How AI Agents Let GTM Teams Scale — Justin Joyce, Cloudflare
Cloudflare’s GTM workflow shows a reusable agent pattern: preprocess business data, encode domain knowledge in skills, then separate drafting, verification, and tone.
Cloudflare preprocesses business data by time, organizational slice, and metric, then connects it to curated role-specific skills. The speaker says this structure handles **80%+** of common performance requests and reports **2× efficiency** across the broader workflow.
For coding-agent systems, separate responsibilities: let one agent draft from prepared data, another verify the draft against that data, and a third adjust tone. Centralize and review skills so business definitions do not fragment across teams.
Cloudflare preprocesses business data by time, organizational slice, and metric, then connects it to curated role-specific skills. The speaker says this structure handles **80%+** of common performance requests and reports **2× efficiency** across the broader workflow. For coding-agent systems, separate responsibilities: let one agent draft from prepared data, another verify the draft against that data, and a third adjust tone. Centralize and review skills so business definitions do not fragment across teams. The reported efficiency is an internal aggregate, not a controlled benchmark. Deeper actions such as CRM updates, quoting, approvals, meeting notes, and proactive delivery still require security work and stronger workflow controls.
This supports prepared data and centrally governed skills as the substrate for routine business-agent work, but narrows multi-agent delegation to distinct drafting, verification, and tone responsibilities. Against evidence that fixed specialist chains can lose coherence, the pattern is credible only when each handoff is bounded and checked; deeper write actions remain outside the demonstrated workflow until stronger controls exist.