Ask a nonprofit executive director whether the organization uses AI, and the answer is often “not really.” Ask the staff, and you get a different picture. A program manager drafts grant reports in a free chatbot. Someone in development pastes a donor list into a tool to write thank-you letters. The accounting system now suggests how to code transactions and which bank lines to match. Nobody decided any of this. It simply arrived.
The research backs up what we see in the field. The 2026 State of Nonprofit AI report from NTEN and The Bridgespan Group found that 98% of the 917 organizations surveyed use AI in some form, yet only 32% of staff have had formal AI training, and fewer than half of executives have written guidance on what data can go into these tools.
Closing that gap doesn’t take new software. It takes oversight, and oversight is the board’s job.
Where AI is already inside your organization
Before deciding what to do about AI, it helps to know where it already is. In most nonprofits we work with, it shows up in three places:
- Staff accounts. Free and personal versions of AI tools are the most common entry point. Depending on the tool and its settings, what staff type may be used to train the vendor’s models. Business versions generally exclude your data from training by default, but only if you are actually on one.
- Your accounting system. QuickBooks Online now includes Intuit Assist and AI agents that suggest categories, help with reconciliations and flag unusual activity. Sage Intacct includes Sage Copilot. These features touch your general ledger, often without anyone deciding to turn them on.
- Donor and fundraising platforms. Blackbaud and most donor CRMs have added AI that drafts appeals, scores donors and suggests ask amounts, using the most sensitive data you hold.
So the real question is not whether your organization should adopt AI. It is whether anyone is overseeing the AI you already have.
Why this belongs on the board’s agenda
Boards are responsible for internal control, donor trust and compliance. AI now touches all three.
- Federal awards. If you receive federal funding, the Uniform Guidance (2 CFR 200.303) requires effective internal control over the award and reasonable measures to safeguard sensitive information, including personally identifiable information. An AI tool that sees beneficiary data or posts entries to grant-funded accounts is part of that control environment.
- Funders. Grantmakers now publish their own AI rules. NIH will not accept applications substantially developed by AI. NSF allows AI in proposal preparation but requires disclosure of how it was used. Private funders range from prohibition to silence.
- Form 990 and public reporting. In April 2026, Treasury announced a Form 990 transparency initiative focused on government grants, government contracts and fiscal sponsorships. Proposed rules and a comment period come first, and no effective year has been set, but the direction is clear: cleaner tracking of where funds come from and how they are used. AI tools that summarize or categorize that data need a human check.
- Your own advisors. In June 2026, the IRS Office of Professional Responsibility issued guidance confirming that tax professionals who use AI remain fully responsible for verifying the work and protecting client data. It is a fair standard to expect from anyone who prepares your Form 990.
A one-page AI policy is enough to start
You do not need a long document. You need clear answers to five questions, written down and approved:
- Which tools are approved? Name them, and require business or enterprise accounts rather than personal logins.
- What data is off-limits? Donor and beneficiary personal information, payroll and HR records, and anything a grant agreement restricts should never go into an unapproved tool.
- Who reviews AI output before it is used? Anything that leaves the finance office, goes to the board or goes to a funder gets reviewed by a named person.
- Who can let AI take action? If a tool can post entries, send email or approve payments, its permissions should be set and limited like any other user’s.
- Who owns the policy? One person keeps it current and reports to the finance or audit committee at least once a year.
If you apply for grants, add a short standard statement describing how your organization uses AI, so every application gets the same accurate answer.
In the finance office: let AI draft, never decide
Used well, AI saves a finance team real time. Good places to start are tasks where a mistake is easy to catch:
- First drafts of month-end commentary and budget-to-actual explanations
- Plain-language summaries of the financial statements for the board
- Pulling key terms, reporting dates and restrictions out of long grant agreements
- Flagging unusual transactions for a person to look into
Where AI should not have the final word: the numbers themselves, functional expense allocations, net asset classifications, and the governance answers on your Form 990. Those come from your books, your records and your judgment.
The rule we apply is simple. AI can prepare a draft. A person who understands the programs and the money signs off on it, and there is a record of who did.
What to bring to your next finance or audit committee meeting
- A short inventory of where AI is in use today, including features inside your accounting and donor systems
- A draft of your one-page AI policy
- The list of data that staff may never put into an unapproved tool
- Who reviews AI-assisted work in accounting, and how that review is documented
- Your standard AI disclosure statement for funders
None of this requires new software or a large budget. It requires someone to ask the question, and a board that expects an answer.
Want a second set of eyes?
Mandala CPA provides outsourced accounting to nonprofit organizations across the United States: month-end close, financial reporting, Uniform Guidance compliance and single audit preparation, on a single monthly fee. We can help you see which AI features are already switched on in your accounting system, draft a one-page policy your board can approve, and set up a review process your auditors will recognize.
Sources
NTEN and The Bridgespan Group, 2026 State of Nonprofit AI: Adoption and Governance.
2 CFR 200.303, Internal controls (Uniform Guidance).
NIH NOT-OD-25-132; NSF Proposal & Award Policies and Procedures Guide (PAPPG).
U.S. Department of the Treasury, “Treasury Announces Form 990 Transparency Initiative,” April 23, 2026.
Journal of Accountancy, “IRS outlines AI risks, Circular 230 duties for tax practitioners,” June 2026.
This article is general information and is not accounting, tax or legal advice for any specific organization.
