AI agents seem to be springing up everywhere at an exhilarating pace, promising to carry out all kinds of tasks autonomously –– much like human users would. As exciting as this paradigm shift is, it also comes with a need for scalable and standardized integration infrastructure. But what does that look like when all of this is so new?
The reality is that most of today's AI agents authenticate using the same methods as traditional applications. They are stuck relying on outdated auth patterns such as OAuth, and dealing with brittle integrations that don’t meet modern security or usability expectations. The result? AI agents that remain shiny new toys that fail to fulfill their true problem-solving potential. Sagi will explore common security pitfalls that occur when building agentic AI, and discuss new groundbreaking solutions to overcome them.