How to implement AI for your nonprofit

For nonprofit professionals, time is valuable. Especially when you’re juggling multiple roles or projects. Despite its limitations, AI can create tremendous efficiency in your work. Therefore, I recommend that all nonprofits use it to assist where it can help most (writing, analysis, note taking, routine tasks), which then allows you to spend your time on areas where it helps less (human connection).

What AI can help you do

The right way to think about AI

Think of AI as a highly skilled, brand-new assistant on your team. It can hand you a strong starting point in seconds, but it needs:

• Your direction.
• Your knowledge and expertise.
• Your final edit.

The goal isn’t to “hand off” your work. It’s to reach a solid draft faster, so you have more time and energy for the human parts of your job.

A simplified approach, but a great way to get started:

Step 1: A simple AI policy is better than none. Start by creating an AI policy today. The reality is that your staff are using AI to complete their work, whether they are authorized to do so or not. Therefore, to start, I suggest developing an AI policy for your entire organization. This policy creates the protective framework and guidance for all your employees. A simple AI policy is better than nothing, and to help you get started I developed an AI policy template that you can download below.

Step 2: Consider the scope. When it comes to official AI usage, I’d suggest starting with one team at a time. From experience, starting with a team such as fundraising, grants, or marketing is particularly helpful as these teams have a clear use case for AI (writing) along with looming deadlines that promote the desire for greater efficiency. In addition, these teams are typically handling less sensitive data than direct service team members, which helps mitigate risks when testing early.

Step 3: Choose your platform. Check out my detailed guide on which platform to choose for your nonprofit. The short of it: any platform can do a great job assisting you with your work. Whichever platform you choose, obtain enough licenses to ensure all selected users receive a license.

Step 4: Have authorized AI users sign an Ethics Pledge. All individuals that have been authorized to use AI should sign an Ethics Pledge. This creates clearer guidance and boundaries for their AI usage. See below for an AI Ethics Pledge template or what to include when creating your own.

Step 5: Train your team on fundamentals. Have teams use Projects, CustomGPTs, or Gems to get the most out of their AI usage. Check out more about how to use Projects in my Using AI for grant writing resource. Educate your staff on how to check AI-generated content for accuracy, including watching for biases and ensuring that outcomes and impact data come from a human. Also, teach them about the types of information that should never be input into AI.

Step 6: Share feedback and learnings. Have staff share their successes and what they’ve learned along the way. If you notice that some staff are using AI at a high level, highlight their usage and have them demonstrate to others. This is especially important as more users start using AI as they can learn from others who have found success.

Key components of an AI policy

• Data protection guardrails. Turn abstract “don’t share sensitive data” guidance into a clear framework that helps, regardless of the specific tool being used.
• A governance structure. For smaller nonprofits, the governance might land on one or two people. For larger organizations, spread governance across a diverse team. The goal is to ensure that someone is evaluating and tracking AI use and approving tools for your organization.
• An “approved-tools-only” approach with a clear approval process. Make it clear which AI tools are approved, especially for staff that are handling sensitive participant data. Create additional clarity on how an employee can secure access to approved AI tools.
• Mandatory transparency to managers about AI use. Requiring employees to disclose AI use in their work creates organizational awareness and a culture of openness. It also protects against the scenario where AI-generated content with errors gets treated as human-verified work.
• Halting any automated employment decision tools until further notice. This is especially important for nonprofits in workforce development, social services, or any space where hiring, screening, or promotion decisions affect vulnerable populations. Having this policy creates a general hold rather than a case-by-case judgment, which is the right thing to do at this time, given the evolving legal landscape around AI in employment.

Key benefits of an AI policy

• It can help protect your client, participant, and staff data. It is very common for nonprofits to store personal information about clients and participants, including housing status, immigration status, health information, employment history, incarceration history, and more. Therefore, an AI policy can help avoid the risks or mitigate the outcomes associated with a data leak or with sensitive information being entered somewhere it shouldn’t be.
• It gives your team permission to actually use AI instead of operating in a gray zone. Without a policy, what happens is one of two things (often both): staff use AI without direction or guidance; leadership sends vague emails restricting the usage of AI. By having a policy, people know what’s allowed, what’s not allowed, and how to request access.
• It positions you ahead of funder and regulatory expectations. Eventually, government agencies and foundations will require AI governance documentation from grantees. Having a policy in place means you’re ahead of the game and prepared for when that question gets asked in the next government RFP.
• It creates a way to experiment without overcommitting resources. The pilot framework helps you get started quickly and affordably. In addition, it allows your organization to discover innovations, new ways of working, and greater efficiencies.

Note that this AI policy is not a legal document and does not inherently offer legal protection or guidance. It is offered as a starting point for your organization to build on and adapt as needed.

When an individual is given official access to AI, have them sign an “AI Ethics Pledge.”

Key components of an Ethics Pledge

• A transparency requirement. The pledge gives clear guidance on appropriate AI usage that protects both the organization and the employee. Also, creating a culture of transparency helps with awareness in the event something goes wrong.
• A human-review obligation for all AI output. It’s important to frame the AI relationship correctly: AI is a tool that creates the first drafts, with a human required to review, edit, and refine before completion.
• Distributed accountability through department leads. Team supervisors should be responsible for collecting signed pledges and monitoring compliance for their own teams. This is especially important as supervisors are more involved in the daily work and are more realistic compliance checkpoints than the IT department.
• A commitment and active role in mitigating bias. Require employees to actively identify and mitigate biases in AI output. Nonprofits often serve marginalized communities where biased AI output can be detrimental to your final product.
• An anonymous reporting mechanism. Most AI policy violations won’t be done with malicious intent. Instead, it will be an honest mistake where someone didn’t realize they were feeding PII into AI. The anonymous reporting option keeps the door open for individuals to flag issues without feeling like they’re personally attacking a fellow employee.

Key benefits of an Ethics Pledge

• It converts organizational policy into personal commitment. The act of signing the ethics pledge creates a psychological shift: the employee moves from a general awareness of the rules to a more direct commitment that they agreed to and read the rules they’ve signed. A policy tells people what the organization expects; a pledge makes them confirm they understand and accept personal responsibility for meeting those expectations.
• It gives you a documented, defensible compliance trail. When a funder asks how you’re governing AI use, when a board member raises concerns about data security, or when an incident actually happens, you can point to signed, dated documents showing that every approved AI user acknowledged the rules before they had access.
• It creates a natural checkpoint for onboarding new users as your AI scales. As more teams get approved for AI tools, the pledge becomes a built-in piece of the onboarding process: before you get access, you read the expectations, you sign, your supervisor collects it. Paired with the policy’s training requirements, it gives you a repeatable system for growing AI adoption responsibly.

Note that this Ethics Pledge is not a legal document and does not inherently offer legal protection or guidance. It is offered as a starting point for your organization to build on and adapt as needed.

Offer the following guidance to your staff as they begin using AI.

• Ground it in your knowledge. This is your number-one defense, and the real reason Projects matter. When AI draws from Projects, it works from your real information instead of guessing. Learn more about Projects in the Using AI for grant writing resource.
• Know your subject. You must understand what you’re asking for. AI is the co-pilot; you’re the pilot. It works best when you can tell whether its output is correct. Be familiar with your organization’s programming, research around your work or clients, and trends in your local community. This familiarity will result in you extracting the best information from AI.
• Vet every source. AI can fabricate citations, studies, and statistics that look legitimate — real-sounding journals, plausible authors, even working-looking links. Never repeat anything from AI you haven’t confirmed at the original source. If AI has pulled something from your knowledge file such as a research statistic, double check the number to verify.
• Watch for bias and derogatory terminology. AI leans toward the majority viewpoint or terminology used in its training data, which can flatten nuance and underrepresent or stigmatize the communities you serve. Read with that lens and correct it in your instructions.
• Check the date. AI’s knowledge has a cut-off and may miss recent funders, events, research, or regulation changes. Verify anything time-sensitive against a current source.
• Ensure that outcomes and impact data come from a human. AI does not know what’s feasible for your program to accomplish over the next year. Never let AI estimate or guess what your outcomes should be — those figures can slip through, leaving you on the hook for accomplishing more than you can.

One accidental paste can send private information somewhere you can’t pull it back from. You don’t need to memorize a policy (though your organization should adopt one) — use one rule of thumb:

If it’s already on your public website — or if you’d be willing to add it to your website — it’s fine to share.

Never put into AI

• Personal information about anyone (participants, donors, board, staff): names, addresses, contact info, demographics, health or SUD status, conviction history, résumés
• Confidential funder relationships or anything under an NDA
• Private financial details beyond your public, audited statements
• Unannounced programs, innovations, or strategies that aren’t public yet
• Anything proprietary, or that you’d treat as a trade secret

When in doubt, leave it out — or anonymize it first.

Right now, your organization is learning to write grants with AI. Soon, so will everyone’s. On top of that, grantmakers are already beginning to use AI to help review and score proposals. We’re heading toward a world where AI helps write grants and AI helps review them.

The result: the differentiator shifts even more toward relationship building and human connections that AI cannot assist with. In fact, I’ve already seen funders move from a written proposal process to a video call process. I expect more of them to follow as AI becomes further embedded in this work.

The work and the world is evolving to adapt to AI’s presence. My advice: keep an ear to the ground and stay adaptable. Use AI where you can to advance your work and your mission.

Cold applications rarely win grants. With that, leverage AI to help you write grants faster and more efficiently. The time you save should shift toward building relationships and fostering connections. This is precisely what AI is worst at: empathy, trust, negotiation, and genuine relationship-building over time. So the human part of fundraising isn’t becoming less important — it’s becoming your edge.

Grantary can help track all outreach efforts and cultivation deliverables to ensure that you are being strategic in your connections. When you’re ultimately invited to apply for a grant, Grantary can store that information as well.

Find out more about Grantary. See how it works →

• Which AI should your nonprofit use? — platforms, nonprofit pricing, and whether to buy AI bundled into other software.
• How to use AI for grant writing — setting up a project and writing a grant using AI.
• How to use AI for grant prospecting — prospecting and researching funders using AI.

Do we need an AI policy if only one or two people are using AI?

I would strongly suggest having an AI policy regardless of the number of individuals who are officially authorized to use it. An AI policy offers protections and some basic guidance for employees to use in their AI practices.

What if our board or leadership doesn’t want us using AI at all?

People will find a way to use AI, regardless of whether a board or leadership team disavows it. If they’re blocked from using it on company Wi-Fi, they will use AI on their phones. As a member of the initial AI launch group for a very large national nonprofit, I have seen this firsthand. AI is here, and AI is here to stay. Therefore, I would strongly advise that the board and/or leadership team think through how to ethically and responsibly implement AI at their organization.

Do we have to tell funders we used AI?

Only if they directly ask. This is not an effort to hide anything, but grant applications really don’t offer the space or the question to respond to this fully. In my experience, only perhaps one in 20 funders will ask if AI was used in the development of a proposal, so it isn’t that prevalent a question at this time. That said, I do feel that funders can tell when AI is used, especially when it is used poorly. Therefore, I’d suggest looking at my AI for grant writing resource to learn how to best utilize AI when writing grants.

Which team should get AI access first?

I’d suggest that a non-direct service team doing something repetitive gets access first. By repetitive, I mean something like grant applications, social media posts, spreadsheet analysis, and so on. This minimizes your risk and reduces the likelihood that sensitive data is entered into AI. It also gives you time to figure out how it works, how the team uses it most effectively, and how to better implement it for future teams.

How much does it cost to roll AI out for a team?

It depends on the size of your team. Claude and ChatGPT offer a fairly generous nonprofit discount, which both begin at $8 per month per user. Therefore, for a team of 10, it would cost $80 per month. I dive deep into pricing on my Which AI should your nonprofit use? resource.

Can we use free AI accounts instead of paying?

You could. However, free accounts do not give access to the Projects, CustomGPTs, or Gems feature, which is the feature I recommend most for nonprofit organizations to utilize. In addition, free accounts do not provide the ability to restrict AI from being trained on the data you’re entering.

Is it safe to put our organization’s information into AI?

Yes, as long as you would put that same information onto your website. That’s my personal rule of thumb. If something is already on your website, or if it is something you’d be willing to put onto your website, then yes, it is safe to put into AI. If it’s information that you would not be willing to post publicly, then I do not recommend entering it into AI.

In the grant writing space, for the most part, the grants we’re writing contain information that’s already publicly available, so that information is safe to enter. This is also true for any work in communications, marketing, and beyond.

Where things start getting sensitive is client and participant information. I would 100% avoid putting any PII information into AI.

Who should own the AI policy at a small nonprofit?

It depends on how small your nonprofit is, and this answer, of course, varies depending on the complexities of every nonprofit organization. I’d recommend that one or two individuals own the AI policy directly, ensuring it’s up to date and that employees are made aware it exists and should be followed. With that said, it’s ultimately every leader’s responsibility to ensure that AI is being used ethically and responsibly in the organization. For example, if I’m supervising a team of grant writers, it is my responsibility to ensure they are using AI properly.