Anthropic’s $1B to $19B growth run: how Claude became the fastest-growing AI product in history | Amol Avasare

We’re kicking off with a rocket-ship story: Anthropic’s revenue blasted from roughly one to nineteen billion in about fourteen months, and it keeps climbing. My guest is Amol Aversari, head of growth at Anthropic and a master of building momentum, with a personal arc that includes founding, banking, leading growth at Mercury and MasterClass, and recovering from a serious brain injury—he’s here to explain how this future is unfolding.

We started as the smallest, least funded player, so getting here feels improbable. It’s the hardest job I’ve had, and you have to throw out most old playbooks to keep up with the pace and the stakes.

How did you land this role at a company growing faster than any I’ve seen?

I cold-emailed Mike Krieger with a tight note on why Anthropic needed a growth team, using subject lines and timing I’ve honed for years, and followed up until I got a response. They weren’t even hiring yet, but the timing clicked and one thing led to another.

What’s it like steering growth inside that kind of surge?

Research is the engine—our models and infra teams deserve most of the credit—while growth tries to keep pace. Internally we live on log scales, the line keeps tilting up, and even those big public revenue numbers are already stale.

With models this strong, why does a growth team matter, and what wins have moved the needle?

We still own acquisition, activation, routing, monetization, and pricing, but about seventy percent of our time is handling “success disasters” where growth breaks something else. The rest is classic growth work—prioritizing products, packaging, and timing when to lean in.

That import-your-memory move from ChatGPT was clever—what was the intent?

Activation is the choke point, so we focused on killing the cold-start by making Claude understand you sooner and guide you to value fast.

How critical is activation for AI, and what actually works?

It’s everything, and it’s hard because capability jumps make yesterday’s learnings obsolete. The durable play is to add the right friction that profiles a user, routes them to the right feature, and sets clear expectations.

Any onboarding changes that materially lifted activation?

At Mercury we spent a quarter fixing quality issues end-to-end and saw a big completion lift; sometimes splitting dense forms and removing tiny annoyances beats any shiny experiment. At Anthropic we ask who you are and what you’re trying to do, then tailor the experience—even if it lengthens the flow—because the data shows it pays off.

So how do you think about good versus bad friction?

Cut friction that adds no value, but keep steps that reduce cognitive load and personalize the path; that’s why flows like MasterClass and Calm use quizzes, and why breaking one busy screen into two often improves conversion.

If I want to teardown competitor onboarding today, what would you use?

Co‑work with the Chrome extension can crawl flows, capture evidence, and summarize what’s working and what’s not.

How big is your growth org, and how’s it structured?

About forty people across platform and monetization horizontals, plus pods focused on audiences and products like B2B, Claude Code, knowledge work, and the API.

And in terms of functions and where you place your bets?

We’ve got engineers, designers, PMs, and data, and we bias toward big swings over micro-optimizations. Our team even built the Chrome extension that now powers key co‑work and coding use cases.

Should other AI teams also bias toward big bets?

If AI is your core value, yes—product value compounds quickly, so aim for leapfrogs; if AI’s just a side feature, you may not need to swing as hard.

You mentioned automating growth—what’s actually running?

We’re piloting a system called Cache that uses Claude to suggest, build, test, and analyze small growth experiments; it’s early but already delivers wins like a solid junior PM. Humans still gate brand and handle stakeholder alignment, which remains stubbornly human.

So it proposes and ships tests with human review, and gets better over time?

Roughly that; we encode brand rules as skills to reduce review, and track week-over-week time saved and win rates across the loop.

How are PM, engineering, and design changing with AI in the mix?

Engineers get the biggest lift from tools like Claude Code, which squeezes PM and design capacity; we deputize product-minded engineers to PM sub–two-week projects while PMs focus on larger, cross‑functional bets.

So PMs should spend less time coding and more time raising the bar?

In bigger orgs, yes—insight, prioritization, and telling the team what matters is higher leverage than shipping one more small feature; still prototype to think with your hands, and keep docs lightweight.

How do you kick off fast without heavy PRDs?

Tiny changes run in Slack and ship; for larger efforts we do a crisp cross‑functional kickoff and spin up a short co‑work draft to align legal, safeguards, and partners.

I keep imagining a strategy bot that watches everything and nudges the plan—how close are we?

Close; co‑work already monitors charts, books rooms, files expenses, triages my inbox, finds misalignment across Slack, and sends weekly manager-style coaching summaries.

You even ask for feedback in your manager’s voice—how does that work?

Claude builds a profile from public writing and internal context, then mirrors what that leader would flag for me; it’s not perfect, but it has surfaced misalignments that would have burned weeks.

How can listeners set this up?

Install the co‑work desktop app, connect the Slack MCP with the right permissions, and then just ask it to run the workflows you want on a schedule.

Anthropic’s focus on B2B and coding feels unusually disciplined—who set that tone?

Leadership did from day one; even in 2021 there were docs arguing to go deep on coding to speed research, and constraints forced us to focus; we also delayed an early chatbot for safety, which nudged our path.

It’s wild how far you’ve come—what explains that?

Extraordinary leadership and talent density, full stop.

Why bet so hard on coding beyond market size?

Great coding tools accelerate researchers, which improve models, which then make coding even better—a tight loop that compounds.

How do you balance growth with Anthropic’s safety mission?

We’re a public benefit corporation, so we’ll pass on growth that risks safety, brand, or user trust; we’re comfortable leaving money on the table now to earn durable growth later.

For PMs and growth folks anxious about AI, what should they do to stay valuable?

Use the tools and retry workflows after every model jump, lean into your superpower and go interdisciplinary, and stay adaptable because old playbooks won’t hold.

Anything else about Anthropic’s culture worth highlighting?

It’s intensely mission-driven, open, and high trust—our notebook channels make leadership thinking visible and debate welcome—and the talent density feels like an all‑star roster; that culture is the real moat.

Quick tangent: you still run on Slack, Figma, Workday, and friends—does SaaS survive this wave?

We build some things ourselves, but we rely heavily on those tools and don’t see that changing soon.

People question your safety messaging, but it sounds deeply held.

We think in exponentials, so we speak plainly about both risks and upside, and we stay in the arena to pull the race toward higher standards.

Let’s visit failure corner—what’s a time things went sideways, and what did you learn?

My mental health startup ran for three years, raised money, hired a team, and then we shut it down; sending honest updates while failing was brutal, but that experience forged the skills and resilience that power my work today.

I love how your story flips the 'it’s all over' career spiral on its head, and it reminded me of the guest‑post call I put out before parental leave that drew more than five hundred pitches; yours—about how a traumatic brain injury reshaped you as a product manager—stopped me cold, so please share that journey.

In early 2022 a routine sparring session went wrong, and the concussion wrecked me for months—I couldn’t handle screens, music made me nauseous, my wife had to run my life, and for a while I wasn’t sure I’d ever work again; I clawed back through careful exposure, then got re‑injured in 2023 by an overhead bag right after joining Mercury, which benched me again, and I’m still not fully healed.

Oddly, it made me better: I cut alcohol and caffeine, take real breaks even on launch days, and retreat to meditate—Spirit Rock changed my life—so I can hold awareness apart from the chaos and keep a steady head without unhealthy crutches.

It’s wildly inspiring—yes, from a kick to the head to leading growth—and the constraint you live with has become a feature, not a bug, which feels like a lesson for all of us.

Freedom through constraints is real: you can fight reality and suffer or accept it, do everything you can, and refuse to let outcomes control your happiness.

A teacher put it simply for me: learn to be content even when you don’t get what you want, which I’m still practicing every day.

Lightning round time—what books do you recommend most?

Joy of Living helped me reshape how I relate to experience, Awareness complements that from a different angle, and Thinking in Bets sharpened my probabilistic thinking for product calls.

I caught a recent movie that was delightfully unhinged and great, and I haven’t watched TV in a while beyond the Olympics.

Name a product you love right now.

A bead‑filled pillow I found at a hotel in Japan—its height shifts as you move, which eased my neck pain and works whether I’m on my back or side.

Two mottos I lean on are “she’ll be right” and “just go for it,” the latter drilled into me by Eros Resmini from Discord.

Perfect combo—final one: new hobby since stepping away from martial arts?

Time is tight, so I prioritize work, health, and my closest relationships, and I’ve fallen hard for watching football.

My wife took me to the Big House and I became a Michigan Wolverine on the spot, and I’m also a die‑hard 49ers fan.

This was fantastic—where can people find you, apply to your team, and be helpful?

I’m on LinkedIn under my full name, and we’re hiring across growth, engineering, product, and design.

Try our products, send us frank feedback, and point great people our way.

Dream job for many—thanks for listening, and if you found this useful, subscribe on Apple Podcasts or Spotify, drop a rating or review, and find every episode at lennys podcast dot com; see you next time.