The method

In a world short of attention, the experience is the product.

Most corporate training programmes are content delivery. We design experiences. Real-time, facilitated, behaviour-surfacing experiences where teams practice the thing instead of being lectured about it. Below is why we work this way, and what we have learned from ten years across two continents, 60 organisations, and 12 industries.

01 · The problem

The AI training industry is solving the wrong problem.

Every organisation we have spoken to in the last twelve months is buying the same thing. A ChatGPT workshop. A prompt engineering course. A library of tutorials. They are spending real budgets, and the results are quietly underwhelming.

The standard programme teaches people to use a tool. Click here. Prompt like this. Try this template. It treats interaction with an LLM the way you would treat a software update. That framing is already obsolete in 2026. The leading edge is agentic workflows, multi-model orchestration, and human-AI teams where the human designs, evaluates, and intervenes rather than types clever instructions. None of that works without an accurate mental model of how the underlying systems actually behave.

Your AI colleague has no memory of yesterday. It is confidently wrong in ways uncorrelated with how confident it sounds. It optimises literally for what you asked, not what you meant. People who internalise those traits work with the system differently. People who do not, do not. The research supports this directly.

BCG (2024, via IBM): 89% of companies say their workforce needs improved AI skills. Only 6% have begun upskilling in any meaningful way. Not finished. Started.

SHRM (2024): roughly 1 in 5 organisations believe their current upskilling efforts are working. The common failures are training irrelevant to actual roles, sessions nobody has time to attend, and content disconnected from real work.

02 · What we have observed

The teams that perform are the teams that understand how their peers think.

In ten years of designing engagement systems, one pattern has held more consistently than any other we have observed: the teams that performed were not the ones with the best tools or the smartest individuals. They were the ones where peers genuinely understood how each other thought. Not what colleagues did. How they thought. What they noticed first, what they missed, what they got defensive about, what they were unusually good at.

That shared understanding is a more reliable predictor of output quality than individual skill or formal incentive structures. It is the difference between a team and a group of people sharing a Slack channel. Today every organisation is adding a new kind of colleague to that team, the AI one, and pretending the principle no longer applies.

Most senior leaders can describe their human colleagues' thinking patterns in extraordinary detail. Then they open a ChatGPT tab and treat it like a vending machine.

03 · Why simulation

You cannot read your way into intuition.

You build it by doing the thing, failing at it, doing it again, and noticing what changed. This matches what the broader training research has found: simulation and experiential formats consistently rank as the most effective for skill transfer into actual job performance. We learned language by speaking it badly. We learned to walk by falling. We learn the working psychology of any new colleague, human or otherwise, by interacting inside a context where the stakes are real but failure is safe.

A simulation where you have to collaborate with an AI to solve a real problem will teach you things no slide deck can. You will see the moments where its confidence misleads you. You will feel the difference between a well-framed question and a lazy one. You will build the intuition.

Knoth et al. (2024, peer-reviewed): AI literacy, specifically a user's understanding of how AI works and the roles it plays in interaction, predicts prompt quality, which in turn predicts the quality of output people get. The chain holds.

04 · How we work

How we design the experience.

Same four steps every time. No off-the-shelf packages, no scripted facilitation, no two engagements identical.

01
Listen
Discovery call with the people accountable for the outcome. Real problems, real KPIs, real constraints. We will not propose anything in the first call. We are listening for the behaviour you actually want to change.
02
Design
We pick mechanics that surface the behaviour we want to study. Energy budgets surface prioritisation. Hidden information surfaces communication patterns. Time pressure surfaces fold discipline. The game is the instrument; the behaviour is the data.
03
Live
Facilitated session, 60 to 120 minutes depending on format. Players play. We observe. The system captures structured signals (turn-by-turn actions, messaging, fold decisions). No surveys. No self-report.
04
Mirror
Structured debrief while the experience is still fresh, written team commitments, and a manager-ready report with evidence references. The thing that bothered your team last quarter will appear in the report with a turn number next to it.

05 · Track record

The compounding kind of number.

10: Years building engagement systems
60+: Enterprise organisations
3,000+: Participants engaged
40+: Game mechanics in the library
12: Industries covered
100%: Of sessions end with a written team commitment