Head of Growth (Anthropic): Claude is Growing Itself at This Point
Head of Growth (Anthropic): Claude is Growing Itself at This Point
ELI5/TLDR
Anthropic’s head of growth, Amol Avasari, explains how the company went from $1 billion to $19 billion in annual revenue in 14 months — and the number is already out of date. He got the job by cold-emailing the CPO, runs a team of 40 that spends 70% of its time on “success disasters” (things breaking because growth is too fast), and has started using Claude itself to automate growth experiments. He also survived a traumatic brain injury that left him unable to walk for months, and credits meditation and forced constraints with making him better at the job.
The Full Story
How to Get Hired at the Fastest-Growing Company in History: Send a Cold Email
Amol was not referred. He did not apply on the website. He was not sourced. He cold-emailed Mike Krieger — the co-founder of Instagram, now Anthropic’s CPO — and said, essentially, “Great product, you obviously need a growth team, want to chat?” Krieger responded. Amol is the only PM Krieger has ever hired from a cold email. The role wasn’t even listed.
The cold email itself was methodical: a high-conversion subject line he’d tested over years as a founder, sent to a personal email (not LinkedIn, not work email, where everyone gets outreach), kept short. His rule of thumb on follow-ups: keep going until they tell you to stop. In this case, one email was enough.
The Scale of Things
The numbers bear repeating because they are genuinely difficult to process. Anthropic hit $1B ARR at the start of 2025, $4B by mid-year, $9B by year’s end, and $19B ARR by February 2026. That last figure, Amol notes, is already stale. For context: companies like Atlassian and Snowflake, built over 15-20 years, do $4-6B ARR. Anthropic adds that much every few months.
“The funniest thing is I’ve noticed internally linear charts are just not cool. Everything is log linear. Just show me at log linear scale.”
The 10x year-over-year trend has held since the beginning: 2023 was $0 to $100M, 2024 was $100M to $1B, 2025 was $1B to roughly $10B. When Amol joined in late 2024 for revenue planning, Dario pushed the aggressive case scenario. People were “freaking out.” Dario thought they could go higher. They did.
Success Disasters: The 70/30 Split
Amol’s team is about 40 people, structured like a traditional growth org — horizontals for growth platform and monetization, plus audience-focused pods for B2B, Claude Code, knowledge workers, API. The work categories are standard: acquisition, activation, monetization.
What is not standard is the ratio. About 70% of his time goes to what he calls “success disasters” — things breaking because everything is growing too fast. The charts are all green, all up and to the right, and everyone is still miserable because scaling at this rate creates cascading problems.
“You need to sort of step back and just realize like we’re very lucky to have these problems.”
The remaining 30% is more conventional growth work: deciding which products to invest in, long-term pricing and packaging, figuring out when to lean in on new products like Co-work.
Why Anthropic Swings Big (and Why You Might Want To)
A traditional growth team might spend 60-70% on small-to-medium bets and 20-30% on big swings. Anthropic flips it — 50/50 or even 70/30 toward larger bets. The logic is mathematical: if you’re a grocery delivery app, your product value might increase 30-50% in two years. If your core product value is driven by AI models on an exponential improvement curve, the product value in two years is 100-1,000x what it is today. Small optimizations can capture a meaningful share of a linear increase. They cannot capture an exponential one.
The Chrome extension, now foundational to Co-work and Claude Code, was built by the growth team. A bullish engineer, nobody else doing it, so they did. That is not a move Amol would have made at a previous company.
His caveat: this advice applies specifically to AI-first products where the core value proposition rides the exponential. If AI is a side feature, the standard playbook still holds.
Activation: The Right Friction Helps
Importing ChatGPT memories was a clever acquisition play. But the deeper challenge is activation — the cold start problem. The industry suffers from “capability overhang”: models improve faster than products can diffuse those improvements to users. You might have something approaching AGI, but if people’s instinct is to ask “what’s the weather in SF?”, the capability is wasted.
At Mercury, Amol’s team spent an entire quarter ignoring metrics and just fixing quality in the onboarding flow. Forget conversion. Just make it good. It ended up being the single most impactful quarter he’d had as a growth PM until joining Anthropic.
His principle on friction: cut friction that doesn’t help users understand why the product is for them. But if asking a few extra questions helps you recommend the right feature to the right person, add that friction without apology. Masterclass does it. Calm does it. Anthropic does it. The data from those onboarding questions flows downstream into lifecycle marketing, lookalike targeting, everything.
“The most simple understanding people have is like just solve time to value. Cut all the steps and just get them into the product. And like that doesn’t work most times.”
Claude is Automating Growth Experiments
The growth platform team runs an initiative called CASH — Claude Accelerates Sustainable Hypergrowth (Amol did not name it; he wants you to know this). The idea: use Claude to automate the growth experimentation loop.
There are four stages to evaluate: identifying opportunities, building the feature, testing quality and brand compliance, and analyzing results post-ship. They score Claude on each. Right now it handles copy changes and minor UI tweaks. The win rate is roughly junior PM level — a 2-3 year PM, not a senior one.
“You can push it, press play with it, and it ultimately prints money.”
Before Opus 4.5, it wasn’t viable. With Opus 4.6, it started working. Human review is still in the loop, partly for brand protection. But the amount of human time needed drops week over week. The piece that stubbornly remains human: cross-functional stakeholder management.
“We will have AGI and it will still be impossible to get six people in a room to get aligned.”
The PM-Engineer Ratio is Inverting
Claude Code has 2-3x’d effective engineering output. PMs and designers have gotten leverage too, but less. The result: a default team of five engineers, one designer, one PM now operates like 15-20 engineers, 1.5-2 PMs, 1.5-2 designers. PM and design are “absolutely squeezed.”
Anthropic’s response is twofold. First, hire more PMs (they are hiring). Second, deputize product-minded engineers as mini-PMs. The framework: if a project is two engineering weeks or less, the engineer is effectively the PM — talks to security, legal, stakeholders, drives it. The PM advises. Over two weeks, the PM takes the wheel.
For PMs, the question becomes: with 20 engineers, is the highest-leverage use of your time shipping the 21st feature, or making the other 20 slightly better-aimed? At larger companies, Amol argues for the latter.
“If you then get a really good PM who can improve the why and the what by 5%, that is such a high leverage.”
PRDs are mostly dead at Anthropic. Maybe 60-80% of what they ship has no PRD. For the 20-30% that really matters, documentation should be excellent. For everything else, Slack messages and a good kickoff meeting.
Co-work as a Management Tool
Amol has a Co-work task that runs every morning, checks 20-25 charts, and tells him what to pay attention to. He has another that scans Slack weekly for areas of potential misalignment across teams — people working at cross purposes, overlapping efforts, brewing disagreements.
He uses Claude to review his direct reports’ work against team OKRs and suggest feedback. He even has it simulate feedback from his own manager, Ami Vora, by building a model of her priorities from her public writing and Slack behavior.
“You’re like working with a coach who’s kind of like drunk at times. Not drunk, but… sometimes says something you’re like, why would you bring that up? But other times you’re like, wow.”
Claude also books his meeting rooms, does a first pass on his inbox, files reimbursements in BenPass, and handles expenses in Brex.
Focus as Competitive Advantage
Anthropic had a chatbot before ChatGPT launched. They chose not to release it for safety reasons — they didn’t want to kick off a global AI arms race. ChatGPT launched, got massive traction, and pulled OpenAI toward consumer. Anthropic went deep on B2B and coding instead.
This wasn’t just a commercial bet. A 2021 internal doc from co-founder Ben Mann argued for focusing on AI coding, years before anyone knew the market size. The thesis: better coding models accelerate research, better research produces better models, the flywheel spins. Dario has apparently held this view for years.
The necessity helped too. Anthropic was the smallest, least-funded player. No distribution like Meta or Google. No first-mover advantage like OpenAI. Constraints forced focus.
“Freedom through constraints — when you have a bunch of constraints applied on you, it can bring a lot of freedom because it frees up all this excess choice.”
Safety is Not Marketing
Anthropic is structured as a Public Benefit Corporation, not a standard Delaware C-Corp. This legally allows them to optimize for something other than shareholder value. The safety commitment is, if anything, understated externally relative to how seriously it’s taken internally.
“I think we actually believe in this stuff more strongly than we say externally.”
Amol frames the commercial approach: be comfortable leaving money on the table. Don’t squeeze every last dollar. Controversial tests get sorted into two buckets — those you should never run regardless of results (red lines, including safety), and those that make you uncomfortable but aren’t a red line (test it, but demand high returns for high cringe).
A Brain Injury, a Failed Startup, and a Pillow from Japan
Before any of this, Amol founded a mental health startup. Three years, a couple million raised, seven to ten employees. It failed. He describes calling investors to say he’d lost their money as brutal. It took years to get over. But without the founder experience, he wouldn’t have learned cold email, sales, product — the skills that made his career possible.
Then in early 2022, a routine Muay Thai sparring session went wrong. A kick to the head. Nine months off work. The first few months, his wife did everything including texting his friends. He couldn’t listen to music for more than 20 seconds without needing to vomit. Couldn’t look at screens. It was unclear if he’d ever work again.
A month after publishing a guest post about the experience in Lenny’s newsletter, he got re-injured — a bag hit his head getting off a plane. Two more months off work, one month into his Mercury job. He’s still not fully healed. Occasional dizziness and headaches.
He doesn’t drink alcohol or caffeine. Takes breaks twice a day, even during model launches. Does meditation retreats at Spirit Rock at least once a year.
“True freedom in life is learning how to be content when you don’t get what you want.”
His book recommendations: Joy of Living by Yongey Mingyur Rinpoche, Awareness by Anthony de Mello, and Thinking in Bets by Annie Duke. His life motto: “She’ll be right” (Australian for “it’ll be fine”) combined with “sometimes in life you just have to go for it.” His favorite recent product discovery: a Japanese pillow filled with beads that lets you adjust height unconsciously while sleeping. The Maruhachi Pro. Only available in Japan. He brought it back on the plane.
Claude’s Take
There is something genuinely unusual about this conversation. Most growth leader interviews are an exercise in attribution theater — claiming credit for things that would have happened anyway. Amol does the opposite. He repeatedly says the research team, the model quality, and the product teams deserve most of the credit. He calls his 70/30 split “firefighting success disasters” versus actual proactive growth work. This is either unusual honesty or very sophisticated signaling. Probably both.
The CASH initiative — Claude automating growth experiments — is the most interesting claim here. The “junior PM win rate” framing is specific enough to be credible. The timeline checks out: Opus 4.5 released in early 2025, so “a couple months ago” for not working aligns, and Opus 4.6 arriving more recently as the inflection point tracks. The honest admission that cross-functional stakeholder management remains stubbornly human is the kind of detail someone invents only if they’re telling the truth.
The revenue numbers ($1B to $19B ARR in 14 months) are externally corroborated and genuinely without precedent at this scale. The comparison to Atlassian and Snowflake is apt — those are decades-old companies doing less ARR than Anthropic adds quarterly.
The advice on friction in onboarding runs counter to the Silicon Valley orthodoxy of “reduce time to value” and is well-supported by his track record across three companies. The framework of “friction that helps users understand why the product is for them” is a more precise articulation than most growth practitioners manage.
Where things get softer: the culture section leans hagiographic. “I have not met a single person who’s checked out” is a strong claim for any company of Anthropic’s size. The Real Madrid analogy is charming but it’s also the kind of thing people say when they’re recruiting on a popular podcast. The safety narrative — “we believe it even more strongly internally” — is unfalsifiable from the outside.
The brain injury story is the emotional core and it’s genuinely remarkable. The connection between forced physical constraints and professional effectiveness is the most original insight in the conversation. Most productivity advice is about adding habits. This is about what happens when your body removes your choices and you discover the remaining options were better all along.
One thing worth noting: Amol is clearly an exceptional self-marketer (the cold email, the guest post pitch, this podcast appearance). This is not a criticism — it’s the job. But it’s worth accounting for when weighing claims about humility and leaving money on the table. The man brought a pillow back from Japan on a plane. He knows how to tell a story.