In this case study, we’ll reflect on how the Patrick J. McGovern Foundation's Products & Services team baked social responsibility principles into the design and development of Grant Guardian.
<aside>
AI Journey Phase: Curiosity
Module: Social Responsibility
About the Patrick J. McGovern Foundation:
Domain: Philanthropy
Organization Size: 11-50
Region: Global
Website: www.mcgovern.org/
</aside>
Earlier this year, our team at the Patrick J. McGovern Foundation (PJMF) announced the launch of Grant Guardian, an AI-powered financial due diligence tool intended to help grantmakers streamline the due diligence process and focus on meaningful engagement with nonprofits. A central goal of our product development efforts is to share what we learn about AI and technology development more broadly. So as we did with our previous product, JournaPilot, we’ll use this post to give you a sense for how our Products and Services team took a principles-first approach when integrating social responsibility into Grant Guardian’s development process.
To help set the stage, let’s first share more about what Grant Guardian is and why we built it.
Conversations with our own colleagues and our peers in the philanthropic sector surfaced an interesting challenge in the grantmaking process: how to consistently and efficiently assess a nonprofit’s financial stability as part of a broader analysis of the organization’s eligibility for a grant. The existing practices we heard about were time-consuming and in some cases varied from program to program within the same institution. And this inefficiency had consequences for nonprofits as well, with different funders requesting different documents, or even asking organizations to fill out an entirely new form with the funder’s preferred financial indicators. This felt like a perfect use case for AI. With Grant Guardian, foundations receive sophisticated financial health analyses, while nonprofits can simply submit their existing financial documents without the task of filling out multiple custom forms. This streamlined approach helps both parties focus on what matters most — their shared mission of creating impact.
What’s happening under the hood? A large-language model — in this case, Anthropic’s Claude 3.5 — rapidly extracts data from financial documents like 990s and audited financial statements. Then, the tool analyzes a nonprofit’s financial stability based on the set of criteria defined by the foundation using the tool. These criteria — various indicators along with weights and thresholds — are fully customizable, allowing the grantmaking organization to tailor metrics specifically to their organization’s funding priorities. These metrics are combined to give a total financial score, alongside an AI-generated qualitative summary of the financial assessment.
A screenshot of a financial report generated by Grant Guardian.
As we embarked on building this product, our team knew how important it would be to bake our social responsibility principles into the design, development, and deployment processes. Five social responsibility principles were particularly relevant for Grant Guardian:
Anyone who has built digital products before can likely attest to the fact that these principles don’t magically happen on their own, they require a lot of intentionality and strategic planning. Let’s dive in and see how PJMF’s Products and Services team addressed each principle during the development of Grant Guardian.
Our human-centeredness principle focuses on two critical aspects: 1) engaging the communities we’re looking to serve throughout the development of the solution, and 2) keeping a human-in-the-loop at critical decision-making moments within the product, specifically as it pertains to reviewing the output of AI models.
To adhere to this principle, we sketched out a robust, multi-phased user testing plan and solicited feedback incrementally throughout the development of the product, from very early prototype right up to general availability (GA). Here are the stages of user feedback that we implemented: