How to Use AI for nonprofits: A 2026 Starter Guide

The Adoption-Impact Gap: Why 92% AI Use Produces 7% Transformation

Here's the paradox at the heart of nonprofit AI in 2026: 92% of organizations use AI in some capacity, yet only 7% report major strategic impact. Nearly everyone is doing something with AI. Almost no one is transforming how their organization works because of it.

The data behind that gap comes from the 2026 Nonprofit AI Adoption Report, which maps the sector onto a three-stage maturity ladder:

• 65% reactive and individual — one staff member using ChatGPT to draft donor emails or summarize meeting notes, with no shared process.
• 18% operational — AI use woven across team workflows, with some coordination and norms.
• 7% embedded — AI tied to organizational goals, budgets, and KPIs, treated as infrastructure rather than a side experiment.

Most organizations never climb past the first rung. Only 4% have documented, repeatable AI workflows, and 81% use AI individually and ad hoc. As Gabe Cooper, CEO of Virtuous, puts it, one person drafting an appeal with ChatGPT while the rest of the team stays buried in manual processes "isn't a strategy. It's a workaround."

Why does the gap persist? The root causes are structural, not technological.

Nearly half of nonprofits have no formal AI policy — other surveys put the figure as high as 86% of nonprofits. Without policy, there's no governance, no data boundaries, and no shared expectations. Meanwhile, shadow AI fills the vacuum: SSIR researchers note that 70% of professionals using generative AI do so without their boss's knowledge, and that imagining staff aren't using AI because guidelines don't exist "is foolish."

Training is the second hole. 60% of nonprofits lack in-house expertise to assess AI tools, and only 4% have AI-specific training budgets. Staff learn by trial and error, which produces small efficiency wins — faster drafts, quicker research — but never the compounding gains that come from redesigned workflows.

The third hole is documentation. When AI knowledge lives in one person's head rather than in written, repeatable processes, every departure resets progress to zero. That's also the point where AI stops being a chatbot trick and starts becoming automation: reusable, multi-step workflows that any team member can run. This is where agent skills — instruction packages that let AI agents handle tasks like donor-report drafting or data cleanup — become relevant, and why verifying them matters; published research found roughly 26% of community-published skills carry at least one vulnerability. SkillzBench is building a free, independent directory that security-scans, license-checks, and benchmarks agent skills, plus a free MCP connector to attach vetted skills to your agents.

So where does your organization sit on the ladder? If AI use is individual, unsanctioned, and undocumented, you're in the 65% — and the path upward starts with policy and training, not new tools.

Step One: Write Your AI Policy Before Scaling Your Use

Here's the paradox facing nonprofit leaders in 2026: nearly everyone is already using AI, but almost no one has written down the rules. 92% of nonprofits use AI in some capacity, yet between 50% and 86% have no formal AI policy at all, depending on which survey you trust.

The stakes are higher than they look. According to research from the Stanford Social Innovation Review, 70% of professionals who use generative AI do so without their boss's knowledge. As those authors bluntly put it, imagining your staff aren't using AI because no guidelines exist "is foolish." Your team is already pasting donor data into chatbots — the only question is whether you've set boundaries.

A one-page policy beats a perfect policy. The goal isn't legal exhaustiveness; it's giving staff clear, usable rules so shadow AI becomes sanctioned, governed AI. Keep it to a single page covering:

• Approved tools — which chatbots and platforms staff may use, and for what tasks.
• Data privacy rules — what may never go into an AI tool: donor financial details, beneficiary records, anything confidential.
• Human-in-the-loop requirements — who reviews AI output before it reaches donors, grantmakers, or the public.
• Donor disclosure — how and when you tell supporters that AI was involved.

That last item deserves special attention. Sector research shows 76% of donors believe nonprofits should disclose their AI use, yet only 26% of nonprofit professionals say their organization actually does. That 50-point gap is a trust liability sitting quietly in your donor file.

The privacy concern is not theoretical. The same research found 70% of nonprofit professionals worry about data privacy and security with AI — and those concerns shift from hypothetical to urgent the moment staff use tools without oversight. A policy converts that anxiety into procedure.

If your team is ready to move beyond ad-hoc chatbot use into automated workflows, the same governance logic applies: verify before you deploy. SkillzBench is building a free, independent directory of agent skills with security scanning and benchmarking — worth exploring when you're ready to equip agents that touch donor data.

Draft the policy this week. Circulate it, invite edits, and make it a living document. Governance isn't the step before AI adoption — it is the first step of AI adoption.

Start With Exquisite Pain Points: Proven Use Cases and Human-in-the-Loop Boundaries

The fastest way to waste an AI subscription is to point it at everything at once. Stanford Social Innovation Review's 8-step framework recommends the opposite: target what its authors call your "exquisite pain points" — the time-consuming, repetitive tasks that quietly consume your team's week.

The strongest documented use cases map neatly onto nonprofit pain. Fidelity Charitable's practitioner guidance and sector survey data point to six areas where AI consistently delivers:

• Grant writing — drafting, formatting, and tailoring proposals to funder requirements
• Donor communications and prospect research — 82% of fundraisers are comfortable using AI for donor research, though 63% remain unsure about generative AI for donor-facing messages
• Data analysis — 42% of nonprofits already use AI here, making it the most common analytical application
• Document management — the one area where Shannon Farley of Fast Forward says humans can genuinely step back
• Board automation — minutes (82%), document summaries (75%), and director prep (63%) benefit most

One boundary matters more than the rest. According to Candid's research, 23% of foundations will not accept grant applications created with generative AI, while 67% remain undecided. Never submit AI-drafted proposals without checking each funder's stated position — and without substantial human rewriting.

That's the essence of the "humans in the loop" boundary both frameworks converge on. Fidelity Charitable's guidance keeps people in charge of privacy, strategy, donor relationships, and final decisions. SSIR calls this "co-boting": augmenting staff rather than replacing them. As NTEN CEO Amy Sample Ward puts it, "tools should not make decisions."

Both frameworks also insist on piloting before scaling. Run one use case with one team for a defined period, measure the results, then expand. SSIR's authors are blunt about the alternative: 43% of professionals already use generative AI, and 70% do so without their boss's knowledge — so pretending staff aren't using it because no guidelines exist "is foolish."

Once a pilot proves itself, the natural next step is turning one-off prompts into repeatable workflows. Agent skills — reusable instruction packages for tools like Claude Code or Cursor — can automate multi-step tasks such as donor-report drafting or grant formatting. SkillzBench is building a free directory that security-scans and sandbox-benchmarks these skills before you trust them with sensitive donor data, which matters when published research found roughly a quarter of skills carry at least one vulnerability.

From Chatbot Drafts to Repeatable Workflows: Your 2026 Adoption Plan

Most nonprofits already use AI somewhere — the problem is that only 7% report major strategic impact, according to the 2026 Nonprofit AI Adoption Report. The difference between the 92% who dabble and the 7% who transform is a plan. Here's one you can start this month.

Step 1: Name a pilot owner. With 81% of nonprofits using AI individually and ad hoc, someone needs to own the shift from personal workaround to team workflow. Pick one person — a communications lead, a development manager — and give them authority to run a 30-day pilot on one "exquisite pain point": a repetitive, time-consuming task like donor email drafting or board meeting summaries.

Step 2: Write a one-page AI use policy first. Roughly half of nonprofits (some surveys put it as high as 86%) have no formal AI policy, and shadow AI is rampant — 70% of professionals who use generative AI do so without their boss's knowledge. Your policy should cover what data staff may paste into tools, which tools are approved, and when to disclose AI use to donors (76% of donors want that disclosure; only 26% of nonprofits provide it, per Nonprofit Tech for Good).

Step 3: Document the workflow and measure time saved. Only 4% of nonprofits have documented, repeatable AI workflows — that documentation is what separates operational use from one-off prompting. Write down each step of the pilot task, note where AI assists and where a human reviews, and log hours saved weekly. Keep humans in the loop on strategy, donor relationships, and final decisions; that's the consistent guidance from both SSIR's human-centered adoption framework and Fidelity Charitable's practitioners.

Step 4: Pick affordable entry points. Cost shouldn't block a pilot. Anthropic's Claude for Nonprofits program offers discounts of up to 75% on Team and Enterprise plans, connectors to tools like Blackbaud and Benevity, and a free AI Fluency for Nonprofits course built for non-technical staff.

Step 5: Expand into repeatable, automated workflows. Once a documented workflow proves out, agent skills — reusable instruction packages that automate multi-step tasks like donor-report drafting or data cleanup — are the natural next rung. Caution matters here: published research found roughly 26% of agent skills carry at least one vulnerability, which is why SkillzBench is building a free, independently verified skills directory with security scanning, version pinning, and sandbox benchmarking, plus a free MCP connector. Everything will be free, funded by ads around its blog rather than user fees.

If your pilot saves real hours, repeat it across team workflows — that's how the 18% of "operational" nonprofits climbed the ladder. Want first access to verified, safe agent skills for your workflows? Join the SkillzBench waitlist and be there when the directory and free MCP connector launch.