Security shouldn't be the "waiting room" that keeps a great idea from becoming code.

Traditional security reviews often become a bottleneck not because of failure, but because of the sheer volume of engineering teams ideating simultaneously. At Twilio, we solved this by building Threat Model Assistant (TMA)—an in-house tool that decentralizes expertise and allows teams to bake in threat mitigation the moment they start designing. This session dives into the architecture of TMA, exploring how we moved beyond generic reviews to specialized assessments for Agentic AI, all while maintaining developer velocity.