Stripe Engineers Boosted Output Tenfold via AI Agents

The firm reported a massive increase in code output through the use of internal agentic coding tools throughout 2025 and 2026.

Updated on Sept. 28, 2026 in Artificial Intelligence

Isometric editorial illustration of a structured stack of server hardware and conduit blocks, representing high-throughput software engineering infrastructure. — Stripe reported a ten-fold increase in engineering output throughout 2026, driven by the adoption of autonomous agentic AI tools for code development. AI Illustration. Upload story photo >

Live Poll

Do you believe AI tools will significantly increase your own productivity at work?

Stripe engineers achieved a ten-fold increase in output by using internal AI tools throughout early 2026. One engineer successfully merged over 600 pull requests during that six-month period, a significant shift from 2024 performance levels.

Why it matters

This shift illustrates how large-scale engineering organizations are restructuring team sizes and project timelines around agentic AI. The surge in velocity highlights the growing reliance on automated coding assistants to handle routine development tasks.

Engineers are leveraging Minions, a set of in-house agentic tools that currently generate 1,300 pull requests per week. Reliability is maintained through infrastructure that executes six billion test runs daily to validate every codebase change.

The players

Stripe

A global financial technology firm providing an API-first stack for payment processing and developer-focused infrastructure.

Patrick Collison

The CEO of Stripe who has overseen the integration of AI agentic tools into the firm's core development workflows.

The details

Stripe uses Minions, a suite of agentic tools—AI systems capable of executing multi-step tasks autonomously—to draft and refine code. To manage the volume of changes, the company employs a high-throughput reliability infrastructure that verifies each pull request against six billion automated test runs per day. These tools operate within data guardrails established in 2017 to ensure structural consistency across the repository.

2017: Stripe began implementing data guardrails.
2024: Average developer output was 2.3 pull requests per week.
2025: Production reached over 1,000 pull requests per day.
H1 2026: One engineer merged over 600 pull requests.

The Tech Race

Stripe's internal adoption of Minions represents a shift toward AI-native software development pipelines. This move places the firm among the first large-scale organizations to successfully integrate automated agents into production at this velocity.

This development signals a transition toward faster deployment cycles for complex financial infrastructure and services. Developers and teams using similar agentic workflows should anticipate increased pressure on automated testing infrastructure to handle higher volumes of machine-generated code.

The takeaway

The rapid acceleration of engineering output through AI suggests that manual code entry will increasingly be reserved for high-level architecture. Observers should track the rate of production incidents as organizations transition from human-written to agent-generated codebases.