few weeks |
Training works, including on behaviour at work: a meta-analysis based on 122 effect sizes found a medium to large effect (Arthur et al., 2003). In error management training, where participants actively explore and are allowed to make mistakes, the effect on application to structurally different tasks was highest (Keith and Frese, 2008). The meta-analysis by Blume and colleagues does add a nuance: what sticks after training is linked to motivation and to a work environment that supports its use (Blume et al., 2010).
This is what it looked like at biotech company VF Bioscience. ClickForest gave the management team Claude training in three on-site sessions of 2 hours each, from the basics to Projects, MCPs and Skills. Between sessions, there was room to experiment and bring questions to the next meeting. After the three sessions, the management team works with Claude independently, with its own Projects for reports, emails and analyses. How the Claude app works is explained in the article on Claude Desktop; what Skills and MCP do, you can read in Claude Skills and MCP.
Which format suits your team: keynote, workshop or programme?
A keynote creates insight and buy-in, a workshop teaches a team to work with the tools, and a programme embeds AI in daily work. Choose based on what you want to achieve. If you want people to really work differently afterwards, a programme with several sessions and time to practise in between is the most obvious choice.
| Format | Duration | What it delivers | Suitable for |
|---|---|---|---|
| Keynote or inspiration session | Short, broad | Insight and buy-in | Events, a start for the whole organisation |
| Workshop | Half a day or a full day | Working with the tools on own tasks | A team that wants to get started concretely |
| Programme | Several sessions, spread over time | AI embedded in daily work | Teams that want to use AI structurally |
ClickForest offers all three, in-company on site or online via video call, for teams of SMEs in Mechelen, Antwerp and across Flanders. After the training, the team has reference material to fall back on. If you want to roll out AI more widely, with a shared workspace and connections to your own software, an AI implementation is the next step. If you are unsure which format suits your team, a short conversation about the AI training can work it out with you.
How do you choose a provider for an AI workshop?
Choose a provider that asks about your tasks in advance, practises with your own work, is honest about what AI cannot do, sets agreements on data and tools, and provides follow-up. Also ask who runs the session, and whether that person works with AI every day. A standard programme with slides on the history of AI delivers little.
Useful questions for a provider:
- What do you ask about our team in advance? A good provider wants to know which tasks, tools and levels there are.
- What material do we practise with? Preferably with examples from your own work.
- How do you teach where AI fails? A workshop without exercises on errors and made-up answers is incomplete.
- Which tools are covered? Choose the tools your team actually uses or wants to use, such as ChatGPT, Claude, Gemini or Make.com. An overview per task is in the article on AI tools per task.
- What happens after the session? Reference material, a follow-up moment or a follow-up session make the difference.
- Who runs the session? Someone who works with AI every day can answer questions from practice.
“Companies cannot simply roll out GenAI tools and expect transformation.”
— Sylvain Duranton, Global Leader of BCG X
If you also need help choosing tools or an approach for the whole company, the article on AI consultancy for SMEs goes deeper into that. A comparison of the most popular tools is in the AI tools comparison.
Is AI training mandatory for your team?
A certificate or a specific course is not mandatory. Since the Omnibus, in force from 27 July 2026, anyone using AI must still take measures to support the AI literacy of their people, without having to guarantee a particular level. A practical workshop is a logical way to fulfil that.
The text of Article 4 of the AI Act is in Regulation (EU) 2026/1744. The European Commission writes in its questions and answers: “There is no need for a certificate” (European Commission). According to the Commission, employees who, for example, write ad copy with ChatGPT should be aware of specific risks such as hallucinations. What else the AI Act asks of your chatbot and your AI content is covered in the article on the EU AI Act. What your employees need to know at a minimum and what you keep track of is explained in our guide to AI literacy.
Conclusion: an AI workshop only succeeds if your team works differently
Your team is probably already using AI, with or without your knowledge. The question is whether it does so well: with the right tools, with agreements on data, and with the judgement to see where AI gets it wrong. That is what an AI workshop is for.
Choose one that practises with your own work, teaches where AI fails and provides follow-up. Then measure what changes in the work in the weeks that follow.
ClickForest delivers practical AI workshops and programmes for teams of SMEs in Mechelen, Antwerp and across Flanders, tailored to the team’s tasks and tools. Want to know which format suits your team? Get in touch.
Grow with AI, practically and without unnecessary complexity
Ready to use AI to save time and accelerate your marketing? Discover our AI services
Discuss your challenge directly with Frederiek: Book a free strategy call or send us a message
Prefer email? Send your question to frederiek@clickforest.com or call +32 473 84 66 27
Strategy without action remains theory. Let's take your next step together.
Frequently asked questions
That depends on the goal. A keynote is short and broad and mainly gives insight, a workshop takes half a day or a full day and teaches people to work with the tools, and a programme spreads several sessions over time. In a BCG survey, people who received five or more hours of training used AI significantly more often on a regular basis; whether the training caused that, the survey does not measure.
The price depends on the format (in-company, online or keynote), the duration, the group size and the topics. ClickForest delivers tailored AI training to SMEs in Mechelen, Antwerp and across Flanders, and discusses a clear estimate in the introductory call based on the specific situation.
By giving AI a fixed place in the work. Spread sessions over a few weeks, let people practise with their own tasks in between, agree on which data may go into which tool, and have a manager actively follow up on usage. Learning research shows that support from the work environment is linked to what people actually apply after training.
No. A good workshop starts from the level of the group and tailors content and pace in advance. ClickForest therefore starts an AI training for teams of SMEs in Flanders with a conversation about the team's tasks and tools, so that both beginners and people who already work with ChatGPT or Claude take something away.
Sources and references
Figures for Belgium and Europe:
- Statbel: "ICT en e-commerce bij ondernemingen" (2025) · https://statbel.fgov.be/nl/themas/ondernemingen/ict-en-e-commerce-bij-ondernemingen
- Eurostat: "Use of artificial intelligence in enterprises" (2025) · https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises
- Eurostat: dataset "Artificial intelligence by size class of enterprise" (isoc_eb_ai) · https://doi.org/10.2908/ISOC_EB_AI
- Belgian High Council for Employment: "Artificiële intelligentie op de Belgische arbeidsmarkt" (2026) · https://hrw.belgie.be/nl/home/verslagen-adviezen/verslagen-2026/artificiele-intelligentie-op-de-belgische-arbeidsmarkt-februari-2026
- Acerta: "Aantal bedrijven dat AI-opleiding aanbiedt aan werknemers op één jaar tijd met 15% toegenomen" (2026) · https://www.acerta.be/nl/in-de-pers/aantal-bedrijven-dat-ai-opleiding-aanbiedt-aan-werknemers-op-een-jaar-tijd-met-15-toegenomen
- Acerta: "Werknemers in openbare sector krijgen vaker AI-opleiding op het werk dan werknemers in privé" (2026) · https://www.acerta.be/nl/in-de-pers/werknemers-in-openbare-sector-krijgen-vaker-ai-opleiding-op-het-werk-dan-werknemers-in-prive
- Inetum: "Tech & AI 2026 Observatory" (2026) · https://www.inetum.com/be/nl/about-us/media-center/press-release/barometer-jobs-ai.html
International studies:
- Microsoft and LinkedIn: "2024 Work Trend Index: AI at work is here. Now comes the hard part" (2024) · https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part
- KPMG and University of Melbourne: "Trust, attitudes and use of artificial intelligence: a global study 2025" · https://kpmg.com/xx/en/media/press-releases/2025/04/trust-of-ai-remains-a-critical-challenge.html
- BCG: "AI at Work 2025: momentum builds, but gaps remain" (2025) · https://www.bcg.com/press/26june2025-beyond-ai-adoption-full-potential
- Dell'Acqua et al.: "Navigating the jagged technological frontier", Organization Science (2026) · https://www.hbs.edu/ris/Publication%20Files/dell-acqua-et-al-2026-navigating-the-jagged-technological-frontier_5c589c8c-fbb5-458f-b285-c944746cd717.pdf
- Brynjolfsson, Li and Raymond: "Generative AI at work", NBER (2023) · https://www.nber.org/papers/w31161
- Noy and Zhang: "Experimental evidence on the productivity effects of generative artificial intelligence", Science (2023) · https://europepmc.org/article/MED/37440646
Learning research and regulation:
- Arthur, Bennett, Edens and Bell: "Effectiveness of training in organizations: a meta-analysis of design and evaluation features" (2003) · https://europepmc.org/article/MED/12731707
- Keith and Frese: "Effectiveness of error management training: a meta-analysis" (2008) · https://europepmc.org/article/MED/18211135
- Blume, Ford, Baldwin and Huang: "Transfer of training: a meta-analytic review" (2010) · https://jasonhuangatwork.com/papers/Blume%20Ford%20Baldwin%20Huang%202010%20JOM%20-%20Transfer%20Meta.pdf
- Association for Psychological Science: "The science of training and development in organizations" (2012) · https://www.psychologicalscience.org/publications/journals/pspi/training-and-development.html
- EUR-Lex: Regulation (EU) 2026/1744 (2026) · https://eur-lex.europa.eu/legal-content/NL/TXT/HTML/?uri=OJ:L_202601744
- European Commission: "AI literacy, questions and answers" · https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers
- Ethan Mollick, One Useful Thing: "Making AI work: leadership, lab, and crowd" (2025) · https://www.oneusefulthing.org/p/making-ai-work-leadership-lab-and
- Ethan Mollick, One Useful Thing: "Centaurs and cyborgs on the jagged frontier" (2023) · https://www.oneusefulthing.org/p/centaurs-and-cyborgs-on-the-jagged
- MIT Sloan: "Workers with less experience gain the most from generative AI" (2023) · https://mitsloan.mit.edu/ideas-made-to-matter/workers-less-experience-gain-most-generative-ai
