What actually changed by the end?

The situation

The team was not new to AI. But by their own account the work revolved around a couple of familiar tools; the question had not yet moved from "which tool do I open" to "how do I do this piece of work better with AI". The programme was built around that gap: not a tour of tools, but the team's own work on the table.

What we did

• →Writing effective prompts with the CRAFT framework: framing the question before choosing the tool
• →Live work across four tools: Perplexity, Claude, Gemini and Gamma, comparing how the same task diverges in each
• →Hands-on practice on real scenarios participants brought from their own desks
• →Ethical dilemmas and data security, without which none of the rest is usable inside a company

The outcome

By the end of the day two thirds of participants reported feeling more confident using AI. What came through in the feedback was not having learned another tool, but noticing how much wider the set of options was.

The situation

Marketing, sales and general managers across the Citroën, Peugeot, Opel and DS brands. What was missing at this level was not knowledge of the tools but calibration: at the start of the session most participants overestimated what AI could do and underestimated how much human oversight it still needs. Investment decisions rest on exactly those two misjudgements.

What we did

• →An overview of the AI ecosystem: which tool exists for what, and where its limits begin
• →Myths against reality: that agents run without human oversight, that a larger context window improves performance by itself, that hallucination will be solved outright in the next generation, that a fine-tuned model always beats a general-purpose one
• →Hands-on prompting. In the exercises a well-framed prompt was seen to improve output quality by as much as 150 per cent
• →A lab on real business scenarios, close to the decisions the group actually makes

The outcome

What the day produced was not a list of tools but a recalibrated expectation: telling apart where AI is genuinely strong from where human oversight remains non-negotiable. Participant satisfaction was measured above 95 per cent.

The situation

Recorded online courses have a well-known problem: people start them and do not finish. The industry average sits near 15%, meaning eighty-five of every hundred people drop out. The issue is rarely the quality of the material. It is whether a course nobody is chasing you about can hold on to a place in your working week.

What we did

• →Six modules: generative AI foundations, AI governance, tools and their applications, prompt engineering, AI in department-level work, and human and AI collaboration
• →Each module was built to stand on its own, so people could move through it in fifteen to twenty minute sessions between blocks of work
• →Every section closes with something to try, so what was just learned gets used immediately

The outcome

951 people completed the course and earned the certificate, at a 60% completion rate: four times the industry average. Satisfaction came in at 4.7 out of 5. It is the strongest piece of evidence on this site, because it shows people choosing to finish in a setting where nobody was making them.

The situation

Participants in Hepsiburada’s programme for women entrepreneurs: women from a range of fields, most without a technical background, running their own businesses. The obstacle here is not the number of tools but the threshold to starting. Nobody spends time learning this before seeing what it does for their own work.

What we did

• →An entry point that assumes no technical knowledge: less what the tools are, more what they make easier
• →Hands-on practice on examples from participants’ own businesses
• →Concrete uses showing where AI fits into how they already work

The outcome

This was outreach rather than a corporate team programme. The aim was not measurable behaviour change but getting entrepreneurs without a technical background to try it once. No outcome measurement of the kind collected on corporate programmes was taken here, so there is no figure.