Ruder Finn has transitioned its corporate website to an LLM-first interface, abandoning traditional static menus in favor of a conversational engine. By integrating Anthropic’s Claude models via Amazon Bedrock, the global communications agency is now requiring visitors to query its services and expertise through natural language dialogue.
The shift marks a departure from standard web architecture, replacing keyword-based searches with a Retrieval-Augmented Generation (RAG) system. This setup pulls information from a proprietary knowledge graph, which agency strategists manually curate to ensure responses remain accurate and grounded in verified data. While traditional web pages remain accessible for legacy navigation, the primary interaction model now prioritizes direct inquiry.
CEO Kathy Bloomgarden noted that the redesign reflects a fundamental change in user behavior, where audiences prefer asking questions to browsing. "People don't browse anymore, they ask," she said. The agency intends for the platform to serve as a proof of concept for the AI services it provides to clients, demonstrating a commitment to integrating technology directly into its own operational infrastructure.
To manage the inherent risks of generative AI, the system operates within a closed-loop environment. CTO Tejas Totade emphasized that the differentiator is the underlying data strategy and human oversight rather than the model itself. A dedicated moderation layer requires human approval for system updates and outputs, ensuring that client confidentiality and corporate compliance are maintained. The agency plans to further refine this by linking its content management system directly to the retrieval index, automating the flow of information between its operational databases and the public-facing interface.
Comments (0)
No comments yet. Be the first!