Vana Vibes Brisbane

It usually goes like this. The licences got bought. The town hall happened. A training module went out and the completion rate looked fine. Six months later, the same handful of enthusiasts are the only people using anything, and the board is asking what the AI budget actually bought.

If that sounds like your organisation, the research says you are the norm. MIT’s State of AI in Business study found that 95 per cent of enterprise generative AI pilots deliver no measurable return. The researchers blamed what they called a learning gap: organisations deploy tools faster than their people can absorb them. Model quality barely featured in their failure analysis.

Your employees would probably agree. Nine in ten workers now use AI at work in some form, yet only one in six feels fully prepared to use it well (Study.com, 2026). And the training you have run so far has likely missed: 85 per cent of workers say they cannot connect their AI training to the actual job they do every day (Docebo, 2026).

On Thursday 16 July in Brisbane, Open Data Labs ran a session that tested a different model. We took eighteen professionals from nursing, law, accounting, marketing, publishing and professional services, spanning daily AI users to people who had never opened a chatbot, through three hours of hands-on work. Your workforce looks a lot like that room. Three things we saw in that room explain the corporate numbers.

Your training misses because it is about AI, and their day is not

Adult learning research has said this for fifty years. Adults engage when the material is problem-centred, connects to their own experience, and applies immediately. Look at your last AI training module against those three tests.

What the Brisbane room loved most was solving a practical challenge together, one that was relatable without being work. This is why we partnered with the Vana Foundation. Their Vana Vibes series picks a societal problem and invites people to hack at different aspects of it, and this edition picked the cost of housing in participants’ own city.

The first challenge asked every team to draft a tenant’s reply to a steep rent increase. The scenario was fictional, built on Brisbane’s real rental figures, and it worked because rent is a sore subject for almost anyone who lives here. People threw themselves at it without worrying about looking underskilled, because nothing in it touched their actual job.

That is the design. Skill and confidence get built on a shared problem with low stakes, then carried into the business problems that matter. Most corporate training starts at the second step and wonders why nobody engages.

Your people use AI at different levels. Put them in the same room.

AI is a different tool to a beginner drafting an email than it is to someone building an app, and useful training has to meet both. We structured the evening as a ladder across four use-case levels: drafting, research, content creation and application building. Everyone found a challenge pitched at their level.

The mixed teams did the rest. Professionals at different levels, deliberately seated together and rotated through the night, taught each other constantly. Most people have never watched a colleague use AI, which is why so few habits ever improve. A publisher taught a table of accountants how she directs tone. A lawyer showed his table how he checks an AI’s legal claims before relying on them. None of that needed a trainer, and the same knowledge is already sitting in your workforce.

“I normally reach for the same AI tool for everything. The person beside me ran on a completely different set, and she was producing finished content in a few clicks. I wrote her list down on the spot and I’m keen to try all of it.”

Kathy, Customer Success Manager, publishing industry

Your people get creative when they choose their own tools

We deliberately made the session tool-agnostic. Participants could use any AI tool they liked, and that turned out to be one of the things they valued most about the night. People explored beyond their job-mandated suite, compared products across tables, and left with new tools and techniques they chose themselves.

Your organisation is having this experience whether you designed for it or not. MIT’s research found employees at over 90 per cent of firms use personal AI tools at work, while only around 40 per cent of firms have an enterprise subscription at all. Locking the suite down mostly pushes the behaviour out of sight, where the habits are ungoverned and the learning stays private.

Top-down lockdown is one answer to that, and it is usually the wrong one. The Brisbane room points at alternatives that work with the behaviour instead: give people sanctioned space to explore, and teach the data and confidentiality judgement that makes their choices safe. Peers will find and share what works without being asked. We design tool-choice frameworks and capability programs that solve the personal tooling problem without taking the tools away. Choice and security trade off far less than most policies assume.

The same design works inside your organisation

AI capability building works when it is designed the way adults actually learn: peer-focused, active, experimental, staged and social. That is a change management discipline, and it cannot be bought as a licence bundle.

This is where Open Data Labs innovates. Our training runs on data as much as on AI: sessions use real datasets, participants learn to interrogate sources and verify claims rather than trust outputs, and the governance conversation covers both the tools and the data moving through them. Our team has spent the past decade building and training large distributed workforces for the global AI industry, and we build on the Vana network, open data infrastructure that now serves more than a million people.

An organisational program follows the sequence Brisbane demonstrated. It opens with a shared, relatable challenge that builds skill and confidence across mixed-level teams, then bridges those skills onto your live business problems in staged follow-on sessions. Tool-choice guardrails are co-designed with your security and data teams before anyone opens a laptop, and application is measured at 30, 60 and 90 days, because capability that does not show up in the work a month later was never built.

The gap between the 95 per cent of stalled pilots and the successful few comes down to whether anyone built the human capability layer. The lowest-risk way to find out what that looks like for you is the way Brisbane did it: one session, one problem, your people. We scope pilot sessions in a 30-minute call, and the call itself will tell you more about where your rollout is stuck than the last quarter of dashboard reviews has.

---

Vana Vibes Brisbane was delivered by Open Data Labs in collaboration with the Vana Foundation. For training and capacity building enquiries: oshin@opendatalabs.com.

Sources: MIT NANDA, The GenAI Divide: State of AI in Business 2025 (pilot failure rate; personal AI tool use vs enterprise subscriptions); Study.com workforce AI readiness survey, 2026; Docebo corporate AI training survey, 2026.