Stripe publishes details of Kai, its internal knowledge AI platform

Stripe has published details of Kai, its internal knowledge AI platform, built to give non-engineers the leverage that coding agents gave software teams. Within two weeks of the April launch most of the company was using it; today 83% of employees are weekly active users, including nearly all of go-to-market.

Sales representatives, finance analysts and technical account managers had been left behind: no existing tool met Stripe's data security requirements or covered work such as querying warehouses or preparing compliance reviews.

Why knowledge work needs a different architecture

Coding tasks share a uniform shape — edit files, run tests, commit — so a single agent architecture works. Knowledge work is the opposite: every task needs different tools, different data and its own definition of done.

Before Kai, Stripe had two options: a NoCode Agent Builder that produced more than 4,000 agents, and coding agents, which raised security concerns. Three requirements followed: scale expertise without centralizing it, meet users wherever they work, and enforce guardrails that do not exist in code. One Stripe invariant holds that data from two unrelated customer contexts must never share a single analysis.

APIs, AgentStudio and the execution environment

Kai is a platform, not a product. Surface-agnostic APIs power a web application and a Slack integration; internal tools embed the same agent, and Chrome extensions surface it inside third-party web tools. AgentStudio lets domain owners build, test and monitor their skills and tools, with usage signals beside each asset.

The execution layer shares its core primitives — harness, sandbox, workflow orchestration and access control — with Stripe's product-facing agents. The harness is built on LangChain's deepagents and runs on Kubernetes with a per-session sandbox and virtual filesystem; one recent session reached 932 turns. Kai connects to more than 1,000 tools, from BI dashboards to Zoom and Google Workspace.

Impact and what comes next

New GTM hires use Kai 2.7 times more than other employees; account executives who use it show twice the sales activity, create 17% more opportunities, generate 26% more revenue opportunities and close 39% more deals. In total, Kai has shifted 25,000 hours a year from administrative work to revenue-generating work.

Stripe says it has not won yet. Planned work includes better state management, a loop where Kai proposes skill improvements for review, and collaboration tools for sharing context across sessions.