Boosting Fundraising with Hyper-personalization powered by Amazon Bedrock

Introduction

• The University of California Los Angeles (UCLA) Anderson School of Management, in partnership with Slalom Consulting and AWS, used AWS and Anthropic's Claude Sonet Large Language Model (LLM) to revolutionize their alumni engagement strategy.
• This project led to a significant increase in donation conversion (86%), donation size (26%), and overall dollars raised (132%).

Key Takeaways

AWS and Generative AI in Higher Education

• AWS's mission in higher education is to create an environment where institutions can thrive by building resilience and empowering innovation.
• AWS is seeing universities adopt generative AI capabilities to improve student success and organizational effectiveness, such as the UCLA Anderson project.
• Advancement offices play a vital role in higher education, as they are responsible for increasing philanthropy to close short-term operational costs and improve long-term institutional stability.

UCLA Anderson's Approach

• UCLA Anderson's challenges included engaging their 45,000 alumni and increasing fundraising efforts.
• The project was a collaborative effort between UCLA Anderson, UCLA's IT external affairs, and Slalom Consulting, enabled by AWS.
• The goal was to use generative AI to improve email subject lines, personalize content, and suggest appropriate donation amounts to boost alumni engagement and fundraising.

Technical Implementation

1. Model Selection: The team evaluated various models and settled on Claude 3 from Anthropic's Bedrock platform, as it provided a consistently good, human-like tone.
2. Prompt Engineering: The team spent significant time constraining the LLM's behavior and ensuring the output met their requirements, such as appropriate tone, personalization, and conciseness.
3. Validation: The team implemented a robust validation process, including both automated and human validation, to ensure the emails were high-quality and safe to send.
4. Technical Architecture: The solution integrated Bedrock, AWS services, and Salesforce to generate, validate, and send personalized fundraising emails to a sample of UCLA Anderson's alumni.

Results and Next Steps

• The initial test with 3,000 emails (1,500 AI-generated, 1,500 standard) showed promising results, with significant improvements in open rates, donation conversion, and overall dollars raised.
• The team plans to expand the use of generative AI in fundraising and explore other opportunities to leverage AI and machine learning to further enhance advancement efforts.

Conclusion

The UCLA Anderson project demonstrates the potential of generative AI to transform advancement and fundraising efforts in higher education. By seamlessly integrating customized content with donor data, the team was able to significantly boost engagement and fundraising results. This success highlights the broader opportunity for universities to leverage AI and machine learning to drive organizational effectiveness and support their core missions.