A Non-Programmer Single-Handedly Carried All of Anthropic's Growth Marketing for Ten Months

How much can AI actually improve an individual’s work efficiency?

Recently, a post about Anthropic went viral on social media. The poster, Ole Lehmann, claimed that the entire growth marketing team of Anthropic, a company valued at $380 billion, consists of just one person—a non-technical marketing professional—who has been independently handling paid search, paid social, app store optimization, email marketing, and SEO for nearly ten months.

Shortly after the post was published, it was questioned in the comments, but the person involved quickly confirmed it himself. Austin Lau, the growth marketer, responded: when that article was written, he was indeed the only person doing growth marketing, supporting all those activities alone for almost ten months.

Image: Related tweet (Source: X)

In late January this year, Anthropic released an official case study detailing Austin Lau’s work approach. Around the same time, Anthropic also published an internal white paper titled “How the Anthropic Team Uses Claude Code,” covering use cases across ten teams, from data infrastructure to legal, with growth marketing being one of them.

The white paper states: The growth marketing team focuses on channels like paid search, paid social, mobile app stores, email marketing, and SEO, and is described as a “non-technical one-person team” that relies on Claude Code to automate repetitive marketing tasks, building workflows that traditionally would require significant engineering resources.

(Source: Anthropic)

Austin Lau is not an engineer. In an official Anthropic case video, he mentioned that he has “never written a line of code.” When he first started using Claude Code, he even had to Google “how to open Terminal on Mac.” When Claude Code was first released, his initial reaction was “completely unsure who this product is for,” as a marketer, he found its purpose unclear.

The turning point came when a colleague shared a Claude Code installation guide aimed at non-technical staff in the company Slack group. Out of curiosity, Austin installed it. A week later, he had built two automation workflows that completely changed how he worked.

The first was a Figma plugin. Creating paid social ads and app store marketing requires handling大量视觉素材 in Figma. The old process involved: when creating multiple copy variants for the same design, he had to manually duplicate frames in Figma, switch back and forth between Google Docs and Figma, copying and pasting titles one by one. If there were 10 copy variants to adapt to 5 different aspect ratios, this mechanical work could easily take half an hour.

Image: Austin Lau (Source: Anthropic)

He described this pain point to Claude Code in natural language, asking it to help write a Figma plugin. During the process, he referenced Figma’s API documentation, researching and prototyping simultaneously. The first version of the prototype was imperfect, but enough as a starting point. He kept debugging and refining, eventually creating a usable plugin.

(Source: Anthropic)

The plugin works as follows: select a static image frame, and the plugin automatically recognizes components like titles, call-to-action buttons, code blocks, etc. It then batch-generates separate Figma frames from a prepared list of copy, with each variant corresponding to a new set of copy. It can generate up to 100 ad variants per batch, taking about half a second per batch. What used to take 30 minutes of manual work now takes 30 seconds.

The second workflow is for generating ad copy for Google Ads. Responsive search ads on Google have strict character limits: 30 characters for headlines, 90 for descriptions. Previously, he had to draft in Google Sheets, manually check character counts, then paste each into Google Ads.

Austin created a custom slash command “/rsa” in Claude Code. When triggered, Claude asks for campaign data, existing ad copy, and keywords, then cross-references his pre-set “Agent Skills,” which include Anthropic’s brand tone, product accuracy standards, and Google Ads RSA best practices.

The system uses two specialized sub-agents—one for headlines, one for descriptions—that work within their character constraints, producing higher quality output than stuffing both tasks into a single prompt.

Finally, Claude Code packages 15 headlines and 4 descriptions into a CSV file ready for upload to Google Ads. Austin emphasizes that the generated copy is just a starting point; he reviews each one for value proposition, tone, differentiation from competitors, etc. But at least the boring initial drafting and formatting are fully automated.

These two workflows have already dramatically increased efficiency, but Austin’s system doesn’t stop there. He also built a connection to Meta Ads API via an MCP (Model Context Protocol) server.

Through this integration, he can directly query ad performance, spend data, and results within the Claude desktop app, without opening Meta Ads dashboards. Questions like “Which ads have the highest conversion rates this week?” or “Where am I wasting budget?” can be answered directly by Claude with real-time data.

More importantly, it’s a closed loop. Austin set up a memory system that records assumptions and test results from each ad iteration. When starting a new round of variations, Claude automatically retrieves all previous test data—what worked, what didn’t—allowing the next generation to build on past experiments. This system improves with each cycle. Such systematic tracking of hundreds of ads would normally require a dedicated data analyst in traditional teams.

According to the white paper from Anthropic, the results of this approach include: reducing ad copy creation time from 2 hours to 15 minutes, increasing creative output by tenfold, and enabling a single person to test ad variants across channels and volume that surpass most full-scale marketing teams.

In that white paper, growth marketing is just one of ten case studies. The data infrastructure team uses Claude Code to debug Kubernetes cluster failures—solving issues that would normally require network specialists in minutes; the inference team, without ML backgrounds, uses it to understand model functions and settings, reducing documentation lookup from an hour to 10-20 minutes; the product design team directly modifies frontend code with Claude Code, discovering that designers are making “large state management changes you wouldn’t normally see from designers”; the legal team used it to create a predictive text assist app for family members with language barriers in just one hour, despite having no prior coding experience.

Different roles, different uses, but the conclusion is consistent: Claude Code is blurring the line between “possible” and “impossible,” a boundary that was almost entirely determined by technical ability in the past.

Austin Lau summarizes in the case: “The gap between ‘I wish this existed’ and ‘I can build it myself’ is much shorter than most people think.”

Of course, it’s important to note that growth marketing does not equal the entire GTM (go-to-market) strategy. Anthropic has a full brand, product marketing, and communications team. Austin Lau is responsible for performance marketing—paid campaigns, app store optimization, SEO—quantifiable channels.

In February this year, Anthropic ran a TV ad during the Super Bowl, which obviously isn’t something a single person can handle alone. The copy and brand assets his workflows rely on were initially produced by the product marketing and copy teams, with Claude generating variants and scaling testing on top.

Recently, Austin added some background on LinkedIn. He pointed out that the widely circulated article describes his experience as the sole growth marketer in Q2 2025, nearly eight months ago. The team did expand later, though still much smaller than the outside perception. As he put it, “Our combat effectiveness far exceeds our headcount.”

Even so, the signals are strong. A company valued at $380 billion post-investment, with annual revenue of $14 billion, during its fastest growth phase, had a marketer with no coding experience managing core growth channels for ten months with good results. This should already demonstrate that AI’s amplification of knowledge workers’ capabilities may be much greater than what our current organizational structures and hiring habits suggest.

How broadly this model can be replicated remains uncertain. Growth marketing is highly data-driven, process-oriented, and API-friendly, making it naturally suited for automation. In fields requiring more human judgment or creative intuition, the situation could be quite different.

Anthropic’s white paper offers three recommendations at the end of the growth marketing chapter: identify repetitive workflows with API interfaces for automation; break down complex processes into multiple specialized sub-agents rather than trying to handle everything with a single prompt; and before coding, thoroughly think through the overall process design in Claude. Essentially, these tips emphasize that the bottleneck in efficiency often isn’t technical ability but whether you’re willing to spend time dissecting your workflows and delegating parts that can be machine-managed.