AI in GTM at Notion — Flora Liu
Notion treats GTM automation as a shared context system for humans and agents, not an AI layer over disconnected tools. The key design choice is keeping risky customer actions human-approved.
Notion reduces GTM workflows to **Know → Decide → Act → Learn**. Snowflake produces versioned customer entities through daily and sometimes real-time transforms; DynamoDB serves denormalized profiles in milliseconds, alongside agent-generated notes and research.
Give humans and agents one inspectable context layer, retain timestamps and ownership, and trace every model step. Let agents gather context and draft work, while humans approve risky customer-facing actions. Early context-aware recommendations were **63% more likely** to prompt a next step.
Notion reduces GTM workflows to **Know → Decide → Act → Learn**. Snowflake produces versioned customer entities through daily and sometimes real-time transforms; DynamoDB serves denormalized profiles in milliseconds, alongside agent-generated notes and research. Give humans and agents one inspectable context layer, retain timestamps and ownership, and trace every model step. Let agents gather context and draft work, while humans approve risky customer-facing actions. Early context-aware recommendations were **63% more likely** to prompt a next step. The system is still being built, and that percentage does not establish broader deal impact. Notion also chose to own its context layer while relying on vendors for functions such as CRM and email, a boundary that may differ for smaller teams.
This turns GTM automation into a context-ownership problem: recommendations become safer when humans and agents share versioned, attributable customer state and risky actions remain approval-gated. It confirms process-first, inspectable automation, while narrowing the 63% result to increased next-step activity rather than revenue impact or a universal case for building the context layer in-house.