Behind the Craft – Episode Summary

Episode Title: Inside Ramp, the $32B Company Where AI Agents Run Everything
Host: Peter Yang
Guest: Geoff Charles, CPO of Ramp
Date: March 15, 2026

Overview

In this episode, Peter Yang interviews Geoff Charles, Chief Product Officer (CPO) at Ramp, one of the fastest-growing, most AI-native companies in the world. They discuss how Ramp uses AI agents across all product development stages—from understanding customer needs, to shipping code, to running internal operations—resulting in unprecedented velocity and efficiency. With more than 50% of Ramp’s production code generated by AI (and aiming for 80% soon), the conversation explores the implications for product managers (PMs), product development, company culture, and the future of work.

Key Discussion Points & Insights

1. The AI-Driven Product Development Workflow

• AI Agents Integral to Every Role:

  • Geoff emphasizes that not using “Claude code” (Anthropic’s advanced LLM agent) means underperforming—regardless of your role (00:00).
  • Quote:
    “If you're not using Claude code this year, no matter what your role is, you're probably underperforming compared to others on the company.” – Geoff (00:00)

• Product Process Fundamentals (01:14–02:29):

  • Core principles (customer understanding, identifying the right solution, building/testing) remain unchanged.
  • AI drastically lowers the cost and effort at every stage.

• Prompt to Product Loop (02:38–03:12):

  • PMs no longer write specs only for engineers; AI now reads and implements specs.
  • “It's actually AI that's reading the spec now versus engineers.” – Geoff (00:06, 02:38)

• PMs and Builders Empowered:

  • Prototypes and even production-level products are spun up rapidly by PMs via prompting, with engineers/other roles validating and iterating (16:37–18:31).

2. Leveraging Data and Voice of the Customer, Automated

• Voice of the Customer Agent (03:36–06:40):

  • AI bots aggregate and analyze all customer interactions: support tickets, sales notes, call transcripts, analytics data, etc.

  • Example: Extracting 90 days’ worth of feedback in under 10 minutes—something that would take a human analyst days.

  • Quote:
    “This was done in about eight minutes—something that would have taken eight days for a human.” – Geoff (05:07)

  • Bots can continue the conversation, draft outreach emails, schedule meetings, and fetch detailed customer insights upon request.

  • Memorable Moment: Using a Slack interface to simply “DM” the agent—mirroring how you’d assign a human analyst (06:34–06:40).

• Quantitative Analysis via AI (07:47–10:27):

  • The “Ramp Research” assistant lets anyone query internal databases conversationally.
  • Has been upgraded to use Snowflake CLI plus Claude “skills” for more complex, goal-driven reporting (e.g., generate a full HTML report with actionable insights).
  • Quote:
    “At the end of the day, like, it’s funny, you know, you ask a question but you have a goal. You should just tell AI what your goal is and ... you’ll actually be surprised at the questions that the AI can actually ask themselves to get to the goal.” – Geoff (10:32)

3. The Technical Transformation

• AI-Generated Code at Scale (15:18–18:31):

  • 50% of production code at Ramp is now AI-generated, up from 30% recently—projected to reach 80%+ shortly.
  • AI (Inspect tool) handles both front- and back-end, understands codebase and design system, and can implement PRs (pull requests) autonomously.
  • Quote:
    “We’ve hit coding escape velocity and it’s a brave new world out there.” – Geoff (15:18)
  • “If you have infinite coders … you are actually the bottleneck.” – Geoff (15:18)

• Shipping Faster Across Functions:

  • PMs, designers, operators, account managers, and even sales staff participate in shipping features and improvements.
  • Engineers remain primary users, but “the first pass” now usually comes from AI agents (20:38–20:39).

4. Maintaining Product Quality & Cohesion

• Automated Reviews & Ship Process (20:41–23:02):

  • PRs are triaged for complexity.
  • Large changes go through progressively broader user groups, from internal “dogfooding” to an opt-in beta (10% of customers) to full release.
  • An internal “Release Agent” automates help center articles, enablement docs, impact analytics, and pulls together context for leadership.

• Leadership Evolving: Process over Micromanagement (23:19–25:19):

  • Leader’s role is to analyze what broke in the process when quality issues arise, and fix the process, not just the output.
  • Quote:
    “My job is to automate my job and all our jobs is to automate our jobs.” – Geoff (25:19)

5. Rethinking Planning, Strategy, and Organizational Culture

• Short-Term, Adaptive Planning (25:27–27:21):
  • Formal planning only projects out 3 months now—anything further risks obsolescence given product velocity.
  • Planning serves: (1) Aligning on strategy / trade-offs; (2) Creating commitment/accountability; (3) Supplying sales with reliable roadmaps (all, again, aided by automation).
  • Leadership uses LLMs to synthesize and translate operational-to-strategy communication for all stakeholders efficiently.

6. The Evolving Role of Product Managers

• PMs “at Risk”—Shift to Builders or Strategists (28:12–32:07):

  • Traditional PM skills (stakeholder management, frameworks, documentation) are quickly becoming obsolete.

  • Future PMs will either:

    • Become hands-on builders leveraging AI (“because code is free now”), or
    • Take on business/GM roles focusing on competition, distribution, and monetization.

  • Quote:
    “All that matters now is: Are we going in the right direction? How fast can we go? How do we remove bottlenecks?” – Geoff (28:28)

• Management & Meetings: Cut the Fat (32:07–33:45)

  • Fewer committees, signoffs, or alignment meetings—prove value by building and shipping.
  • Managers are expected to retain hands-on builder skills and model adoption for their teams.

7. How to Build an AI-Native Organization

• AI Proficiency Levels Framework (34:05–36:50):

  • L0 = Occasional users, L1 = Custom GPT/Notion agent users, L2 = Automation/committing code, L3 = Systems builders.

  • The goal is to get everyone up the ladder via:

    • Public sharing of projects and wins
    • Removing barriers to access
    • Providing internal skills repositories
    • Highlighting non-builder use cases (legal, finance, marketing)
    • Holding office hours and mentorship
    • Mandating proficiency in hiring and performance management

  • Quote:
    “If you're not a self-starter and you don't have that growth mindset, it’s gonna be very, very hard to train you out.” – Geoff (34:05–36:50)

• ROI vs. “AI Budget”

  • Ramp is unrestrained in AI agent access—cost is not a concern compared to productivity and learning speed advantages (37:22–38:50).

8. PM Careers and the Future of Work

• Optimize for Builder Skill, Not Management (39:15–41:16):

  • Management skill is less valuable; focus on being the best AI-powered builder.

  • All knowledge work—including PM, design, legal, etc.—is being reshaped; adopt, adapt, and iterate.

  • The “aha moment” (the red pill) for AI productivity is career-changing.

  • Memorable Quotes:
    “Your career is about impact. And right now, the impact that you can have is to ship great products faster and move more metrics for customers and the business.” – Geoff (40:05)
    “The time it takes to go from your taste and vision to a product is shorter than ever. … It’s really, really, really exciting time to be a builder here.” – Geoff (42:49)

• Products Becoming Co-workers (41:27–41:49):

  • Tomorrow’s software isn’t apps/tables/buttons, but AI agents that feel like coworker-experts.
  • “Fundamentally, software is dead. It’s all going to be like coworkers.”

Notable Quotes & Memorable Moments

• “If you're not using Claude code this year, no matter what your role is, you're probably underperforming compared to others on the company.” – Geoff (00:00)
• “It's actually AI that's reading the spec now versus engineers.” – Geoff (00:06, 02:38)
• “We’ve hit coding escape velocity and it’s a brave new world out there.” – Geoff (15:18)
• “My job is to automate my job and all our jobs is to automate our jobs.” – Geoff (25:19)
• “All that matters now is: Are we going in the right direction? How fast can we go? How do we remove bottlenecks?” – Geoff (28:28)
• “The time it takes to go from your taste and vision to a product is shorter than ever.” – Geoff (42:49)

Key Timestamps

• 00:00–01:14 — AI as standard for all roles; PMs prompt AI, not engineers
• 03:36–06:40 — "Voice of the Customer" AI agent demo and workflow
• 07:47–10:27 — Ramp Research bot and evolution to Claude code "skills"
• 15:18–18:31 — How AI generates production code and the Inspect workflow
• 20:41–23:02 — Automated shipping, review, and QA process
• 23:19–25:19 — Leadership's new role: fixing process via automation
• 25:27–27:21 — Three-month planning cycles and instant sales enablement
• 28:12–32:07 — Future of the PM role: builder or strategist, not documenter
• 34:05–36:50 — The four AI proficiency levels at Ramp
• 39:15–41:16 — Advice to PMs: prioritize building, learning, and impact
• 41:27–42:49 — From "product" to "coworker": the future of software

Useful Takeaways for Product Leaders and Creators

• Adopt a hands-on AI-focused mindset: Building—more than stakeholder management or docs—is the core PM skill of the future.
• Automate everything possible: Your job is to automate your job; the only way to keep up is to make yourself and your team more productive with AI agents.
• Shortest path from customer pain to shipped feature: AI’s true advantage is compressing the cycle from feedback/idea to shipped, real product.
• Empower all roles to use and build with AI: Not just engineers; everyone from finance to marketing to sales must become LLM-fluent.
• Plan in radically short cycles: Three months is the new horizon—velocity wins.
• Organize for transparency, sharing, and bottom-up innovation: Remove policy and technical friction; share best practices constantly.
• AI’s impact will soon touch all knowledge work: Skills need constant upgrading, and career paths will emphasize impact over hierarchy.

For Listeners Who Haven’t Tuned In

This episode presents a clear, compelling vision of how a world-class startup operates with AI deeply embedded into every workflow, making human creativity and velocity the ultimate competitive edge. Geoff Charles offers a toolkit for future-proofing your product and career—by embracing AI-powered building, focusing on impact, and continually learning and sharing. Ramp’s story is both blueprint and call to action for companies and builders ready to thrive in the next era of software.