Why Stripe Puts AI Agents in Isolated Cloud Sandboxes
Stripe engineering manager Sharadh Krishnamurthy explains why AI agents threaten core infrastructure and how sandboxes prevent total outages.
Stripe built a company brain: Meet Kai
With Claire Vo, Sharadh Krishnamurthy · Sunday, September 13, 2026
Sharadh Krishnamurthy, engineering manager at Stripe, joins Claire Vo to break down Kai, Stripe's internal AI agent used by over 10,000 employees weekly. He explains why Stripe decided to build rather than buy their company agent, how pre-existing developer and data infrastructure enabled rapid enterprise deployment, and how they use project-level governance, secure cloud sandboxes, and a self-serve skills platform to scale AI safely across both technical and non-technical teams.
Key takeaways
- Stripe deployed its internal AI agent, Kai, to more than 10,000 active employees every week. Read more →
- Over 10,000 Stripe employees use Kai every week, creating more than 2,000 custom skills by converting ad-hoc chat sessions into saved workflows. Read more →
- Over 10,000 Stripe employees use Kai, the company's internal AI agent, every week to run operational tasks and inspect internal data. Read more →
- Stripe launched the initial prototype of Kai, its internal AI agent, with 1.5 engineers in two weeks. Read more →
- Over 10,000 Stripe employees use Kai, the company internal AI agent, every week across technical and non-technical teams. Read more →
5 articles from this episode
Why Stripe Prunes AI Skills to Protect Model Quality
Stripe scaled Kai to 10,000 weekly users by letting staff build 2,000 skills. Here is how they prune dead tools to stop context rot.
Why Stripe Stops AI Agents from Writing SQL Directly
Stripe processes internal data requests across 10,000 employees using a 3-layer triage architecture that protects data warehouses from rogue agents.
Why Stripe Built an Internal AI Agent Instead of Buying One
Stripe built Kai with 1.5 engineers in two weeks. Today 10,000 employees use it weekly. Here is why buying off-the-shelf AI failed.
How Stripe Uses Project-Level Controls to Govern AI Agents
Sharadh Krishnamurthy explains how Stripe uses project-level tool policies and human-in-the-loop controls to govern 10,000 weekly AI agent users.
More How I AI episodes
- GPT-6 Astra blew away every one of my benchmarksSep 6 · 5 articles
- 7 Grok Bot agents I use every daySep 6 · 5 articles
- I built a Claude Cowork system that does a week of PM work in a daySep 6 · 5 articles
- How I manage 15 AI agents 24/7 as a solo founder | Ryan CarsonAug 30 · 6 articles