You'll Manage AI, Not the Other Way Around

Artificial intelligence (AI) has unfolded a world of unprecedented possibilities for the social sector. Your team can now use ChatGPT to write the first draft of a report, blog post, or outreach email to donors. Tools like Vista or Midjourney can help you create unique video and visual content for more impactful storytelling.

Across diverse mission areas, AI is accelerating programmatic goals in nonprofits — driving personalized learning support to improve educational outcomes, making wildfire risk predictions to enhance prevention efforts and proactive response, and analyzing mobile data to preempt and reduce military suicide.

But despite these advances, you’ve probably also noted significant barriers to digital transformation and AI adoption: from cost to ethical risks to organizational readiness. Many nonprofits are navigating these barriers today, including Polaris, a data-driven nonprofit dedicated to ending sex and labor trafficking and helping survivors reclaim their freedom, with the support of its philanthropic partner, the Patrick J. McGovern Foundation (PJMF), a 21st-century foundation bridging civil society with technology.

Together, we envision an AI future that integrates nonprofit wisdom with community resolve to tackle our greatest challenges. But we need participation across the sector. Based on our experiences, here are five key steps that can launch your organization ahead and make 2024 the year nonprofits harness AI for good.

1. Build AI Readiness Across The Organization Your organization’s leadership needs a clear vision of how and what it will take to apply AI in your unique context — from the time and resources required, to the talent that might not fit your compensation structure. These base-line questions about organizational readiness can help as you prepare to dive in.

You might also need to think through re-organizing and maximizing current staff expertise in different roles within your AI transformation. Transparent communication with your analysts and front-line workers can help reduce fears that AI will replace their jobs, while revealing how AI can surface new opportunities, such as enabling more targeted service to communities, extracting data-driven insights, and validating theories of change. It’s also helpful to create multi-disciplinary teams, who can bring wider-ranging insights into problems and solutions and support greater buy-in across your organization.

2. Make Strategic Decisions About Priorities AI gives you an overwhelming number of options — and yet not every problem should be solved with AI. Furthermore, AI models are not all the same, from machine learning and computer vision, to natural language processing and large language models, to deep learning. To optimize AI benefits for your organization, identify high-priority areas where a particular AI approach will be most helpful and where you have reliable datasets for building AI tools that complement your staff expertise.

Also consider the outcome of the tools you build and any potential impact to the vulnerable groups you serve. For example, you might have 10 years of historical data to which you can apply machine learning to develop a model validated by human expertise. But you might decide answering your client hotline with a ChatGPT-operated chatbot, a black-box technology that learns on its own, might present risks that you are not currently equipped to handle.

3. Experiment And Scale Up, Where Appropriate As with any other technological improvement, progress is rarely a straight line. Most AI projects never go beyond the experimentation stage, requiring patient investment in the process and the flexibility to innovate, fail, and pivot. Even when you create a successful model, you still need the staff, time, and resources to test and validate outcomes — before even thinking about scale-up. And just because something is successful does not mean you should scale it. Scale-up brings its own set of challenges, such as overall integration with data systems.

4. Invest In Ethical Practice And Culture As you incorporate AI, be ready to take steps (and sometimes steps back) to ensure new tools don’t compromise the data of those you serve and support. You might need internal and external safeguards to help staff and partners appropriately use new tools to analyze raw data. For example, the Counter Trafficking Data Collaborative uses AI to create synthetic datasets that can be shared with research partners without the privacy risks that accompany raw data. This approach enables analysts, academics, practitioners, and policymakers in 150+ countries and territories to leverage the power of multiple partners’ data.

5. Dream Beyond Incremental Changes — And Beyond AI You might be thinking of discrete AI models to fix specific data problems, which is a great place to start when resources and expertise are limited. However, thinking in specific, small projects can constrain your team and obscure potential innovation outside of the box. Stepping away from your current processes and ways of thinking can make room for transformative solutions that may even save costs in the long term.

Reviewing your systems, processes, and goals to see what bigger strides you can take opens new pathways to impact. But don’t forget that AI is just one tool — albeit a powerful one — that should be used alongside social science, statistical analysis, content expertise, and lived experience.

Partnership As Your Launching Pad As you consider the lessons above, you’re probably wondering how to resource these time- and money-intensive efforts. Through its proximity to wealth and power, philanthropy can help bridge the gap between technological progress and nonprofits as the key to community-driven change. Serving as an active partner, foundations can help nonprofits safely navigate the digital landscape, become tech-empowered visionaries, and leverage AI for good.

First, you can pursue traditional grant partnerships. With PJMF support, Polaris contracted a data scientist to run two AI pilot projects: one that employed machine learning to develop a labor exploitation risk classification model using data from the National Human Trafficking Hotline, and another that used natural language processing to enable staff to search case notes for emerging trends.

The latter found that despite prevalent social media narratives about trafficking abductions occurring in white vans, only 0.04% of victims in Trafficking Hotline situations experienced this form of abduction — a statistic that helps combat such myths around the major trends and warning signs in trafficking.

Those two projects built Polaris’s internal AI capacity, and prompted it to take a much more holistic approach to implementing AI across its data-to-analysis pipeline.

You can also leverage philanthropic expertise in the field. Besides funding, foundations can support AI-driven projects by providing technical support. For example, York University developed a safe water optimization tool for refugee camps in consultation with PJMF’s Data Solutions team, while Climate TRACE, a global partnership of climate actors, built a comprehensive and highly granular data set on greenhouse gas emissions through incentivized data sharing.

Similarly, you can seek out philanthropic initiatives like PJMF’s Data Practice Accelerator Program, which offers nonprofit hands-on training, webinars, and network building. Alumni can access expert feedback from a robust community of like-minded practitioners as they build an ecosystem of ethical policies for responsible data management. Girl Effect, an international nonprofit working to build trust and empower girls and women in Africa and Asia leveraged accelerator learnings to improve the quality and relevance of information provided to girls via their Big Sis chatbot. They continue to work with philanthropy to improve outcomes and protect users’ privacy and safety.

Finally, see if any technology companies have pro bono resources or grant programs that might be a fit for your organization. You can start with your own vendors, who might be able to provide credits, volunteer consulting, or even corporate funding.

The Path Forward Implemented well, AI presents unique opportunities to accelerate toward your goals. It can help you produce real-time, actionable insights and intelligence that improve response and delivery efforts. It has the potential to evolve the questions you ask your data and the depth of your findings to inform innovations in your approach.

When leveraged with other tools and human expertise — from lived experience to social science — and supported by philanthropy, AI has the potential to help you tackle a range of often-intersecting challenges — from trafficking to health inequities to climate change. And the solutions you design today can help build a more equitable, just, and sustainable tomorrow.