AI in Grantmaking: Early Lessons from the Hilton Foundation
The world of philanthropy can be vast and complex, especially spanning numerous initiatives around the globe and with dozens of partners operating in multiple languages. Established in 1944, The Conrad N. Hilton Foundation has awarded more than $3.6 billion in grants, and paid out nearly $300 million in grants in 2024 alone.
In a guest expert session in the June 2025 Applied AI (XD131), Jeff Krentel, the Director of Strategy, Learning and Evaluation shared insights into leveraging AI to understand the progress and impact of over 800 active grantees worldwide, with diverse strategic initiatives in safe water, refugee support, early childhood development, and more.
To tackle this immense knowledge challenge and enhance efficiency, the Hilton Foundation partnered with Exchange Design, initially focusing on basic data tools to analyze grant investments, before expanding to chatbots and agent workflows to help quickly unlock valuable information often "locked up in folder structures and people's heads".
Unlocking Knowledge with Custom AI Chatbots
Starting in 2022, one of the first implementations was to build on top of internal libraries and dashboards by prototyping a “grounded” chatbot platform for Retrieval-Augmented Generation (RAG). Developed nearly a full year before the release of Microsoft Copilot, this innovative tool is designed to curate knowledge from a wide array of documents, including research and evaluation reports, strategy documents, and grantee reports. This chatbot been particularly instrumental during their strategy development phase, allowing them to rapidly understand their diverse portfolio much faster than traditional methods, such as assessing hundreds of grantee reports. The ability to quickly theme and understand what's happening across their initiatives marks a significant "before and after" compared to their previous processes.
Augmenting Human Work with Agentic AI
Beyond knowledge curation, the Hilton Foundation is also experimenting with agentic AI for workflow automations. The goal is not to replace human staff but to augment human work, speeding up initial processes so that program officers can devote more time to crucial community engagement and contextual understanding. For example, agentic AI is being tested to assist program officers with the creation of grant approval documents, potentially generating a "70% product" to get them started more efficiently. They are also exploring its use for processing grantee reports at a portfolio level, with the key principle of human judgment remaining in the loop.
Navigating Adoption, Data Privacy, and Organizational Oversight
The journey into AI has not been without its complexities. The team's response has been varied across organizational, departmental, and individual levels. While there is growing curiosity, initial adoption was slow due to factors like the novelty of the technology, concerns about job security, and the need for a shift in behavior.
On an organizational level, a cross-organizational working group was formed to develop an AI policy focused on safety and security, involving general counsel, grants management, and communications teams. Just like many organizations, this prudent, cautious approach led to limitations on the types of tools that could be used due to associated risks.
Data privacy is a paramount concern. The foundation classifies information into private, proprietary, and public categories. Strict guidelines prohibit the use of private data in AI tools. Due to security concerns, they lack access to publicly available tools like ChatGPT. Instead, they rely on Microsoft's Co-Pilot, which is becoming increasingly useful as part of their Microsoft ecosystem. Their approach to data privacy involves a contained database environment and close collaboration with their IT team to ensure security.
The Future of AI in Philanthropy: Learning, Evaluation, and Grantee Support
Looking ahead, a significant aspiration is to leverage AI to further the learning agenda by identifying trends in grantee reports and validating insights. While this is currently an aspiration, the Learning and Evaluation team is testing tools, in partnership with XD, with larger data sets using a combination of AI tools to gather accurate insights that can inform grantmaking and strategy. Building confidence in these tools among users is crucial for their effective integration into work and learning processes.
The foundation also hopes to extend its AI journey to support its grantees. While currently focused on "getting our own house in order," there's an interest in exploring how to assist partner organizations with their digital transformation and capacity building, especially given that technical assistance is a recurring request from grantees.
The Hilton Foundation's proactive exploration of AI demonstrates a clear vision for how technology can enhance the strategic impact of philanthropy. By embracing innovation while prioritizing safety, security, and human judgment, they are paving the way for a future where administrative burdens are lightened, and human creativity is amplified, ultimately leading to more effective and impactful grantmaking.
