Anthropic makes $6.6M per employee


Hello Reader

When I was a SaaS marketing VP, our company was working toward $1 million in revenue per employee.

If we weren’t hitting that number, then we weren’t approved for new headcount. Which meant I needed to push my team to be as productive as possible.

Anthropic recently reported more than $6.6 million in revenue per employee for the last 12 months on revenue of $16.3 billion.

I can imagine the conversation in leadership meetings now: “If Anthropic can do that, why can’t we get to $3 million? We have AI too.”

But it’s not that easy.

​Johan Abadie, VP Demand Generation & Revenue Strategy at SalesIntel says, “I mean, I'm driving a Camry, they are driving a Lamborghini. There is no way I win that race.”

Anthropic’s advantage is that it develops the technology its employees use, and operates in an unusual market, but most marketing leaders are working with established teams, existing systems and budgets that come with plenty of constraints.

Still, Anthropic’s practices offer useful lessons about team structure, skills and responsibilities that we can apply to our own teams— regardless of our situation.

How Anthropic’s workforce is changing

Anthropic’s internal employment research tracks how work changes as employees use more AI.

Four of these changes have direct implications for how we staff, manage and develop our own teams.

Here they are:

Broader roles need clear ownership

Growth marketer Austin Lau, whose work I’ve covered previously, is a useful example.

He used Claude Code to build a Google Ads copy agent.

He feeds the agent campaign performance data, existing ads and keywords, and Claude drafts headlines and descriptions using brand skills and product knowledge docs.

Austin reviews and refines the drafts, then Claude packages them into a file ready for upload to Google Ads.

His role now includes building the tools he uses to run campaigns, with colleagues contributing expertise when needed.

None of this was in Austin's original job description.

For a marketing team of five or more, expanding responsibilities and cross department workflows creates a challenge.

People can take on work that previously belonged to another specialist—but that makes assigning accountability a bit murky.

Consider a demand generation manager who can now produce landing pages, email sequences and campaign analysis.

  • Who approves the product claims?
  • Who checks the analysis?
  • Who edits and proofreads the copy?

Each responsibility needs an owner with time to do the work.

Otherwise, your copy editor could be buried in landing page reviews while the demand team celebrates how much faster it’s producing them.

One team’s productivity gains have become another person’s backlog. That’s how resentment builds, especially when the extra work goes unrecognized.

Identify the capacity gap before hiring

As AI increases output, your team needs time to review the work and maintain the supporting systems, but before adding headcount, identify where work gets stuck.

Do you need another producer, a specialist reviewer or someone to maintain your AI systems?

That answer will tell you who to hire next or what you need to train for.

If we go back to the previous example, it's clear the work is getting stuck in copy review. You would either train your product marketers and demand gen specialist to review their own copy, or you hire another copy editor.

Teach the whole team to direct and evaluate AI

Anthropic teaches all employees the 4D AI Fluency Framework during onboarding and expects them to apply the rules to their work.

So what are the 4Ds?

It gives people a shared way to make decisions:

  • Delegation: Which parts of the work should AI handle, and where should people stay involved?
  • Description: What context, requirements and examples does AI need?
  • Discernment: Is the result accurate, relevant and good enough to use?
  • Diligence: Are we using data responsibly and taking responsibility for the work we share?

If your team already have the foundations then consider role specific training:

  • Campaign leads need practice directing work and coordinating decisions.
  • Specialists need to assess outputs against professional standards.
  • Marketing operations needs time to maintain integrations, permissions and shared systems.

Tell people what productivity gains mean for their roles

If AI is giving team members untapped capacity they need to know what to do with that time.

Do you expect them to conduct more customer research, do additional campaign testing or a take on a broader role?

Someone who builds and maintains a system used by six colleagues has taken on responsibilities that deserve space in their workload and development plan.

As tasks change, agree on what each person will take on, what they’ll stop doing and what support they’ll receive.

Work with HR where those changes affect job descriptions, progression or compensation.

People need a credible picture of their future as they help improve the work.

Review your team’s capabilities before changing its size

For your next team planning session, answer these four questions:

Compare the answers with your team’s delivery time, rework, campaign results and workload.

Revisit them monthly as people’s capabilities grow.

Anthropic’s revenue figure will attract attention in boardrooms, but your team’s progress will show up in better decisions, stronger campaigns and fewer delays.

In my 90-minute workshop, I’ll walk you through a Claude workflow that tracks your competitors, delivers a weekly briefing and turns that research into the final drafts your team can review.

You’ll get my setup guide, Claude Skill templates and both workflow templates so you can adapt them to your brand.

I’ll show you the system in action using two real marketing workflows:

  1. A market intelligence agent that tracks your competitors and category, and delivers a weekly briefing.
  2. A content engine that turns that briefing into email, newsletter and social drafts in your brand voice.

A week later, I’m hosting a 60-minute build clinic and we fix anything that didn’t work.

We have a great group of marketers already signed up—come join us!

You get $100 off with the code AIATWORK.

​Register now.​

Did some one forward you this email? You can subscribe here.​

2120 Contra Costa Blvd #1059 , Pleasant Hill, CA 94523
​Unsubscribe · Preferences​

AI at Work

AI at Work is a weekly newsletter on how marketing teams redesign workflows, roles, and systems with AI. Real examples, practical frameworks, and repeatable processes operators can use immediately. Join thousands of successful marketing leaders by subscribing below!

Read more from AI at Work
AI and performance reviews

Hello Reader A quick note: Next week, you’ll notice a new sender email address: tahnee@perryco.ai. It’s still me and your weekly AI at Work. Add it to your contacts so you never miss an issue! ~ Performance reviews are coming up, and AI can make the preparation considerably easier. It can gather project records, summarize feedback and produce a polished draft in minutes. That’s appealing when reviews have long been criticized for taking too much time and doing too little to improve...

AI ROI

Hello Reader Imagine hiring someone to redesign your kitchen. They arrive with a beautiful plan, a list of expensive appliances and a date to start work, but they haven’t asked how you use the space or what frustrates you about the kitchen you have. Would you approve the plan? As AI advisor Liza Adams says, “It makes more sense to reimagine the work first. A built-in pet-feeding station and an eight-burner range might look impressive, but why would you put them in a kitchen for a family that...

Using AI

Hello Reader I teach AI for a living, and I still sometimes wonder whether I’m learning the right things. You might already use AI for research, campaign planning and content production. You’re getting better results. Then someone shares an AI agent that connects their customer data, runs research and produces a campaign plan, and suddenly your own work feels basic. Now you’re wondering whether to improve what you’ve built, learn a different tool or investigate something you hadn’t considered...