AI EQUITY PROJECT 2025
From cautious curiosity to active experimentation
In 2025, AI was no longer theoretical for the nonprofit sector. Usage surged, familiarity grew, and organizations began asking harder questions about leadership, governance, funding, data stewardship, and responsible use.
850
Organizations
76%
Using AI
80%
Familiar with AI
15%
Had an AI policy
CONTEXT
The shift from 2024 to 2025
In 2024, the sector was still trying to understand what AI meant for nonprofit work. By 2025, the conversation had changed. Organizations were experimenting even when they did not feel fully ready, and the questions became more practical: How do we train staff? Who owns the decisions? What should an AI policy contain? How do we fund readiness?
How do we use AI without leaving equity behind?
We think of 2025 as the design year. The sector had moved beyond awareness-raising, but the infrastructure needed for responsible adoption was still uneven. The challenge was no longer only whether nonprofits would use AI. It was whether they could build the conditions to use it well.
THE DATA
2025 at a glance
INDICATOR | 2024 | 2025 |
|---|---|---|
Organizations participating | 708 | 850 |
Using AI | 59% | 76% |
Feeling ready for AI adoption | 27% | 35% |
Familiar with AI | 53% | 80% |
Familiar with AI bias | 44% | 64% |
Familiar with Data Equity | 50% | 58% |
Have Data Equity practices | 46% | 36% |
Have an AI policy | 6% | 15% |
The most striking pattern was not simply that AI use increased. It was that adoption accelerated much faster than the organizational systems needed to guide it.
FINDINGS
What did we learn?
01
Navigating readiness in a complex landscape
Organizations were increasingly learning while adopting. AI use climbed to 76%, but only 35% described themselves as ready for adoption on the project’s year-over-year measure. Leadership uncertainty also showed up repeatedly: organizations that named fear or misunderstanding among leaders tended to report lower readiness.
This made readiness feel less like a technology checklist and more like an organizational condition. Confidence, leadership, staff learning, internal champions, time, and shared expectations all shaped whether experimentation could become responsible practice.
02
A disconnect between AI purpose and funding language
Nonprofits were trying to fund AI-related work while still learning how to describe the work itself. Respondents frequently named difficulty identifying or pitching clear AI use cases, while technical assistance and operating support were more attractive than narrowly defined technology investments.
At the same time, relatively few organizations prioritized AI governance policy development as a funding need—even while concerns about bias, ethics, and harm were widespread. The data suggested that many organizations understood the risks but had not yet connected those risks to the infrastructure that governance requires.
03
Data stewardship without enough infrastructure
The 2025 data exposed a basic tension: organizations were being asked to make more sophisticated decisions about AI while many were still working with uneven data systems. Small and medium-sized organizations frequently described unclear processes or data stored on individual computers, and several sectors serving marginalized communities reported uncertainty about their own Data Equity practices.
This matters because responsible AI does not begin with the AI tool. It begins with the data, processes, permissions, capacity, and values an organization already has.
04
Fear, curiosity, and dreaming existed at the same time
Open-ended responses were filled with concern about bias, misinformation, privacy, and data protection—but also curiosity about what AI could make possible. Organizations serving BIPOC communities and people with disabilities placed stronger emphasis on keeping data safe, while Canadian respondents more often framed their questions around community-centered possibilities.
The sector was not simply “pro-AI” or “anti-AI.” It was holding hope and caution at the same time. That tension is useful. Responsible adoption requires enough imagination to explore—and enough caution to refuse what does not fit.
EQUITY
The equity signal we could already see
In 2024, familiarity with AI was higher than familiarity with AI bias, while awareness of Data Equity was only beginning to translate into consistent practice. The project’s first AI Caution Ratio—measuring how many people familiar with AI were also familiar with its risks and biases—was 83%.
Smaller organizations often showed strong curiosity but weaker readiness, infrastructure, and governance capacity.
People from historically marginalized communities reported higher awareness of AI bias, even when their organizations had limited resources to act on those concerns.
Organizations serving communities already exposed to inequity were often carrying some of the strongest concerns about data protection and potential harm.
That is why AI equity cannot be treated as a side conversation about ethics after adoption. The people most able to see potential harm also need power, resources, and decision-making space to shape what happens next.
CUSTOM INDICATORS
What changed in the project's indicators?
CUSTOM INDICATOR | 2024 | 2025 | WHAT IT SIGGESTED |
|---|---|---|---|
AI Caution Ratio | 83% | 80% | AI familiarity rose faster than awareness of bias. |
Policy Preparedness Ratio | 10% | 20% | More AI users had policies—but from a very low base. |
Readiness Conversion Ratio | 46% | 46% | Readiness did not grow faster than usage. |
Equity Practice Ratio | 92% | 62% | Data Equity familiarity rose while practice fell. |
AI Equity Awareness Index | 49% | 67% | Overall awareness of AI, bias, and Data Equity increased. |
A DELIBERATE CHOICE
Why we chose not to create an AI Equity Score
In the first year, we explored whether a single AI Equity Score could summarize readiness, equity alignment, and responsible use. In 2025, we chose not to pursue it.
Equity is too contextual to flatten into one number. An organization can be highly confident and still lack community voice. It can have a policy and still lack accountability. A score could reward visible activity without showing whether the underlying practice is thoughtful, participatory, or just.
So instead of asking organizations to rank themselves, the AI Equity Project offers patterns and questions that help them reflect.
Readiness is not just a mindset. It is leadership, time, training, governance, resources, and the ability to act on what you learn.
SUPPORT
What the sector was asking for
Across 2024 and 2025, the top three supports remained remarkably consistent: leadership support, staff training, and technology tools. But the 2025 responses made the request more specific. Nonprofits wanted practical learning, technical assistance, and space to build internal capacity—not simply another tool license dropped into the organization.
The emerging message was clear: if the sector wanted responsible AI adoption, it would need to fund the human and organizational work around the technology.
CONTINUE THE STORY
Toward accountability
2025 moved the AI Equity Project from a baseline into a multi-year body of learning. By 2026, AI use would become widespread enough that the central questions shifted again—from adoption and readiness toward accountability, impact, funding, community voice, and power.