Behind the Craft Inside Ramp, the $32B Company Where AI Agents Run Everything | Geoff Charles
87 snips
Mar 15, 2026 
Geoff Charles, Ramp's Chief Product Officer known for leading AI-native product work, shares how his team uses Claude Code and internal agents to speed product discovery and delivery. Hear demos of voice-of-customer, data-analysis, and inspect agents. Topics include automated PRs, release automation, Ramp’s L0–L3 AI framework, and the evolving PM role into hands-on builders or strategic GMs.
AI Snips
Chapters
Transcript
Episode notes
Specs Are Now Prompts That Produce Products
- AI dramatically lowers the cost of product development so PMs shift from writing specs to specifying prompts that produce working prototypes.
- Geoff Charles explains Ramp treats the spec as a prompt and collaborates with LLMs to turn prompt-to-product, not just design documents.
Voice Of Customer Agent Turned 8 Days Of Research Into 8 Minutes
- Ramp built a Voice of Customer agent that ingests Gong calls, Salesforce notes, support tickets, in-app surveys and Snowflake data to answer PM questions.
- In a demo Geoff ran 90 days of support tickets and chat logs and produced prioritized topics and links in about eight minutes instead of eight days.
Data Analysts Become Natural Language CoWorkers
- Ramp's data-analyst LLM bridges schema knowledge and analytics so anyone can ask business questions and get full reports, not just single queries.
- They evolved from a Q&A bot to Snowflake CLI + Claude Code skills that generate HTML reports tied to Ramp's data and recommended growth ideas.