Amol Avasare says he cold-emailed Mike Krieger, then Anthropic's chief product officer, after heavy Claude use, landing the growth role before a team formally existed.
Anthropic runs CASH, a Claude-powered system that finds, builds, tests and analyses growth experiments, and Avasare says it performs like a junior PM with a human still approving experiments.
Avasare says Anthropic will leave money on the table to protect its AI safety red line, predicting that safety stance becomes a long-run competitive advantage.
Anthropic's growth team lets Claude run its own experiments
Anthropic's growth team is betting that AI products improve so fast that growth work should favor large bets and automation. Models outpace what users know to do with them, so activation matters more, and Claude already handles junior PM-level experiment work. Success came mainly from research, with safety held as a red line.
Takeaways
Activation matters more as AI models improve. Models improve faster than people discover what they can do, which Amol Avasare calls capability overhang. The right friction usually lifts conversion, as with MasterClass's purchase quiz and Mercury's onboarding fix. Anthropic's interest questions stay because the data supports the long flow.
Claude now runs growth experiments itself. The CASH project, led by engineer Alexey Komissarouk, uses Claude to find opportunities, build, test and analyse growth experiments. Amol says it performs like a junior PM two or three years in, and a human still approves each experiment.
Anthropic's growth team favors big bets. Product value should rise by orders of magnitude, so the roughly 40-person team tilts 70-30 or 50-50 toward larger bets. A grocery app might gain 30 to 50% in two years. AI-first firms like Lovable and Cursor should copy this.
Engineers gain leverage while PM work shifts. Claude Code lifts engineers more than designers or PMs, so engineers act as mini PMs on projects of two engineering weeks or less. Between 60 and 80% of shipped work has no PRD. PMs stay accountable on larger, cross-functional projects.
Research drove success; strategy came from necessity. Amol says growth deserves little credit. Anthropic lacked funding and distribution, so it focused on coding and B2B, following a 2021 Ben Mann document. It held back an early chatbot for safety reasons, and ChatGPT's launch pulled it toward consumer.
Safety is a red line for growth. As a public benefit corporation, Anthropic treats AI safety as off limits for growth experiments, and growth teams should accept leaving money on the table. Amol predicts the safety stance will become a significant competitive advantage over the long run.
How the cold email landed the job
Amol was a heavy Claude user, and Anthropic listed no growth PM roles though it was starting to consider a growth team.
He sent a cold email to Mike Krieger, then chief product officer.
Krieger replied that he was interested and wanted to talk.
Amol advises following up a few times, and continuing until the recipient asks to stop.
Tap a name to go there · tap a line to see why they are linked