Anthropic's Growth Team Is Just One Person

What It Covers

Anthropic's growth team reportedly operated with one person for ten months, prompting Neil Patel and Eric Siu to examine how Silicon Valley startups actually handle marketing through engineers, contractors, and embedded talent — and how AI should replace workflows rather than workers across organizations of all sizes.

Key Questions Answered

• •Lean team reality check: When a company claims a one-person growth team, look deeper. Anthropic ran four Super Bowl commercials and contracted multiple external agencies simultaneously. The single headcount figure excludes engineers doing marketing work, external consultants, and agency partners — a pattern consistent across major Silicon Valley startups valued above $100 billion.
• •Forward-deployed marketer model: Embed generalist marketers directly inside client organizations full-time, similar to Palantir's forward-deployed engineer model. These marketers operate from a custom-built agent stack and customize it per client. NP Digital has run this model for over five years across multiple regions, with embedded staff returning to the agency once engagements conclude.
• •Enterprise ad spend hierarchy: Startups prioritize paid search first, then organic social, then content and SEO. Mid-market companies shift SEO to second place and add paid social third. Enterprise companies flip the model entirely — brand advertising ranks first, paid search second, paid social third — because large brands derive the majority of customers from brand recognition alone.
• •AI workflow replacement over human replacement: Brett Taylor, Sierra CEO and OpenAI board chairman, frames AI adoption correctly: replace processes and workflows, not people. Organizations building dashboards or keyword tools are still designing for human execution. The higher-value approach builds systems that only automated agents can operate — such as spinning up thousands of ad variations without human involvement in each step.
• •Train the AI, not the workforce: Large organizations trying to train hundreds of employees to produce consistent AI-generated content output are using a broken model. A more feasible approach trains the AI itself to recognize different user skill levels and adapt its outputs accordingly — ensuring minimum quality standards are met regardless of individual employee proficiency or technology familiarity.