Invest Like the Best with Patrick O'Shaughnessy

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

October 21, 2025

Key Takeaways Copied to clipboard!

  • The next major phase of AI adoption involves programming LLMs as the core product (agents) rather than just using them as productivity tools to speed up existing code/work. 
  • Building consumer-grade user experiences for business software, obsessing over minimizing user interaction time, is a key differentiator that drives customer adoption over incumbents. 
  • The culture at Ramp is built on mutual accountability where every team member must use and advocate for their own product, reinforcing the tenet that everyone is responsible for selling and improving the customer experience. 
  • Ramp's business model evolved from relying solely on Total Payment Volume (TPV) to incorporating a SaaS revenue model because TPV growth plateaus as businesses mature and move more spending off-card, necessitating a value capture mechanism that scales with complexity. 
  • Technical founders are uniquely positioned to dominate the current era because their understanding of underlying technology allows them to see product possibilities (the 'car' instead of the 'faster horse') that non-technical domain experts might miss. 
  • To accelerate growth, companies must ruthlessly optimize the system for experimentation, prioritizing fast iteration cycles, individual accountability for creative output, and empowering teams with tools that minimize dependency and friction, rather than focusing on perfect upfront estimation. 

Segments

Ramp’s Competitive Stance
Copied to clipboard!
(00:06:08)
  • Key Takeaway: Ramp maintains an upstart mentality by breaking down problems into small, autonomous teams to preserve rapid decision-making speed despite exceeding $1 billion in revenue.
  • Summary: The feeling of being an upstart is maintained by giving small teams full autonomy over the problems they tackle. The current AI landscape is still early, with most companies using LLMs only for productivity gains rather than integrating them as the core product. Ramp is actively entering the phase where LLMs become the code itself, programmed as agents with instructions and infinite loops.
AI Policy Agent Framework
Copied to clipboard!
(00:09:26)
  • Key Takeaway: The ‘policy agent’ uses integrated context (calendar, email) to enforce expense policies 24/7, evolving the policy document into a continuously improving, context-aware system.
  • Summary: The policy agent applies expense policy logic against transactions with more context than human reviewers, enabling live, 24/7 enforcement. This agent also provides feedback to clarify and improve the written expense policy over time. This same principle of agent-based automation applies to undocumented workflows like invoice processing and validation.
Customer Adoption Drivers
Copied to clipboard!
(00:12:45)
  • Key Takeaway: Ramp’s early customer acquisition was driven by an obsession with building consumer-grade user experiences, reversing the trend of business software designed only for decision-makers.
  • Summary: Early success stemmed from prioritizing user experience, aiming for the polish of an Instagram applied to business software. This involved obsessively polishing every interaction, pre-filling forms, and minimizing necessary user input. This design obsession translates into an ongoing practice of asking how to skip steps or answer questions without asking the customer.
Divinely Discontent Management Style
Copied to clipboard!
(00:16:28)
  • Key Takeaway: Karim Atiyeh’s ‘divinely discontent’ style stems from believing current results reflect past efforts, making celebration of lagging indicators risky, and the inherent drive to continuously solve problems.
  • Summary: Celebrating current success is avoided because today’s results were baked in months ago, risking complacency regarding current, critical problems. The job is never done, and the alternative to constant improvement is feeling purposeless. This mindset is fueled by observing massive administrative waste in sectors like healthcare and education, which Ramp aims to reduce by accelerating business operations.
Ramp’s Cultural Tenets
Copied to clipboard!
(00:19:35)
  • Key Takeaway: Ramp fosters a culture of mutual accountability where teams are responsible to peers rather than just managers, ensuring cross-functional alignment on customer value.
  • Summary: The culture emphasizes peer accountability, meaning sales cares about product, and product cares about sales, as everyone depends on each other’s success. The core interview filter for new hires is whether the interviewer would join or start a company with that person, focusing on perseverance and alignment on mission. A company is fundamentally a collection of people solving increasingly difficult problems together.
Ramp’s Business Model Overview
Copied to clipboard!
(00:21:46)
  • Key Takeaway: Ramp is a finance automation platform stitching together disparate workflows—expenses, procurement, AP, and accounting—to provide a live view of spending, aiming to eliminate weeks/months of closing time.
  • Summary: The platform serves finance teams across all company sizes, automating busy work and integrating workflows that typically live across many systems. The ambition is to allow companies to run finance on autopilot, removing the need to be a finance operations expert. The total addressable market is estimated to be over $100 billion annually in wasted administrative spend.
Entrepreneurial Prologue and Early Lessons
Copied to clipboard!
(00:24:07)
  • Key Takeaway: Growing up in conflict-prone Lebanon instilled a sense of external risk management, while early exposure to MIT research shaped a fascination with engineering breakthroughs.
  • Summary: Growing up in Beirut instilled a sense that things are ephemeral, leading to parental pressure to maximize chances of studying abroad. Early exposure to the Research Science Institute (RSI) at MIT involved intense self-teaching in computer science research. Working at a startup using early NLP techniques (Markov chains) provided foundational experience in processing language data.
Parabus Founding and Lessons
Copied to clipboard!
(00:29:29)
  • Key Takeaway: Parabus automated price tracking for consumers against retailer guarantees, teaching the engineering lesson of building systems that break predictably for rapid iteration.
  • Summary: Parabus was founded after observing dynamic pricing on Amazon, using VBA macros to track SKU prices daily. The core lesson learned was engineering pragmatism: build quickly assuming things will break, focusing on fast recovery rather than building systems that never fail. This led to a split approach at Ramp: money movement systems must be built robustly, while UI/UX elements should be iterated on quickly, even if they break temporarily.
Scary Moments and Validation
Copied to clipboard!
(00:39:02)
  • Key Takeaway: Facing cease and desist letters from major retailers, including a security challenge from Amazon, was scary but ultimately validating proof that Parabus was correctly targeting high-value friction points.
  • Summary: Receiving legal threats from large firms while being a small team was terrifying but confirmed the product was effective. The interaction with Amazon involved a call with senior AWS leadership (potentially Andy Jassy) who were supportive, contrasting with the antagonism from Amazon Retail. Being attacked by incumbents is a strong indicator that a company is doing something correct and right.
Selling Parabus to Capital One
Copied to clipboard!
(00:43:04)
  • Key Takeaway: The Parabus sale to Capital One was driven by the need for firepower against legal challenges and the realization that credit card companies offered the best access to purchase data for growth.
  • Summary: The initial goal was partnership with credit card companies to gain access to recent purchase data for targeted marketing, as they knew exactly who was spending where. The acquisition conversations evolved from partnership because Capital One wanted to differentiate its card product using Parabus’s technology. The card business model—revenue increasing directly with customer utility and usage—was seen as refreshing and powerful.
Birth of Ramp and Initial Product
Copied to clipboard!
(00:46:56)
  • Key Takeaway: Ramp began by applying the Parabus principle of saving money/time to business spend, using the corporate card as the initial entry point to understand and automate all related workflows.
  • Summary: The initial ambition was to build technology to save businesses time and money, recognizing that time waste is often greater than direct monetary waste. The first product focused on integrating deeply with email to automatically match receipts to card transactions and cleaning up cryptic merchant identifiers. The pitch to early design partners was simply that Ramp would not be worse than existing cards, with the promise of future improvement based on founder trust.
Investor Selection Strategy
Copied to clipboard!
(00:54:14)
  • Key Takeaway: Ramp treats investors like employees, selecting them based on what they can bring to the partnership beyond capital, and nurturing those relationships continuously.
  • Summary: Investors are viewed as private equity holders of RAMP equity, similar to employees investing time and effort. This requires continuous education about the business challenges, not just reaching out when capital is needed. This strategy allowed Ramp to build a cap table where most investors have participated in multiple rounds, often starting small to build the relationship.
Hardest Fundraising Round
Copied to clipboard!
(00:57:13)
  • Key Takeaway: The most difficult round to raise was not due to investor rejection, but due to the founder’s internal struggle to justify raising capital when valuations were declining in a down market.
  • Summary: The difficulty in a 2022/2023 round was internal alignment on the need to raise when valuations were falling, as Ramp did not critically need the cash. Raising the round served as a necessary price discovery mechanism to establish a new checkpoint for the company’s valuation. The ultimate goal remains maximizing enterprise value created for customers, not maximizing valuation at every single round.
Product Evolution and TAM
Copied to clipboard!
(00:59:37)
  • Key Takeaway: Ramp’s product development follows an algorithm to minimize total time wasted by customers across all finance workflows, mapping this time waste to the total addressable market (TAM).
  • Summary: The core value metric is minimizing the total time wasted by finance teams on reviewing transactions, matching receipts, and procurement research. As new products are built—like predicting transaction memos or providing procurement price benchmarks—customer adoption increases, raising the total value delivered. Early customers were often only interested in the card and cashback features, requiring Ramp to prove value across the entire workflow.
Value Creation Algorithm
Copied to clipboard!
(00:59:47)
  • Key Takeaway: Ramp’s core algorithm focuses on minimizing total time wasted by customers across transactions, reviews, and accounting categorization.
  • Summary: The algorithm aims to reduce time wasted on unnecessary transactions, reviewing existing ones, and manual data entry like memo requests or accounting categorization. This total addressable market of wasted time dictates product development priorities. Early customers often only adopted the card product, but the value of software automation led to a successful land-and-expand motion.
TPV to SaaS Revenue Shift
Copied to clipboard!
(01:02:56)
  • Key Takeaway: Ramp transitioned to a SaaS revenue model because the TPV-based model breaks for large businesses where card spend plateaus relative to overall complexity.
  • Summary: The TPV model incentivizes spending, which conflicts with Ramp’s mission to help companies spend less. For larger, more complex businesses, card spend does not scale linearly with the value derived from their automation software. Charging for software provides a necessary feedback signal from the market, guiding the roadmap better than free products.
Pricing AI Agent Work
Copied to clipboard!
(01:06:48)
  • Key Takeaway: Charging for AI work should ideally be based on the complexity of the task solved, not on resource consumption like tokens or CPU time.
  • Summary: Charging based on agent time or tokens risks incentivizing engineering teams to build inefficient systems that use more resources. The ideal mechanism is to charge based on the value or complexity of the task the AI agent successfully solves for the customer. The exact pricing structure for these new AI agents is still being figured out.
Technical Founders Seeing Possibilities
Copied to clipboard!
(01:07:43)
  • Key Takeaway: Technical founders are better equipped to see product possibilities in the AI era because they understand the underlying technology’s potential beyond current customer requests.
  • Summary: Customers often ask for incremental improvements (faster horses), but technical experts understand the paradigm shift enabled by new technology (the car). The gap between lacking domain expertise and gaining it has shrunk due to LLMs, allowing engineers to quickly become knowledgeable enough to build superior solutions in unfamiliar domains.
Engineering Approach to Marketing
Copied to clipboard!
(01:10:18)
  • Key Takeaway: Applying an engineering mindset to marketing involves fixing the experimentation system to increase the velocity of testing ideas rather than assuming channels fail.
  • Summary: The CTO took over marketing when leading indicators like conversion were high but recognized lead generation growth was plateauing, necessitating a change. The focus was on accelerating the clock speed between idea generation and customer exposure, such as redesigning the brand asset creation process from a two-week brief cycle to near-instantaneous output.
Marketing Velocity and Risk Taking
Copied to clipboard!
(01:15:37)
  • Key Takeaway: Reducing the cost and time required to change marketing assets (like billboards) allows for taking more shots on goal and increasing risk tolerance.
  • Summary: The ability to change a billboard for $1,000 seven times in a week is more powerful than buying two static billboards, demonstrating the value of iteration speed over static investment. This principle of finding system hacks to increase velocity applies across all marketing channels, including paid ads and SEO, where platform changes require rapid adaptation.
Principles of Attention Grabbing
Copied to clipboard!
(01:22:41)
  • Key Takeaway: Grabbing attention requires seeking differentiation for differentiation’s sake, as seen in Ramp’s choice of yellow branding against the industry standard blue/green.
  • Summary: Differentiation, like Dyson’s approach, creates powerful brand recognition where visual cues (like Ramp’s yellow) become instantly recognizable even without legible text. This contrasts with the saturation of common formats, such as the current sea of generic launch videos, which leads to audience fatigue.
Recruiting for Spikiness and Speed
Copied to clipboard!
(01:25:17)
  • Key Takeaway: Early-stage recruiting should prioritize hiring ‘spiky’ individuals with extreme talent in specific areas where the company has asymmetric information, rather than hiring generalists.
  • Summary: Small companies cannot compete with giants on brand or salary, so they must look for extreme talent signals, like excelling in difficult freshman-year classes or specialized programs like RSI. The goal is to assemble an ‘Avengers’ team where each member has a clear superpower, as vetting too many attributes leads to average hires.
Tactics for Business Speed
Copied to clipboard!
(01:29:40)
  • Key Takeaway: Process often hinders speed in high-performing teams; avoiding standard estimation processes prevents the incentive to pad timelines and ensures faster delivery cycles.
  • Summary: The most critical tactical element for speed is shortening the cycle between idea and customer feedback, often by avoiding rigid processes like upfront effort estimation. When estimation becomes a performance metric, teams are incentivized to overestimate task duration to ensure they hit their targets, slowing down overall velocity.
Future of Payments Infrastructure
Copied to clipboard!
(01:32:30)
  • Key Takeaway: For consumers, the underlying payment rail (card vs. stablecoin) is irrelevant; the shift in infrastructure will matter only if payment decision-making moves from humans to autonomous agents.
  • Summary: The high cost of US card payments is largely driven by consumer rewards, not just interchange fees. If agents begin making purchasing decisions, they might optimize for rails based on speed and cost rather than consumer rewards, potentially shifting payment flow dynamics. Ramp’s focus remains on optimizing customer workflows regardless of the underlying rail technology.
Vision for Self-Driving Finance
Copied to clipboard!
(01:36:00)
  • Key Takeaway: The ultimate vision for Ramp is for customers to log in zero times, achieving ‘self-driving finance’ where AI agents automate workflows and optimize decisions continuously.
  • Summary: This future mirrors the evolution of cars from mechanical operation to autonomous driving, where the system handles the complexity of moving from point A to point B. The continuous accumulation of customer data and successful outcomes allows Ramp’s models to learn intent and improve reasoning over increasingly complex financial tasks.
Managerial Growth Through Constraint
Copied to clipboard!
(01:38:36)
  • Key Takeaway: Forcing constraints, such as a broken arm, compelled the CTO to delegate, manage, and focus on maximizing the team’s total output rather than his individual contribution.
  • Summary: Early in Paribus, the CTO’s inability to code due to injury forced him to become better at explaining architecture and delegating tasks to junior engineers. This experience shifted his focus from direct output to maximizing the sum total output of the team, a necessary step for scaling leadership.
Evolution of Company Building Views
Copied to clipboard!
(01:40:20)
  • Key Takeaway: The reward for conquering business challenges is simply more complicated challenges, leading to an intentional focus on enjoying the journey and the people involved.
  • Summary: The initial belief that solving a challenge yields a great, final reward has been replaced by the realization that success only brings the next level of complexity. Therefore, the most important factor for sustained enjoyment is working with brilliant, spiky people across multifaceted disciplines like engineering, design, and capital markets.
Pride in Talent and Diaspora
Copied to clipboard!
(01:42:21)
  • Key Takeaway: The greatest source of pride at Ramp is the caliber of people attracted to the company and the success of the diaspora of talent that moves on to start other great companies.
  • Summary: The goal is for Ramp to be the last job employees ever need to apply for, viewing external interest as a sign of hiring the right people, not a retention failure. The company celebrates the achievements of its alumni, including interns winning international physics medals while employed.
Kindest Act Received
Copied to clipboard!
(01:43:38)
  • Key Takeaway: A friend’s family immediately offered the speaker, a 16-year-old stranger during a crisis in Lebanon, a place to live in New York with full support.
  • Summary: When war closed the airport during the Research Science Institute program, a friend’s family welcomed the speaker into their home without hesitation. They provided meals, pocket money, and contact lists, treating him instantly like family, which established a profound, lasting bond.