Canteen: Building Governance Before Building Tools

The situation

Canteen is a national not-for-profit supporting young Australians living with cancer — as patients, siblings, or bereaved family members. Its board includes five young Member Directors, who bring the lived experience the organisation exists to serve. When members asked Canteen to take AI seriously as a tool for improving services, the organisation faced a governance challenge familiar to many NFPs: how do you move quickly on AI without creating unacceptable risks for a vulnerable population?

What they did

The board created a time-bound, management-led AI Sprint Committee with a direct line to the board. Set to run for twelve months, the committee's remit was threefold: set guardrails for AI use, clarify who is accountable for what, and integrate AI oversight into Canteen's existing governance arrangements — while also identifying opportunities for AI to improve services.

Critically, the committee was not a governance mechanism only. Its brief explicitly included innovation. The board wanted a structure that could govern risk and explore opportunity at the same time, without conflating the two.

The design built in regular, structured consultation with the people AI decisions would affect most — staff working directly with young people, and the young people themselves. This was not a compliance exercise; it was recognition that the most useful AI governance intelligence would come from people close to the work.

What it cost

Canteen has not publicly reported the resource cost of the Sprint Committee. As a governance structure rather than a technology purchase, the main investment is staff time and board attention.

What the board did

The board created the committee as a board-linked governance and innovation body — close enough to the board to report genuinely on risk and opportunity, but managed by senior leadership with the operational knowledge to act. The board defined the committee's purpose, its time limit, and its accountability structure, then stepped back to let it function.

The twelve-month time limit was deliberate. It created urgency, limited open-ended commitment, and gave the organisation a natural review point: after a year, what had the committee learned, and what should change?

The outcome

The committee was in early operation at the time of publication. The design means Canteen goes into AI adoption with governance clarity most organisations lack: defined guardrails before tools are deployed, clear ownership of risk, and structured mechanisms for the people most affected by AI decisions to shape how those decisions are made.

Whether the Sprint Committee model achieves its aims will become clear over the twelve-month period. What is already clear is the choice of sequencing: governance first, then experimentation.

Lessons for directors

• A time-bound committee with a fixed remit can be more effective than a standing AI subcommittee, especially when the organisation is still learning what AI governance should look like.
• Building consultation with staff and beneficiaries into the governance structure — not as an add-on — is both ethically sound and practically useful. The people closest to the work often spot risk earlier.
• Linking governance and innovation in the same body reduces the risk of the board being presented with only the optimistic case for AI adoption.