{"url":"https://youtu.be/91AJ0cpgLlQ?is=TQceVqxIYP2Olepq","title":"Anthropic PMs use Claude for instant data analysis","domain":"youtu.be","imageUrl":"https://img.youtube.com/vi/91AJ0cpgLlQ/maxresdefault.jpg","pexelsSearchTerm":null,"category":"Tech","language":"en","slug":"cc65100c","id":"cc65100c-f760-425c-a946-396488936522","description":"Anthropic product manager Lisa Crofoot explains using Claude AI for fast data queries and evals without SQL.","summary":"## TL;DR\n- Anthropic product manager Lisa Crofoot explains using Claude AI for fast data queries and evals without SQL.\n- Claude connects to BigQuery tables to analyze product data like dark mode usage trends in minutes.\n- This lets PMs work independently, focus on strategy over coordination, and tackle tasks they couldn't before.\n\n## The story at a glance\nLisa Crofoot, a product manager at Anthropic, shows in this short video how her team uses Claude to query product data and build AI evaluations quickly. The company connected Claude Code to BigQuery tables, so PMs skip SQL and data science waits. It's part of Anthropic's \"How we use Claude\" series, posted as their tools evolve rapidly.[[1]](https://youtu.be/91AJ0cpgLlQ?is=TQceVqxIYP2Olepq)[[2]](https://www.youtube.com/watch?v=91AJ0cpgLlQ)\n\n## Key points\n- PMs traditionally ping data scientists or write basic SQL on unfamiliar databases, causing delays.\n- Anthropic's data team set up a BigQuery MCP to link product data tables directly to Claude Code.\n- Demo uses synthetic data to plot dark mode usage over three months, with Claude auto-adding 7-day rolling averages and overall averages.\n- Quick iteration: PM asks to plot light/dark mode by plan type; Claude edits the graph after permission.\n- Claude generates evals by taking a product scenario and a few test examples, expanding them to 50 cases fast.\n- Frees PMs to spend more time on strategy, customer talks, and decisions instead of operations.\n\n## Details and context\nClaude acts as a human-friendly interface: PMs describe what they want in plain language, and it handles the SQL and visualization behind the scenes. In the demo, Crofoot generated synthetic data on light vs. dark mode first, then explored fractions over time—tasks that would take hours manually.\n\nFor evals, which test AI systems, Claude \"expands the option space\" of test cases from one or two to dozens, making product validation faster.\n\nThis fits Anthropic's push to use their own AI internally, as shown in their news post on PM workflows.[[1]](https://youtu.be/91AJ0cpgLlQ?is=TQceVqxIYP2Olepq)\n\n## Key quotes\n- \"What excites me about being a product manager now is that I can move and iterate much faster. I can use Claude to test my product ideas before I even get anyone else in the loop.\"[[1]](https://youtu.be/91AJ0cpgLlQ?is=TQceVqxIYP2Olepq)\n- \"It's more than automation for me. It's things that I wouldn't have otherwise necessarily even been capable of doing independently.\"[[1]](https://youtu.be/91AJ0cpgLlQ?is=TQceVqxIYP2Olepq)\n\n## Why it matters\nAI tools like Claude are changing how teams handle data and testing, cutting out specialist bottlenecks in fast-moving tech firms. For product managers, this means quicker insights and experiments, less waiting, and more focus on high-level work. Watch for more in Anthropic's internal-use series or similar tools from rivals like OpenAI.","hashtags":["#ai","#productmanagement","#anthropic","#claude","#bigquery","#tech"],"sources":[{"url":"https://youtu.be/91AJ0cpgLlQ?is=TQceVqxIYP2Olepq","title":"Original article"},{"url":"https://www.youtube.com/watch?v=91AJ0cpgLlQ","title":""}],"viewCount":3,"publishedAt":"2026-04-09T11:34:48.364Z","createdAt":"2026-04-09T11:34:48.364Z","articlePublishedAt":"2026-03-26T00:00:00.000Z"}