Building AI capability at two scales: the enterprise and the community

Context

Four gaps kept appearing across our engagements, and none of them could be closed by another policy document.

APS staff were facing rapid generative AI change with limited hands-on exposure: plenty of guidance about AI, very little experience of it. Canberra tech practitioners were learning AI in isolation, without peers to test ideas against. Executive teams needed AI fluency to make credible investment decisions, and were making them anyway. And across all of it sat a gap between policy-level AI training and the real, messy, day-to-day practice of using these tools on actual work.

People don't become AI-capable by reading a policy. They become capable by trying things, failing safely, and learning alongside peers who are doing the same.

Our approach

We worked at two scales deliberately, because they solve different problems.

At enterprise scale

We deliver a signature whole-of-staff AI presentation: what is possible, what is risky, what is next. The value of an all-staff session is not depth. It is that an entire agency receives the same message at the same time. Where that does not happen, staff assemble their understanding of what is permitted from rumour and from whatever their branch happens to believe, and the resulting inconsistency is itself a governance risk.

We pair this with executive and SES literacy programs, on the straightforward logic that executives who understand AI make better investment and risk calls, and executives who do not will make those calls regardless.

At community scale

We co-founded AI in the Shadows with an Australian university, a small open meetup for Canberra practitioners. It runs regularly with around thirty people and is open to anyone in tech.

Two design choices define it. Sessions are built around real problems practitioners bring, not abstract theory. And it is curated as a safe, supportive space to push AI skills without fear of getting it wrong, which is precisely what people cannot do inside their own agency, where being visibly wrong about AI in front of colleagues carries a professional cost.

The university partnership keeps the sessions grounded in evidence rather than vendor enthusiasm.

Outcomes

• A whole-of-staff session reached an entire agency with one consistent message on AI.
• AI in the Shadows meets regularly in Canberra, around thirty people, open to anyone working in tech.
• The community produces shared prompts, patterns and war stories that circulate back into agencies. Practitioners take what they learn on a Tuesday evening into work on a Wednesday.
• The university partnership brings academic rigour and keeps content evidence-based.

Why both scales are necessary

Enterprise training creates permission and a shared baseline. Community practice creates confidence and depth. Neither substitutes for the other.

An agency that runs only the all-staff session has staff who know the rules and still hesitate to use the tools. A practitioner community without organisational sanction produces a handful of capable individuals whose skills cannot be deployed on real work. Running both is what turns AI capability from a training line item into something that persists after the engagement ends, which is the only definition of capability uplift we find meaningful.