One Anthropic Marketer's Claude Code System Now Runs Sales Briefings Without Manual Approval

Adam Ward, a field marketer at Anthropic, replaced a Sunday-night ritual of manually building sales-briefing slides with a Claude Code system that now sends every rep a personalized Monday briefing of three priority action items over Slack — and runs autonomously, without manual approval, after enough iteration to earn that trust. The system pulls from Anthropic’s BigQuery marketing warehouse plus HubSpot, Clay, and Salesforce, layering in each rep’s territory and account context from Slack. Ward reports the targeted approach doubled registrations for an executive dinner within a week of deployment, and it has since expanded well past its original scope: what started with events and webinars now covers blog articles, ebooks, customer stories, and partner-ecosystem events, and the workflow has been extended from field marketing to BDR, customer success, and alliance teams.

Getting there took discipline about what not to automate blindly. Ward developed nine explicit content rules from field feedback — catching problems like invalid URLs, mismatched event audiences, and industry-specific filtering — a reminder that agentic workflows compound value through iteration on rules, not a single clever prompt. That discipline echoes guidance Anthropic published the same month in Maximizing the value of your Claude Code sessions, where engineer Lydia Hallie details how output tokens cost roughly five times the price of input tokens, cache reads run about a tenth of standard input pricing, and cache writes can cost up to double — meaning a workflow that runs continuously, like Ward’s autonomous briefing generator, lives or dies on cache discipline as much as on prompt design.

Ward’s own advice for anyone building a similar system: start from a manual process you understand well, use plain-language instructions rather than code, pilot with a small committed group, and turn feedback into explicit rules before scaling. For companies weighing whether agentic tools belong in sales enablement, this is a concrete, first-party case study of what “autonomous” actually looks like in production — not a demo, but a system that has been running long enough to accumulate nine rules and outgrow its original team.