Build your own company brain: the enterprise AI playbook from Stripe’s engineering team | Sharadh Krishnamurthy

Building internal AI agents requires rigorous governance structures and foundational infrastructure to ensure security and scalability. Stripe’s internal agent, Kai, demonstrates this by utilizing "projects" as a configuration layer to manage tool policies, model routing, and human-in-the-loop requirements. Rather than relying on generic tools, successful enterprise AI integration leverages existing data catalogs and resilient querying layers to provide context-aware, high-performance assistance. A critical component of this strategy is a "skill builder" platform, which enables employees to package and share automations, transforming individual workflows into repeatable, company-wide assets. By prioritizing platform-level investments—such as robust data observability and automated skill maintenance—organizations can empower teams to build bespoke agents that drive productivity without introducing chaos or excessive management overhead.