Nonprofit: the full guide

A nonprofit raises money, often with conditions, then spends it on programs and the staff and systems behind them. Most of the money is promised before the work is done, and much of it must be reported on afterward.

Five pools of money decide the year.

Individuals give most. Americans gave $617.2 billion to charity in 2025. Individuals gave $394.2 billion, or 64%. Foundations gave 19% and corporations 7%. Bequests made up 10%.

Government matters more than most donors think. Government money is about 20% of most nonprofits' revenue. About two in three nonprofits had at least one government grant or contract in the most recent data. Total government funding is around three times total foundation funding.

That money has become less certain. About one-third of nonprofit service providers had government funding cut, frozen or delayed in the first half of 2025.

Donations are not most of the money. Nonprofits get less than half of their total revenue from foundations and individual donors. The rest comes from government and from fees, tuition and sales.

Restricted and unrestricted money. Restricted money can be spent only on what the donor or funder named. Unrestricted money is spent where the organization decides.

Why overhead is hard to raise. Private and government funders commonly limit overhead spending to 10% to 15%. Federal awards let an organization without its own negotiated rate charge up to 15% for indirect costs. AI tools, training and the time to check AI's work usually land in overhead.

AI touches the cost of writing and reporting most, then donor communications and back-office work. It touches revenue through donor trust and funders' rules.

Donor letters and appeals: cheaper drafts. Donor communications are the most common use. In one survey, 62% of nonprofits use AI there. A draft takes minutes, but a person still has to check every name, gift amount and story.

Grant writing and reporting: faster, with new rules. Nearly 25% of nonprofits in a 2025 survey used AI for grant writing. Some funders now limit it. NIH does not treat applications substantially developed by AI as original.

Volunteers: better matching, if the data is clean. Tools can suggest which volunteer fits which shift. Anthropic reports a connector to Benevity that searches more than 2.4 million validated nonprofits to support volunteering and donation searches.

Programs and client services: the most promise and the most risk. AI can summarize case notes, translate materials and answer common questions. Each of those touches people's private details. Bias or discrimination in AI tools concerns 50% of nonprofit leaders.

Finance and the back office: quiet savings. About 44% of nonprofits in a 2025 survey use AI to automate routine tasks. Budgets, reconciliations and board minutes are early candidates, with a finance lead checking each result.

Time saved is not money saved. Faster drafts pay only if the time goes to programs, wins more grants, or lowers a cost.

Two patterns stand out. Use has spread faster than rules, so most nonprofits use AI and about half have no policy for it. And funders are starting to watch. One large federal funder has written limits, while almost no foundations have.

These are scenarios, not forecasts. Each could happen, and each has a signal you can watch.

Scenario 1: Funders require AI disclosure on applications. Foundations add a question asking how AI was used in a proposal or report. Assumes: funders see more AI-written applications and want to judge them fairly. Signal to watch: AI questions on the application forms of your largest funders. Last checked: October 2026. NIH already limits applications substantially developed by AI. About 98% of foundations have no policy on grantees' AI use yet.

Scenario 2: AI assistants answer donors' questions about where to give. Donors ask an AI assistant which local charity to support, and act on the answer. Assumes: donors trust those answers more than they do today. Signal to watch: the share of donors who say AI summaries guide their giving. Last checked: October 2026. That share was only 7% at the end of 2025. Donors were most influenced by charities' websites (54%) and third-party monitors (39%).

Scenario 3: Shared regional platforms take over volunteer matching. A few platforms match volunteers to shifts across many organizations in a region. Assumes: the platforms hold good data on volunteers, skills and shifts. Signal to watch: local volunteer centers and giving platforms adding AI matching. Last checked: October 2026. More than 75.7 million Americans volunteered through an organization in 2023. Over 13.4 million volunteered virtually or in hybrid roles.

All three reward the same preparation. Write a short AI rule you could show a funder. Keep your website's facts current and accurate. Keep your volunteer and program data clean and under your control.

Gain: small teams. More than 60% of nonprofits under $1 million are exploring AI for grant writing, donor outreach and admin. A tool that drafts in minutes matters most where one person does five jobs.

Gain: development teams. Donor letters, appeals and grant drafts are where most nonprofits use AI today.

Gain: clients, if the rules hold. Faster paperwork can mean more time with the people a nonprofit serves.

Lose: organizations that surprise donors. Most people say an unchecked AI image in an appeal would discourage them from giving.

Lose: organizations without rules. Client records pasted into the wrong tool, or invented figures sent to a funder, cost more trust than any saving.

Mixed: staff and volunteers. Routine writing shrinks. Checking, judgment and relationships grow. See the course on AI and your workforce.

Contested: does AI free staff for mission work, or add unfunded work? One side: in a 2026 survey, 79% of nonprofits reported small to moderate improvements, mostly faster drafts and time saved on routine tasks. The other side: 58% of nonprofit leaders say staff lack the expertise or capacity to learn AI, and about 90% of foundations give no AI support. Both can hold at once. Drafting time falls, while checking, training and policy work rise with no grant to pay for them.

Contested: do donors and funders trust nonprofits less when AI writes appeals and reports? One side: 55% of people say an AI-image appeal no staff member checked would discourage them from giving. One nonprofit leader told researchers they avoid AI in fundraising materials for fear of losing grants. The other side: in a 2024 study of 1,006 donors, 28.1% said they accept AI personalization if the organization is open about its practices. Millennials were the most positive, with 32% very positive and 25% positive about AI in a charity's materials. The evidence points to disclosure and a named person who checks.

The cost ranges are illustrative and will be replaced with reviewed figures. The order matters more than the amounts. Work that holds no client data comes first, because a mistake there is cheap to fix.

Read the free offer before you accept it. Google reports its Gemini app and Gemini Notebook at no cost for up to 2,000 users, with enterprise-grade data protections. More AI features come with paid plans, which Google discounts by 70% or more for nonprofits. Microsoft reports 75% off Microsoft 365 Business Premium and 15% off Copilot.

Most offers are discounts, with conditions. OpenAI reports ChatGPT Business Standard seats at $8 a user a month, billed annually. OpenAI says it does not train on Business or Enterprise data. Anthropic reports up to 75% off its Team and Enterprise plans. Google, Microsoft and OpenAI each ask you to confirm eligibility first.

One test before any contract. Ask the vendor to run its tool on last year's grant report and compare the result with what your team sent. Check every number.

Funders. One large federal funder already limits AI-written applications. Most foundations have no policy on grantees' AI use yet.

Software that holds your data. Fundraising databases and volunteer tools are adding AI, and connectors now link AI assistants to donor systems. Anthropic reports a connector to Blackbaud's fundraising and donor management tools. Ask each vendor where your donor and client data goes.

October 2026: first published.

July 2026 (reflected at launch): Giving USA reported $617.2 billion in 2025 giving, with the individual share down to 64%. Affects section 1.

February 2026 (reflected at launch): a survey of 346 nonprofits found 92% use AI and 7% report a major improvement. Affects sections 1, 3 and 5, and sets the guide's main conclusion: start with work that holds no client data.

September 2025 (reflected at launch): NIH's limits on applications substantially developed by AI took effect on September 25, 2025. Affects sections 2, 3 and 4.

September 2025 (reflected at launch): about 90% of foundations reported giving grantees no AI support. Affects sections 3 and 5.