How Dropbox Built an AI-Native Development Workflow

Dropbox engineers now accept more than one million lines of AI-suggested code each month. This case study explores how they achieved widespread adoption and transformed their software development lifecycle.

interface GreetingProps {
  name: string;
  age?: number;
}

export function Greeting({ name, age }: GreetingProps) {
  return (
    <div>
      <h1>Hello, {name}!</h1>
      {age && <p>You are {age} years old.</p>}
    </div>
  );
}

const x = 234;

console.log(x)

The Catalyst: Speed as Competitive Advantage

Dropbox CTO Ali Dasdan emphasized that "Speed is the only advantage of any company," noting that slow AI adoption posed greater risk than rapid implementation. The team required an AI system capable of reasoning across the entire codebase with widespread adoption to achieve velocity goals.

Organic Adoption

By 2024, Dropbox engineers began experimenting with AI coding tools informally through Slack and internal documentation. Leadership identified this grassroots activity and created AI champion groups to amplify adoption while removing friction.

Leaders Try the Tools

During April 2025's company-wide hackathon, Dasdan completed his "smart finder" project in approximately two hours using AI coding assistance for the first time. He later shared this experience with fellow CTOs, demonstrating that even a single head of engineering can test these tools and immediately see the impact.

Indexing the Monorepo

Deployment began by indexing the codebase—scanning files, creating structured chunks, and generating embeddings that capture relationships between code sections. At Dropbox's scale of 550,000 files, this semantic index proved critical for contextual code generation.

The result: "People can actually understand the existing codebase really well and far, far faster."

Measurable Impact on Velocity

Over 90% of Dropbox engineers now use AI tools weekly. PR throughput and cycle time reached industry-leading benchmarks. AI assistance now integrates across development phases:

• Writing — Initial code generation
• Reviewing — PR analysis and suggestions
• Testing — Test case generation
• Documentation — Automated doc updates
• Migrations — Large-scale refactoring support

Looking Forward

Dasdan concluded that "We are reexamining and redesigning every part of how we build software in the context of AI." This transformation represents more than tool adoption—it's a fundamental shift in how engineering organizations approach software development.