Karim Atiyeh - Building Ramp - [Invest Like the Best, EP.445]

What It Covers

Karim Atiyeh, CTO and cofounder of Ramp, explains how the company reached $1 billion revenue in five years by building AI-powered finance automation that eliminates bureaucratic waste, competing against AmEx through relentless product iteration and consumer-grade business software design.

Key Questions Answered

• •Policy Agents Replace Manual Reviews: Ramp's AI policy agents run 24/7 with more context than human reviewers—integrated with calendars, emails, and expense policies—to automatically classify transactions as in-policy or not, eliminating the manual back-and-forth that traditionally takes days and improving policy clarity over time through continuous learning.
• •Consumer-Grade Design for Business Software: Ramp converts over 50% of sales conversations into customers within 30-60 days by obsessing over minimizing user time in the app—prefilling every possible form field, skipping unnecessary questions, and automating 85% of expense reviews with 99% accuracy rather than building for decision-makers alone.
• •Speed Through Predictable Breakage: Build systems that break in predictable ways and recover quickly rather than engineering for perfection. Split teams into money-critical functions that never break and innovation areas that can break 10 times daily, allowing 20 fixes per day without customer impact—speed matters more than flawless execution in high-upside areas.
• •Technical Founders See Possibilities First: Engineers can identify what's possible with new technology before domain experts request it—the gap between having subject matter expertise and not having it is the smallest ever due to LLMs. Technical people asking the right questions of AI can become functional experts in marketing, procurement, or accounting faster than traditional learning paths.
• •Marketing Velocity Through System Design: Reduce creative iteration from two weeks to ten minutes by eliminating approval bottlenecks—enable teams to generate on-brand assets instantly with internal AI tools, change billboards seven times weekly for less cost than two static billboards, and let individuals own creative decisions without committee review to maximize experimentation speed.