Elevating internal customer support at Thorn with AWS: A generative AI use case

Delivering top-notch internal customer support is crucial for nonprofit organizations. Efficiently addressing internal customer support tickets can profoundly impact an organization’s productivity and employee well-being, apart from their ability to focus on the mission at hand. To meet these demands, many nonprofits, including Thorn, are turning to innovative technology solutions. In this post, we’ll explore how Thorn used Amazon Web Services (AWS) in conjunction with generative artificial intelligence (AI) to revolutionize their internal customer support for organization-wide IT, security, and engineering.

Thorn is a registered 501(c)(3) organization that builds technology to defend children from sexual abuse. Their vision is to eliminate child sexual abuse material (CSAM) from the internet. Thorn’s tools have detected more than 5 million instances of CSAM on the internet. AWS has supported Thorn with financial and technical support since 2013. This support has helped Thorn build out Safer, a commercial product that helps organizations detect and report CSAM from their platforms.

Thorn’s executive team has kept a keen awareness of the advancements in generative AI because these advancements are already impacting how malicious actors create and share CSAM and exploit children on the internet. Thorn, in partnership with All Tech Is Human, launched Safety by Design principles for generative AI in April 2024. This is a voluntary initiative in which many leading tech companies, including AWS, committed to adopting principles to prevent the use of generative AI to further sexual harm against children.

Improving mission performance with generative AI

Simultaneously, Thorn has been investigating how they can implement generative AI to streamline internal operations so staff can be more effective and better serve their constituents. One particular focus area for Thorn’s infrastructure team has been internal customer support. This takes on a unique significance because ensuring that the team members have the tools and assistance they need to fulfill their mission is paramount. Thorn’s processes for requesting and receiving support through Slack and Jira have worked reasonably well. That said, the team is always on the lookout for ways to improve efficiency and support scaling.

In October 2023, the team responsible for IT, security, and platform engineering—Thorn’s Toolsmiths team—ran a two-day hackathon event to build out a generative AI solution to reduce team toil in answering repetitive technical questions on internal Slack and accelerate time to solution for their colleagues. The team worked with AWS solutions architects and templates to build out a Retrieval-Augmented Generation (RAG) pattern for a Slack chatbot. This architecture relied on Amazon Kendra, Amazon Simple Storage Service (Amazon S3), and Amazon Elastic Kubernetes Service (Amazon EKS) to send a user request to Amazon Bedrock. The team configured Amazon Bedrock to use Anthropic Claude to answer the user’s question, using only the knowledge base context provided by Amazon Kendra. It took Thorn only a few hours to get this proof of concept set up and integrated with their internal team on Slack.

Generative AI chatbot saves time

During the hackathon, Thorn’s Toolsmiths team began evaluating their solution using past questions from staff related to IT and infrastructure support. The team reviewed the results coming from the chatbot and were pleasantly surprised with the quality of responses. The Toolsmiths team assessed that they could save themselves from multiple interruptions in a week and improve response time for their users through automated bot responses to support questions with minimal additional cost. After the hackathon, the team allocated capacity to productionize what they had built and modernize it with the latest features of Amazon Bedrock, such as support for Amazon OpenSearch Serverless. This initial proof of concept provided the necessary impetus for Thorn to explore generative AI–led solutions to multiple other challenges.

“Amazon Bedrock enabled us to rapidly prototype and fine-tune a chatbot adept at addressing internal IT and engineering support queries. The initial success of this solution spurred us to progressively develop additional applications, harnessing generative AI capabilities to enhance productivity for our workforce,” said Jim Pitkow, Chief Technology Officer at Thorn. “This effort has enhanced our ability to support our team effectively, enabling us to further prioritize driving mission impact.”

The Thorn team has added a second Slack integration for summarizing long chat discussions using Amazon Bedrock and Anthropic Claude. In another example, Thorn is also using Amazon Bedrock and Anthropic Claude to create an internal web application for various roles to get familiar with using generative AI in their day-to-day work. All of these solutions help the Thorn team focus on the mission at hand and work more efficiently.

“Rolling out Amazon Bedrock alongside our existing AWS infrastructure has been a smooth process thanks to its native integration with services like AWS Identity and Access Management and OpenSearch. We were able to leverage our internal AWS expertise to stand up generative AI capabilities quickly while maintaining our standard security postures and operational practices,” said Peter Parente, director of infrastructure engineering at Thorn. “With a tiny team investment, we’ve learned how to fold large language models into familiar UX patterns for our colleagues and gained hands-on experience incorporating generative AI capabilities from AWS into cloud architectures.”

Conclusion

Thorn’s commitment to using generative AI and AWS demonstrates its dedication to elevating internal customer support in the nonprofit sector. Efficiency, responsiveness, and employee satisfaction all receive a substantial boost, enabling Thorn to better fulfill their critical mission of building technology to defend children from sexual abuse.

If your nonprofit organization is looking to enhance internal customer support, consider Thorn’s success story. For more information about how AWS supports nonprofits, visit AWS for Nonprofits.