Dylan Ottinger
GTM systems for early-stage AI companies.
Projects
gtm-mcp-server — A single MCP server that makes the full GTM stack agent-accessible.
Wires HubSpot, Clay, Apollo, Slack, and email into one Model Context Protocol server so any MCP-compatible AI client (Claude, Cursor, etc.) can prospect, enrich, update CRM records, and fire notifications from a single prompt. A workflow that normally spans five tools and manual context-switching becomes one instruction: find VP Sales targets, enrich them, create contacts in HubSpot, and notify the team in Slack.
Why it matters: Most AI agents can answer questions — very few can execute GTM workflows. This gives agents a shared execution layer across the systems revenue teams already run on.
Built with Node.js, MCP stdio transport, and modular tool files per integration (HubSpot, Clay, Apollo, Slack, email).
crm-reliability-engine — A human-in-the-loop CRM governance layer for AI-extracted GTM data.
Instead of letting Gong, HubSpot, or transcript-derived signals blindly overwrite CRM fields, the system turns every suggested change into a confidence-scored mutation proposal. Operators can approve or reject changes in one click, and every decision is logged in an event-sourced audit trail.
Why it matters: Solves the "garbage in, garbage out" problem in automated CRM syncing, preventing data drift without requiring hours of manual sales rep auditing.
Built with Next.js, SQLite, and an event-sourced audit log.
revenue-ops-system — Real-time inbound lead routing infrastructure for GTM teams.
AI enrichment adds intent tags and urgency signals, but routing stays deterministic: territory rules, rep capacity, and weighted round-robin logic decide ownership. Failed routes go into an exceptions queue and can be replayed by an operator. Every lifecycle stage writes to an immutable event store.
Why it matters: Replaces fragile, black-box routing setups with an un-brickable, audit-logged infrastructure ensuring 100% lead compliance and zero dropped revenue.
Built with Next.js, SQLite, deterministic routing logic, and operator-facing observability.