Why skill curation matters more than the model for GTM agents

Cloudflare's principal sales ops manager presents a three-pillar framework for scaling GTM teams with AI agents. The first pillar focuses on scaling analysis by embedding business context into skill files, allowing both technical and non-technical users to query data directly. This reduces analysis time from hours to minutes and frees operations teams for strategic work.

The second pillar automates insight delivery through multi-agent workflows that generate weekly summaries, trends, and risk assessments. Cloudflare spent 2-3 months testing a three-agent architecture: data retrieval, verification, and tone refinement. This pushes standardized performance narratives to the business rather than relying on dashboards.

The third pillar provides self-service agentic workspaces where sales teams can build forecasts, QBR decks, and account plans using curated skills. The key insight is that skill curation—embedding business knowledge into deterministic prompts—is the foundation for predictable agentic behavior.

Tradeoffs include the overhead of maintaining skill quality and preventing skill proliferation. The speaker acknowledges they're in a 'Cambrian stage' of agentic experimentation but argues standardization will be necessary for source-of-truth alignment. The evidence is strong on workflow design and practical implementation details, though ROI claims are supported only by internal anecdotal feedback.