Nonprofit AI adoption is no longer the question. Nearly everyone is doing it. The question is why so few organizations are getting anything out of it.
A new report from the Blackbaud Institute, built on surveys of 1,389 social impact professionals and 1,034 donors conducted in March 2026 with Edge Research, puts numbers on something a lot of nonprofit leaders already suspect. Only about 10% of organizations have moved past experimentation into what the report calls AI-Adaptive. Roughly 75% are using AI in fragmented, individual ways, with limited organizational impact. The remaining 15% or so have not adopted or approved it at all.
That middle number is the one worth staring at. Three quarters of the sector has AI in the building. It just isn’t doing much.
What fragmented nonprofit AI adoption actually looks like
Picture a mid-size nonprofit. The development director drafts appeal copy in ChatGPT on a personal account. A program manager summarizes grant reports with a different tool. Someone in marketing uses AI for social captions and has never mentioned it to anyone. The database administrator has been asking for six months whether any of this is allowed and has not gotten an answer.
Nobody is doing anything wrong, exactly. Everyone is saving themselves a little time. And the organization captures almost none of that value, because the time saved is scattered across a dozen private workflows that never turn into shared process, better data, or a stronger donor experience.
This is what the report is describing when it separates AI-Emerging from AI-Adaptive organizations. It is not a story about who has the fancier tools. Both groups have the same tools. Access to AI has been effectively free for two years. What separates them is whether AI is governed, documented, and pointed at problems the organization actually cares about.
The performance spread is not subtle. Among AI-Adaptive organizations, 34% exceeded their revenue goals, compared with 17% of organizations not using AI. On efficiency, 81% of AI-Adaptive organizations reported both freeing up time and operating more efficiently. For organizations with no AI in use, those figures were 18% and 16%.
Doubling your odds of hitting a revenue goal is not a rounding error. But it does not come from buying a tool. It comes from closing two specific gaps.
Gap one: your data is not ready, and AI will not save you from that
The report calls it the Data-Readiness Gap, and it is the least glamorous finding in the whole thing. Fewer than 20% of respondents rate their organization’s data health as excellent. At AI-Adaptive organizations, that figure jumps to 38%.
The relationship runs in both directions. Organizations with clean, accurate, timely data can actually put AI to work on it, and organizations that put AI to work tend to clean up their data in the process. Organizations with a donor database full of duplicate records, dead emails, and inconsistent campaign tags get exactly what you would expect when they point a language model at it, which is confident nonsense at scale.
Most nonprofits underestimate how much of this problem lives on their website rather than in their CRM. Your forms are the intake valve for your data. If your donation form does not validate emails, if your event registration writes to a spreadsheet nobody syncs, if your GA4 setup has been quietly broken since someone rebuilt the site in 2023, then you are polluting the well every single day. No amount of AI cleans that up downstream.
The unglamorous work of fixing form logic, straightening out your analytics, and making sure your website actually writes clean records into the systems you use is the highest-leverage AI project most nonprofits could start this quarter. It just doesn’t feel like an AI project, which is exactly why it keeps getting skipped.
Gap two: donors want to be told, and almost nobody is telling them
Here is the finding that should make communications teams sit up. Of donors surveyed, 76% say it is important for organizations to clearly disclose when and how AI is used. Only 26% of professionals say their organization does this today.
That is a three-to-one mismatch between what donors expect and what the sector delivers, on a topic where getting caught is significantly worse than volunteering. Donors are not asking you to stop using AI. They are asking you to be straight with them about it, which is the same thing they have always asked about overhead ratios and program outcomes and where the money goes.
The organizations that are quiet about AI use are not being quiet strategically. They are quiet because nobody has written anything down. There is no policy to point to, no statement to publish, no agreed-upon answer for what happens when a major donor asks whether the thank-you note they received was written by a person.
Which brings us to the thing you can actually go do.
Building an AI policy for nonprofits that isn’t just a compliance document
An AI policy for nonprofits does not need to be twenty pages of legal hedging. The useful version is short, public, and specific enough that a staff member could read it on Monday and know what they are allowed to do on Tuesday.
Name the workflows you are using AI for
Not categories. Workflows. Drafting first-pass grant narratives. Summarizing program reports. Generating variations on email subject lines. Cleaning and deduplicating constituent records. Writing alt text. Three to five named workflows beat thirty vague ones, and they give you something to measure.
Say what AI never touches
Most organizations land somewhere similar here. AI does not write the final version of donor communications that go out under a person’s name without review. It does not make decisions about individual beneficiaries or grant recipients. It does not go anywhere near protected client data. Write your version down.
Require human review, and mean it
Among AI-Adaptive organizations, 60% require human review or oversight of AI-generated output before it gets used. That is the single most copyable practice in the entire report, and it costs nothing to adopt. It also happens to be the sentence that reassures donors most, so put it in your public statement.
Publish it where donors can find it
A short AI use statement on your website, linked from your privacy policy and your about page, puts you in the 26% instead of the 74%. It takes an afternoon to write and it turns a potential trust problem into a trust signal. If a donor ever asks, you send a link instead of scrambling.
Where the real leverage is
Once the data is trustworthy and the policy exists, nonprofit AI adoption stops being a productivity hack for individual staff and starts showing up in the numbers that matter.
The organizations pulling ahead are using AI to draft and test far more fundraising copy than a human team could produce alone, then letting real performance data pick the winner. They are using it to segment donors by actual behavior instead of the three tiers someone invented in 2019. They are using it to write the structured content and clear answers that get surfaced when someone asks an AI assistant which organizations in their city work on housing, which is quickly becoming a real channel and a real SEO discipline.
And they are using it, quietly and effectively, on the accessibility and performance work that most nonprofits never get around to, because it is exactly the kind of tedious, rules-based, high-volume task that machines are good at and humans avoid.
None of that requires a data science team. It requires someone to decide that AI is an organizational capability rather than a personal shortcut, and to spend a few weeks getting the foundations in order.
The honest version of the pitch
We build websites and run marketing programs for nonprofits, and we use AI heavily in how we do it — not as a novelty, but as part of how the work actually gets done. We’ve built our own tools and workflows to make it effective, with the same governance and human review we’re telling you to adopt. So take this with the appropriate grain of salt.
We’re not going to plug you into those systems — they’re ours, built for how we work. The point is different. Getting this right is hard, we’ve done the hard version on ourselves, and the thing we learned is transferable. The organizations getting real value are not the ones with the biggest tech budgets. They are the ones whose data is clean, whose site captures information properly, and whose leadership has said out loud what is and is not allowed. That’s a diagnosis and an organizing job before it’s a tooling job, and it’s the part most people skip.
That’s the work we can actually help with: looking at how your data flows, where AI would genuinely earn its keep in your specific operation, and how to get your systems in order so it has something trustworthy to run on. If your site is the thing standing between you and clean data — and for a lot of nonprofits it is — that’s a fixable problem, and Launchpad exists specifically because a lot of nonprofits need that fixed in weeks rather than quarters.
If you want to talk through where your data actually stands, and where AI would and wouldn’t help, before you point it at anything, we are happy to look.
Frequently asked questions about nonprofit AI adoption
What does AI-Adaptive mean in the Blackbaud report?
It describes the roughly 10% of social impact organizations that have moved past experimentation and treat AI as a governed organizational capability, with policies, oversight, and defined workflows. The report contrasts them with AI-Emerging organizations, where use is fragmented across individuals.
Do we need to tell donors we use AI?
Donors think so. In the survey, 76% said clear disclosure of when and how AI is used matters to them, while only 26% of professionals said their organization currently discloses. A short, public AI use statement is a cheap way to get on the right side of that.
What should an AI policy for nonprofits include?
At minimum, the specific workflows where AI is approved, the areas where it is prohibited (usually final donor communications without review, decisions about individuals, and protected data), a human review requirement, and a plain-language public statement. Short and specific beats long and vague.
Is our data good enough for AI?
Probably not, and that puts you in the majority. Fewer than 20% of organizations rate their data health as excellent. Start with the intake points, which usually means your website forms, your analytics setup, and how records flow into your CRM, before you worry about models.
Where should a nonprofit start with AI?
Pick two or three workflows with obvious volume and low risk, such as first-draft content, data cleanup, or alt text. Require human review. Write down what you did. Measure whether it saved time. Then expand.
