US study finds nonprofit AI adoption outpaces governance

Generative AI has become a routine workplace tool across the U.S. social-impact sector, but most nonprofits have yet to build the organizational systems needed to use it safely and consistently. A Blackbaud Institute study released June 24 and analyzed in a July 21 Candid report on nonprofit AI practices found that adoption is broad while governance, data readiness, transparency, and workforce preparation remain uneven.

The Blackbaud Institute and Edge Research surveyed 1,389 social-impact professionals and 1,034 donors in the United States in March 2026. Eighty-five percent of professionals said they use AI at work, and half said their organizations were using it more than a year earlier. Common applications include drafting communications, fundraising support, data work, reporting, and administrative tasks.

The study distinguishes between widespread use and effective organizational adoption. About 10% of participating organizations were classified as “AI-Adaptive,” meaning they had moved beyond experimentation toward systemic use supported by governance, data readiness, and transparency. Seventy-five percent were “AI-Emerging,” using AI in more fragmented ways. The results are U.S.-based and do not represent every nonprofit in Canada or Mexico, but they provide one of the clearest recent signals of where North American organizations are heading.

The study found that only about one-third of professionals believed their organization was using AI very effectively. Half of organizations used paid or enterprise versions of AI tools, while 24% relied exclusively on free versions. That split matters because unmanaged tools may offer fewer controls over retention, access, and the handling of sensitive information.

Formal policies are increasing but remain far from universal. Thirty percent of organizations reported having an AI policy in 2026, more than double the 14% reported in Blackbaud’s 2025 research, while another 37% said they planned to create one. The gap is especially visible in privacy and transparency: 76% of donors said organizations should clearly disclose when and how AI is used, but only 26% of professionals said their organizations do so.

The privacy mismatch is sharper still. Sixty-eight percent of donors said protecting sensitive personal data in AI use was very important, compared with 36% of organizations that said they were taking the appropriate steps. Donors also placed greater emphasis than nonprofits on human review, risk assessment, and disclosure. The implication is not that donors reject AI; rather, acceptance depends on whether organizations can explain the purpose, limits, and safeguards surrounding its use.

Workforce preparation is another weak point. A separate July survey of 512 U.S. HR professionals found that 62% of HR teams used AI regularly, but only 45% provided AI-literacy training to all employees. Seventy-eight percent reviewed AI outputs for bias, accuracy, or legal risk, yet only 39% had a formal review process. For nonprofits, that combination can leave staff responsible for decisions without consistent guidance on what to check or who is accountable.

The emerging lesson is that time savings alone do not demonstrate organizational value. Blackbaud’s research estimated average AI-related time savings at $503 per employee per week, rising to $621 among AI-Adaptive organizations. The difference was not simply heavier use; more mature organizations were more likely to reinvest saved capacity in fundraising, data quality, mission delivery, and donor relationships.

That points nonprofit boards and executives toward a practical agenda: inventory the tools already in use, classify what data may enter them, define human-review requirements, train staff, and publish a plain-language explanation of AI practices. Organizations also need to decide how they will measure whether AI improves service quality, financial resilience, or staff capacity rather than merely accelerating production. As Okta’s 2026 nonprofit report notes, the sector is adopting AI while facing a growing need to manage identity, access, and agent-related security risks. The governance question is no longer whether staff will encounter AI, but whether leaders will shape that use before an error, privacy incident, or loss of trust forces the issue.