Summary
Arman Vaziri, who leads product and sales-led growth engineering at Ramp, walks through the architectural building blocks his team developed to turn go-to-market operations into an intent-driven, automated system. The core thesis: if you can describe a motion — target golfers at East Coast construction companies, generate outbound copy, spin up paid ad creative, trigger in-app notifications — the infrastructure should handle distribution across channels automatically.
The talk covers Ramp’s journey from messy, siloed data across GTM systems to a consistent data substrate with reverse ETL feeding a central layer agents can act on. Vaziri goes deep on durable execution: Ramp built its agent orchestration on Temporal, representing every tool call and model call as an activity within a durable thread. If a worker crashes mid-execution, the thread resumes from its last checkpoint rather than restarting from scratch — a critical reliability property for long-running GTM workflows.
A concrete example is pre-meeting briefs for account managers. The system ingests calendar events, hydrates attendee emails and meeting titles against account records (a fuzzy-match problem at Ramp because the same email can represent multiple businesses), and assembles context about product usage, open tickets, and agenda goals. Vaziri’s broader advice: solve for one team first, find the overlapping workflow patterns — automated outbound, meeting prep, QBR generation — then scale the same primitives horizontally across the organization.
📺 Source: AI Engineer · Published August 26, 2026
🏷️ Format: Workflow Case Study
