Designing AI That Connects

Nathan Enderle from Domo emphasized the critical importance of engineering reliability into AI solutions, ensuring outcomes are dependable and minimizing risks. He stressed the need for careful planning, a deep understanding of AI’s limitations, and a clear roadmap before deployment. AI agents, he noted, are best suited for automating high-volume manual tasks and non-value-add activities to boost productivity in core areas. Enderle highlighted the necessity of observability and audit trails for continuous improvement, along with the value of integrating AI seamlessly into existing user workflows to avoid unnecessary complexity—famously stating, “Don’t introduce AI to add extra failure points if you don’t have to.” He showcased a real-world example through Domo’s “AI Sidekick,” developed in just four weeks and iteratively improved through user feedback. Flexibility and customizability were also key themes, with action buttons illustrating how AI systems can be tailored to unique business needs. He concluded by encouraging early adoption and thoughtful implementation, offering practical insights for building trustworthy and impactful AI solutions.