How a nonprofit can use AI responsibly

Many Canadian nonprofits are already using AI, often without a decision ever being made. A program coordinator drafts a newsletter in a free chatbot. A development lead pastes last year's grant report into a personal account to get a first draft. None of that is unusual, and none of it is covered by a policy.

A January 2026 study by Imagine Canada, the Canadian Centre for Nonprofit Digital Resilience (CCNDR) and Microsoft Elevate found that about 80% of Canadian nonprofits reported using AI, with half using it in three or fewer activities. The same research found many organizations aware of reputational, legal and ethical risks. Staff in most organizations will use AI either way, so the decision for boards and executive directors is how to make that use something the organization can stand behind.

What are good first uses of AI for a nonprofit?

The best first uses are tasks where a staff member reviews every output and no personal information about clients or donors is needed. These carry little risk and save real time:

• First drafts of newsletters, social posts, volunteer emails and event descriptions.
• Summarizing long documents, such as board packs, policy consultations or funder guidelines.
• First drafts of grant report sections, built from your own program notes and outcome data.
• Translating plain-language materials into French or other languages your community uses, followed by review from a fluent speaker.
• Volunteer and member FAQs drawn from your existing handbook.
• Meeting notes and action lists from staff meetings, where everyone present has agreed.

A useful test: if the output were wrong, would a staff member catch it before anyone outside the organization saw it? If yes, the task is a reasonable place to start.

What should stay human-led?

Anything that decides what happens to a person should stay with a person. That includes eligibility for services, intake assessments, housing or funding allocations, risk ratings and case notes about individual clients. AI can help a caseworker find a policy or format a report. It should not be the one making the call.

Extra care applies to health information, information about children, and anything concerning people in vulnerable situations, such as those fleeing violence or with uncertain immigration status. For this kind of data, the safe default is to keep it out of general AI tools completely unless your privacy lead or lawyer has reviewed the specific tool and use.

How do we protect donor and client data when using AI?

Protect it by deciding in advance which data can go into which tool, and by using accounts your organization controls. Personal and free accounts often have weaker data protections than business or nonprofit plans, and the organization has no visibility into what was shared.

The federal and provincial privacy regulators' principles for generative AI (December 2023) set out what they expect of organizations using these tools. In plain terms: remove names and identifying details from prompts where you can, only enter personal information where you are authorized to, check outputs for accuracy, tell people when they are dealing with an AI tool, and remember that accountability for decisions stays with your organization.

Which privacy law applies depends on what you do and where:

• PIPEDA, the federal private-sector law, does not apply to nonprofits automatically. The Office of the Privacy Commissioner says it can apply when a nonprofit carries out commercial activity, such as selling services. Many nonprofits follow its principles anyway as good practice.
• Nova Scotia's PIIDPA limits storing or accessing personal information outside Canada for public bodies and the service providers acting for them. If you deliver programs under contract to a Nova Scotia public body, check your agreement, because many AI tools process data outside Canada. The province's PIIDPA FAQ explains the basics.
• Quebec's Law 25 applies if you handle personal information of Quebec residents, and has specific rules on automated decisions and transfers outside Quebec.

This is general information, not legal advice. For health, legal or financial client data, talk to a privacy lawyer before choosing a tool. Our article on putting client information into ChatGPT in Canada covers the same ground for private businesses.

Should we tell funders and clients we use AI?

Yes, in proportion to how it was used. Clients should know when they are talking to an AI tool rather than a person, for example on a website chat. Funders increasingly ask about AI in applications and reports, and a short, accurate statement is better than an awkward question later. Something like: "We used an AI writing tool to help draft sections of this report. All figures and outcomes were compiled and checked by our program staff." Check each funder's guidelines, since some have their own rules on AI-written applications.

A one-page AI policy for a small nonprofit

A small nonprofit does not need a long AI policy. One page that staff will actually read covers it:

1. Approved tools: which tools and accounts staff may use for work, and that personal accounts are not to be used for organizational data.
2. Data rules: three levels. Fine to use (public information, your own published materials). Remove identifying details first (internal notes, aggregated program data). Never (client case files, health information, donor financial details).
3. Human review: a person checks every AI output before it is sent, published or used in a decision.
4. Decisions about people: AI does not decide eligibility, services or allocations.
5. Disclosure: when you tell clients, funders and the public that AI was used.
6. Owner and review date: a named staff member responsible for the policy, reviewed at least once a year.

Are there discounted AI tools for nonprofits?

Several vendors offer nonprofit pricing, but terms change often and eligibility varies by country, so check each vendor's nonprofit page before budgeting. As of November 2026, Google says its no-cost Workspace for Nonprofits plan includes the Gemini app, with discounts of up to 75% on paid upgrades that add more AI features. Organizations with an eligible Microsoft 365 subscription have access to Microsoft 365 Copilot Chat at no additional cost. Start by checking what your existing licences already include before adding another subscription.

Canadian resources for nonprofits

CCNDR runs a Nonprofit AI Impact Hub with bilingual learning resources on responsible use, tools and case studies. Imagine Canada publishes sector research. In Nova Scotia, check Digital Nova Scotia's events calendar for AI sessions open to nonprofits.

Example: an adult literacy nonprofit with 12 staff

This is an illustrative example, not a client case. A Cape Breton adult literacy organization runs tutoring programs funded by provincial and foundation grants. Staff spend many hours each quarter on funder reports and on rewriting the same program descriptions for different applications.

The ED proposes a 60-day pilot to the board. Staff use the AI included in the organization's existing Microsoft 365 licences, not personal accounts. The pilot covers three tasks: drafting grant report narratives from the program coordinator's notes and anonymized attendance totals, summarizing funder guidelines, and producing plain-language versions of learner materials, checked by a tutor. Learner files, intake assessments and anything with a learner's name stay out. At day 60, the ED reports to the board on hours saved, what went wrong, and whether to continue.

A 60-day start plan

1. Week 1 to 2: Ask staff what they already use AI for. You will likely find more than you expect. Pick one approved tool, ideally one you already pay for.
2. Week 2 to 3: Adopt the one-page policy and get board sign-off.
3. Week 3 to 8: Run a pilot on two or three low-risk tasks. Record time spent before and after.
4. Week 9: Report to the board: what you tried, time saved, problems, and the next step. Our guide on measuring whether an AI project worked has a simple review format.

If your team wants help with tool setup, the policy and hands-on training, see our AI adoption and training service.