Visual summary of operating lessons from Karim Atiyeh.

Lessons from Karim Atiyeh

Karim Atiyeh co-founded Ramp and became its co-CEO in 2026. He and Eric Glyman describe the business as a technology-and-systems-design effort extending across functions, including risk, operations, and marketing. — Ramp Co-CEO Announcement.

Part 1: The Builder Mindset & Engineering Culture

  1. Build across functions: Ramp’s co-CEOs describe building as a company-wide responsibility rather than a task confined to engineers. Their 2026 statement says Karim had directly managed risk, operations, and marketing. — Ramp Co-CEO Announcement.
  2. Prepare for more capable models: Ramp’s co-CEOs argue that each function should be positioned to take advantage of improving AI models, so new capabilities can compound into product and execution gains. — Ramp Co-CEO Announcement.
  3. Treat strategy as systems design: The jointly signed co-CEO announcement says technology is not a separate department at Ramp and that more strategic decisions now involve technology and systems design. — Ramp Co-CEO Announcement.
  4. Build tools that remove friction: Karim says Ramp invests in AI development tools so engineers can move from an idea to a reviewed fix more quickly, even as the company grows more complex. — Karim Atiyeh on Inspect.
  5. Share internal improvements: Karim asks team members who discover a better way to work to build it for colleagues. Ramp’s collaborative coding agent is one example of making that practice easier to share. — Karim Atiyeh on Inspect.
  6. Learn the domain, then build on its infrastructure: Karim says the Paribus sale taught him how cards and financial regulation work; by Ramp’s launch, banking infrastructure let the team focus more of its effort on software and product innovation. — 8VC Fireside Chat.
  7. Let the observer investigate a customer problem: Karim says the person who notices a customer problem should dig into it rather than wait for a separate team to take ownership. — Karim Atiyeh on Inspect.
  8. Solve the underlying customer problem: Karim advises listening to the problem customers describe without assuming their requested solution is the best product design. — Karim Atiyeh on Asymmetric Bets.

Part 2: Hiring, "Spikiness," and Team Building

  1. Hire for exceptional strengths: Karim argues that exceptional people can be unevenly skilled; a checklist that penalizes every weakness can miss candidates with unusually high potential. — Karim Atiyeh on Asymmetric Bets.
  2. Look for demonstrated ability: One engineering candidate impressed Karim with an app he had built despite a sparse résumé. He says the team sought contributors with both technical ability and hunger, not only prior fintech experience. — 8VC Fireside Chat.
  3. Give standout hires demanding work: Karim says exceptional hires need hard, interesting problems and a growing ambition to keep their work consequential. — Ramp Founder Chat.
  4. Train hungry engineers early: Karim says Ramp built university talent pipelines and trained engineers who lacked extensive experience but were eager to contribute. — 8VC Fireside Chat.

Part 3: Speed, Iteration, and Startup Velocity

  1. Use speed as a startup advantage: Karim describes speed as one of Ramp’s competitive advantages and says the team chose early software vendors that helped it move quickly. — Karim Atiyeh on Asymmetric Bets.
  2. Shorten the path from idea to fix: Karim says larger organizations accumulate dependencies that slow even small changes; Ramp built shared tooling to move observed problems toward reviewed fixes faster. — Karim Atiyeh on Inspect.
  3. Test tools and adjust quickly: Karim says moving quickly lets Ramp test more tools and stay flexible as AI models and infrastructure change. — 8VC Fireside Chat.
  4. Turn ideas into reviewed product changes: Karim describes Inspect as a way to carry an idea through to a visually validated, reviewed pull request more quickly while maintaining safeguards for customers. — Karim Atiyeh on Inspect.
  5. Rank customer demand against build clarity: Karim says he stack-ranks potential products by weighing customer excitement against how clearly the team can execute the build. — 8VC Fireside Chat.
  6. Make learning visible to teammates: Karim says shared problem-solving sessions let teammates see a better method, add context, and reuse the improvement through an agent or automation. — Karim Atiyeh on Inspect.
  7. Separate fast iteration from high-assurance money work: Karim distinguishes product areas where quick experiments and recovery are appropriate from money movement and risk, where the cost of a mistake calls for much greater care. — Ramp Founder Chat.

Part 4: Systems Engineering as a Business Model

  1. Design the company as a system: Ramp’s co-CEOs say functions including marketing, operations and risk can be approached as systems-design problems, not isolated departments. — Ramp Co-CEO Announcement.
  2. Avoid credit bets with asymmetric downside: Karim explains that a credit loss can outweigh revenue from many successful customers, so Ramp deliberately limits risk in that part of its business. — Karim Atiyeh on Asymmetric Bets.
  3. Apply engineering to prospecting: Karim describes an early growth system that captured visitor signals, enriched prospect data, and used it for personalized outreach. — 8VC Fireside Chat.
  4. Make the easier user path the useful path: Karim describes giving employees transaction-category hints drawn from available context instead of requiring them to start from a blank form and memorize accounting rules. — 8VC Fireside Chat.
  5. Invest in the foundations for reliable AI work: Karim says Ramp’s next AI platform work requires architectures, safeguards and evaluation frameworks that can operate reliably at scale. — Karim Atiyeh on Ramp AI.

Part 5: "Self-Driving Finance" and Corporate Spend

  1. Use agents to reduce routine finance review: Karim describes AI agents that review ordinary expenses against policy, explain decisions and escalate the minority of cases that need human judgment. This is Ramp’s stated product approach, not a guarantee of perfect autonomy. — Ramp Agents Announcement.
  2. Make card data useful for policy decisions: Karim says card transactions gave Ramp an entry point to understand business spending; the company now uses receipt and transaction context to help determine whether a charge follows policy. — 8VC Fireside Chat.
  3. Match receipts with available context: Karim says Ramp built receipt matching early using heuristics and OCR, then began drawing on receipt and transaction metadata to reduce the information employees must enter. — 8VC Fireside Chat.
  4. Free finance teams for strategic judgment: Karim compares automated financial tracking to health tracking: with information already organized, finance teams can spend less time chasing receipts and more time interpreting results and planning. — Ramp Founder Chat.
  5. Use spend context to improve decisions: Karim says card and receipt information can help business leaders understand what, why and when they transact, creating a basis for better spending decisions. — 8VC Fireside Chat.
  6. Make customer savings the mission: Karim says Ramp grew out of a Paribus-like idea for businesses: use spending information to help them save time and money, not merely offer a card. — 8VC Fireside Chat.
  7. Use cards as an entry point, not the whole product: Karim describes corporate cards as one tool inside a broader spend-management product that also addresses purchases, reimbursements, procurement and travel. — 8VC Fireside Chat.
  8. Look for overlapping subscriptions: Karim points to duplicated software purchases—including teams paying for the same AI tool—as a category that integrated card, invoice and usage data can reveal. — Ramp AI Spend Intelligence.
  9. Bring usage and invoices together for reconciliation: Karim describes matching AI usage data with invoices and sending coded spend into the general ledger, so finance teams can focus on discrepancies rather than reconstruct every charge manually. — Ramp AI Spend Intelligence.

Part 6: Integrating AI and Large Language Models

  1. Turn customer calls into internal learning: In the joint No Priors interview, Ramp’s founders discussed using language models to turn many customer conversations into short internal podcasts that help teams hear what customers are experiencing. — No Priors Founder Interview.
  2. Use AI where it reduces user friction: Karim describes using AI and transaction context to suggest expense categories and smooth the back-and-forth between employees and finance, rather than asking employees to remember every rule. — 8VC Fireside Chat.
  3. Use written policy and past decisions as inputs: Karim says Ramp’s expense-review agents can ingest a company’s policy document alongside the exceptions reflected in historical approvals. — Ramp Agents Announcement.
  4. Make the agent act inside the workflow: Karim describes agents that approve low-risk expenses, flag outliers, answer questions and suggest policy improvements inside the tools employees already use. — Ramp Agents Announcement.
  5. Keep agent decisions auditable and escalatable: Karim says Ramp logs an agent’s decisions with a rationale and escalates expenses needing human judgment; the product is intended to move from recommendations toward more autonomy as a company permits it. — Ramp Agents Announcement.
  6. Treat prompts as part of production engineering: Karim says Ramp has used prompt engineering to reduce AI error rates alongside caching and batching to manage cost and latency. — Karim Atiyeh on Ramp AI.
  7. Choose models by use case: Karim says Ramp mixes specially trained models where cost and latency require control with external general models elsewhere; he also says the team tests tools quickly so it can adapt as the model landscape changes. — 8VC Fireside Chat.

Part 7: Asymmetric Risks and Early Entrepreneurship

  1. Automate price-drop refunds: Karim says Paribus monitored consumers’ purchase receipts and sought the price difference from stores when a purchased item became cheaper. — 8VC Fireside Chat.
  2. Seek upside you can afford to risk: Karim calls a bet asymmetric when the possible gain is much larger than the manageable loss, and argues founders should choose such bets while avoiding losses that could threaten the company. — Karim Atiyeh on Asymmetric Bets.
  3. Carry card-industry learning into Ramp: Karim says selling Paribus to Capital One taught the founders about cards, financial-services operations and regulation; those lessons helped them see the scale of the opportunity that became Ramp. — 8VC Fireside Chat.

Part 8: Product Strategy and the "Divinely Discontent" Mindset

  1. Keep asking what the next bottleneck is: Karim says the growth of building tools leaves judgment about which ideas to pursue and ship as an increasingly important constraint; the work is never finished simply because a prior milestone was reached. — Karim Atiyeh on Inspect.
  2. Let useful automation stay in the workflow: Karim describes agents answering policy questions over existing channels and operating within Ramp’s current workflow, without requiring employees to learn a separate application. — Ramp Agents Announcement.