AI for nonprofits.

Start with the mission, not the tool

Pick two or three real bottlenecks, e.g. grant writing, rostering, intake triage or translation, and judge any AI tool only against those.

Most stalled AI projects in the not-for-profit sector start with a product demo rather than a problem. Before you trial anything, write down two or three tasks that consume disproportionate staff time and carry low risk if a draft is imperfect: first drafts of grant acquittals, meeting notes, position descriptions, translating a service flyer, summarising long policy documents.

For each one, note who does it today, roughly how long it takes each month, and what 'good' looks like. That gives you a baseline you can actually measure against, and it keeps the conversation about service delivery rather than technology.

Claim the nonprofit programs you are entitled to

Google for Nonprofits, Microsoft's nonprofit grants and Canva for Nonprofits give most Australian NFPs enterprise-grade tooling at little or no cost.

Google for Nonprofits gives eligible Australian charities Google Workspace for Nonprofits, Google Ad Grants (up to $x0,000/month of search advertising) and the YouTube Nonprofit Program. In Australia, eligibility is validated through Percent (formerly TechSoup). Registered charities with an ABN and DGR or ACNC registration are the usual fit.

Microsoft offers donated and discounted Microsoft 365 licences for eligible nonprofits, which is the cleanest route to Copilot if your organisation already runs Outlook, Teams and SharePoint. Canva for Nonprofits covers design and simple video.

The reason to start here is not just cost. Tenanted, admin-managed accounts give you audit logs, data residency settings, single sign-on and the ability to turn features off, none of which you get when staff sign up individually with personal accounts.

Choose a small, practical toolset

Two or three tools used well beats ten trials. Favour tools that sit inside systems you already administer.

A workable starting stack for a mid-sized NFP is: one general assistant inside your existing suite (Microsoft Copilot or Gemini in Workspace), one meeting-notes tool with consent controls, and one narrow, purpose-built tool for a specific job — accessibility captioning, translation, or donor-data cleanup.

Ask three questions of every vendor: does our data train your models by default, can we turn that off, and where is the data stored? If the answer to the first is yes and the second is no, it is not suitable for client information.

Put light governance in before you scale

A one-page acceptable-use policy, a register of approved tools, and a rule that a human signs off anything client-facing.

Governance for a nonprofit does not need to look like a bank's. What it does need: a short written policy staff can actually read, a register of approved tools and what each may be used for, a clear rule that personal or health information never goes into an unapproved tool, and named accountability for review.

Australian NFPs should read this alongside the Privacy Act and the Australian Government's AI Ethics Principles, plus any funder or state-based requirements attached to your service contracts. If you handle NDIS, health or child-safety data, treat AI tools as you would any other third-party disclosure.

Design with the people you serve

Test outputs with the communities affected; language, reading level, cultural safety and accessibility all fail quietly.

AI systems fail unevenly. Generated text drifts to a higher reading level than most service communications should sit at, translations lose nuance in community languages, and image or voice tools can misrepresent culture. None of this shows up in a vendor benchmark.

Build a habit of checking outputs with the people who will receive them, and keep a plain-language route for anyone who does not want to interact with an automated system at all.

Measure the outcome, then decide

Review after 90 days against the baseline you wrote in step one, and be willing to stop.

Come back to the numbers you recorded before you started: hours spent, turnaround time, error and rework rates, staff confidence. If a tool has not moved any of them, retire it. Reporting a stopped pilot honestly is a governance strength, not a failure.

Common questions

Is AI safe for nonprofits to use with client data?

Only inside tools your organisation administers, with model training turned off and a documented policy. Never paste personal, health or child-safety information into a free external consumer tool outside your organisation.

Is Google for NFPs free in Australia?

Yes for eligible registered charities. It includes Google Workspace for Nonprofits, Google Ad Grants of up to $x0,000 per month in search advertising, and the YouTube Nonprofit Program. Eligibility in Australia is validated through Percent (formerly TechSoup).

Where should a small NFP start with AI?

With one low-risk, high-volume administrative task such as drafting grant acquittals or summarising long documents, using a tool inside a suite you already administer, plus a one-page acceptable-use policy.

Do we need an AI policy before we start?

Yes! At least one page: what tools are approved, what data must never go into them, and who signs off anything client-facing. A full framework can follow once you know what staff actually use.