|
More Than Half of Nonprofits Use A.I., Google.org Survey Finds
Article Courtesy of the Chronicle of Philanthropy - Learn more and review free subscription offer at Philanthropy Today
Most nonprofits now use A.I. for some tasks, a milestone that comes just over a year since the technology became widely available.
The A.I. tool might be proofreading a web page or drafting a social media post. It might be fine-tuning a grant proposal or matching volunteers to projects. Or it might be coaching thousands of young people on how to craft the perfect cover letter, find a new career, or ace their next job interview.
In the case of CareerVillage, an education nonprofit, it might be all of the above.
“I can’t even begin to articulate how much time we’ve saved” through A.I. tools that “make every day just a little bit easier,” says Jared Chung, executive director of CareerVillage.
The average nonprofit now uses A.I. for at least some tasks, a milestone that comes just over a year since the rise of tools like ChatGPT catapulted the technology into the public imagination.
According to a survey of 4,600 nonprofits released today by Google.org, the charitable arm of the multinational tech company, more than half of nonprofits say at least some of their employees use generative A.I. daily, mostly for proposal writing or content creation. Yet, even as A.I. becomes more common in the workplace, many employees — and nonprofits — are just beginning to scratch the surface of what technology is capable of.
“A.I. will be transformative in the way that all of us live and work in our lifetimes,” says Annie Lewin, senior director of advocacy and Asia Pacific at Google.org, which also launched Thursday a $20 million accelerator for 21 tech-savvy nonprofits, including Career Village. Each will receive training, funding, and free consulting from experts at Google to help expand and refine their use of generative A.I. for efforts such as producing high-tech immigration legal resources and tools for combating disinformation. The 4,600 nonprofits surveyed, which represent 65 countries, are participants in the Google for Nonprofits program, which provides free tools for hundreds of thousands of nonprofits.
For nonprofits that are often cash-strapped and understaffed, A.I. has the potential to “address so many operational capacity challenges that we as a sector have been chipping away at for decades,” says Chung, whose organization has been dabbling with A.I. since 2022. “That’s just going to make the work of being at a nonprofit and helping people get better,” he says.
Focus on Productivity
If the accelerator participants represent the cutting edge of nonprofit A.I. adoption, most charities still see A.I. primarily as a productivity tool, especially for writing or research tasks that can be easily delegated to tools like OpenAI’s ChatGPT. They might use A.I. to help draft grant proposals or social media posts, but they’re unlikely to use it for more strategic mission-driven tasks, like analyzing which neighborhoods are most in need of services or building a chatbot to help train volunteers.
Four in five nonprofits surveyed said they thought that generative A.I. could help the communities they serve, but many cited significant barriers to deploying the technology to its fullest. More than three-fifths said a lack of familiarity with generative A.I. made it difficult to put it use, while around half said that finding enough funding and training for staff were major challenges.
With time and more experience with A.I.'s capabilities, the hope is that “more of these organizations will see opportunities for how these tools can actually transform how they deliver their work,” says Lewin.
“If we as a society are not deliberate about it, the sector that could benefit the most from this technology will be at the greatest risk of missing out,” says Lewin, who noted that A.I. could ultimately help ease the burden on some resource-strapped charities — if they’re able to access and understand the tools most applicable to their work.
Though some nonprofits — like those focused on education or science — reported higher rates of A.I. usage, over a third of those surveyed rated their understanding of A.I. as low.
Only one-quarter said they had “concrete ideas” for how generative A.I. could help their organization. Two in five said that none of their staff had training in A.I., while none of the 4,600 organizations surveyed said that the majority of their staff had been trained.
High-Impact Projects
That means more customized or advanced applications of A.I. — like helping high school students apply for financial aid or identifying low-income diaper deserts — may still be out of the question for most nonprofits.
“To really implement these high-impact programs — that touch so closely to vulnerable communities — it takes a lot more comfort with the technology itself and also the potential biases and risks,” says Vanessa Parli, director of research at the Stanford University Institute for Human-Centered Artificial Intelligence.
According to the institute’s research, nearly 80 percent of nonprofits still lack an organization-wide policy for A.I. usage, a foundational step for protecting data privacy, monitoring potential biases, and establishing norms around A.I. use in the workplace.
At the Kenya-based maternal health nonprofit Jacaranda Health, a participant in Google.org’s A.I. accelerator, the policy underlying much of the organization’s A.I. strategy is rather simple: “We collect as little data as possible,” says Jay Patel, head of technology at Jacaranda Health.
Though the organization uses an A.I.-trained helpdesk to triage and answer questions from expectant parents, it does not collect users’ personal information like names or addresses. Any answers generated directly by the A.I. also go through a human desk agent, who can check for quality and accuracy before reaching the parents themselves.
The group has released a large language model for its work, the first of its kind for Swahili speakers, as an open-access tool that other organizations can use for free.
“These communities are getting left behind in the arms race for A.I.,” says Patel, who hopes that other organizations can leverage the Swahili-speaking model to help their users “engage with A.I. tools in their own language in a more equitable way.”
| 
