How Ramp let 400 reps build their own custom GTM tools

How does one growth team give 400 salespeople the tools they each want? Keyan Sarrafzadeh from Ramp's growth platform team explains how Ramp let reps build their own custom GTM tools, then pulled the best ones back into the core app. The answer is a growth MCP: a permissioned layer between AI clients and Ramp's CRM and outreach tools that reps build against directly. What you'll learn: How Ramp went from a single outbound automation team (OATS) to Ramp Revenue, one app where every rep sees who to work, why, and the intent signals on their book What sits inside the growth MCP: 79 tools today (meeting briefs, prospect research, intent signal enrichment, CRM updates), with more added every week The feedback loop: every agent tool call must log a rationale for why, and Ramp clusters those rationales to see what reps actually need How reps ship internal web apps through a platform called Ramplify Ramp's build-vs-buy rule (only build in-house for a real edge) and why logged rationale beats shadowing reps for product research Sources & further reading: Introducing the Model Context Protocol (Anthropic, official) https://www.anthropic.com/news/model-... - the open standard behind the growth MCP How Ramp set up a data cloud to democratize AI (American Banker) https://www.americanbanker.com/news/h... - Ramp's Snowflake warehouse holding sales call transcripts and customer data Ramp on Claude (Anthropic customer story) https://claude.com/customers/ramp - non-technical staff in sales building with Claude Code against Snowflake Ramp AI Index (June 2026) https://ramp.com/data/ai-index-june-2026 - Ramp's company-wide AI adoption push, cited by its head of product More from Deepline: Event recap: 6 GTM teardowns from top teams https://deepline.com/blog/gtm-ai-nyc-... This talk, written up https://deepline.com/blog/gtm-ai-nyc-... Quickstart https://deepline.com/quickstart Use cases https://deepline.com/use-cases 0:00 Intro and Ramp's growth platform 1:31 Ramp Revenue: the seller's toolkit 3:01 Learning from field-built tools 4:31 Internal web apps with Ramplify 6:01 Tracking sales workflows and usage 7:31 Agent rationale and the feedback loop 9:01 Q&A and wrap-up 10:31 Build vs buy and the internal stack 12:19 Team structure and data science 13:33 Evaluating internal vs external tools 15:00 Audience questions GTM + AI NYC Lightning Talks, hosted by Deepline. Full playlist:    • GTM + AI NYC Lightning Talks (June 30, 2026)   Deepline is one CLI for your GTM data: enrichment, scraping, and sequencing across 44+ providers, with every run logged to a database you own. Try it from Claude Code or Codex at deepline.com (use code GTMASCODE).