The enterprise AI stack behind Stripe’s company brain “Kai”

By How I AI Podcast

Categories: AI, Product

Summary

Stripe built Kai, an internal company agent, not to deploy AI tools but to solve the harder problem: governance structures that let employees safely use AI at scale. The key insight: agents are 'creative at bringing your infra down,' so companies must implement tool policies, multi-layer triage systems, and bounded context engines to prevent rogue behavior and data leaks.

Key Takeaways

  1. Implement three-layer triage for agent workflows: first find existing reports, then query analytics layer, only then fall to data catalog queries. This prevents agents from brute-forcing queries that overload data warehouses.
  2. Design tool policies based on roles and data sensitivity. HR teams shouldn't restrict agents entirely—instead use policies that prevent specific failure modes like posting sensitive data to public documents.
  3. Build a bounded context engine that knows employee identity, org chart position, projects, and colleagues' interests. This context awareness enables agents to 'do the right thing' without explicit permission boundaries.
  4. Host company agents on secure, standard infrastructure behind existing security boundaries. Make the agent-building infrastructure benefit your core product—Stripe's Kai patterns become tools for Stripe users.
  5. Recognize that agent infrastructure multiplies failure modes and amplifies problems. Plan for agents going rogue and almost taking down core systems—governance and restrictions are features, not limitations.

Related topics

Transcript Excerpt

Agents are very creative at bringing your infra down. It turns out that agents just like dial up all your failure mode. It just multiplies the amplitude of problems you can get. There were agents that went rogue. There were agents that may have almost taken down core systems. One of the cool things about projects that can be very concrete for people is the idea of tool policies. Let's say you're a person on the HR team who's dealing with a bunch of sensitive information. You really don't want the agent to sort of go rogue and put that sensitive data into some public Google document that all Stripes can access, but you also don't want to tell them, oh, you can't use any tools because your workloads are too sensitive. >> I love this idea of this like three layer triage that a data agent can …

More from How I AI Podcast