CHAPTERS
- 0:00 – 0:30
Faster product iteration with Claude as a PM copilot
Lisa explains how Claude changes the pace and independence of product management work. She highlights using Claude to pressure-test ideas early, without needing to pull in other teams.
- •Iterate on product ideas much faster with Claude
- •Test and refine concepts before involving other stakeholders
- •Increased ability to operate independently as a PM
- •Empowerment from reducing reliance on coordination
- 0:30 – 0:45
Solving the PM data-access bottleneck (without deep SQL)
She describes data retrieval as a common PM pain point, often requiring help from data science or hand-written SQL against unfamiliar schemas. Claude helps bridge that gap by translating questions into analysis.
- •PMs often depend on data scientists for queries and analysis
- •Writing SQL is possible but slow and schema knowledge is a barrier
- •Claude enables querying/analysis through natural-language interaction
- •PM becomes the “human interpreter” of results rather than the query author
- 0:45 – 1:00
Claude Code + BigQuery MCP: Connecting product data directly to the assistant
Lisa outlines the workflow infrastructure that makes this possible: Claude Code connected to BigQuery via an MCP integration set up by the data science team. This creates a direct path from PM questions to product data exploration.
- •Using Claude Code for data analysis workflows
- •Data science team set up a BigQuery MCP integration
- •BigQuery product tables become accessible within Claude Code
- •Reduces friction between question, query, and insight
- 1:00 – 1:15
Demo setup: Synthetic dataset and a concrete usage question
Before the demo, she has Claude generate synthetic product data to simulate light vs. dark mode usage. She frames a specific question: the fraction of dark mode usage over the last three months.
- •Claude generates synthetic product data for demonstration
- •Focus metric: fraction of dark mode usage
- •Time window: past three months
- •Directing Claude to the relevant data table
- 1:15 – 1:30
Instant visualization with helpful analytics defaults
Claude produces a graph immediately and adds enhancements Lisa didn’t explicitly request. The visualization includes a rolling average and overall average, showcasing high-quality analysis output by default.
- •Claude auto-generates a chart from the requested data
- •Adds a 7-day rolling average without prompting
- •Includes an overall average baseline
- •Quality of charting and analysis exceeds typical manual effort
- 1:30 – 2:00
Rapid iteration: Break down usage by plan type
Lisa asks for a segmentation by plan type to refine the insight. Claude requests permission to make edits, then quickly generates the segmented plot—compressing hours of work into moments.
- •Iterating from a single metric to segmented analysis
- •Request: plot light/dark mode usage by plan type
- •Permission/approval step before edits are applied
- •Significant time savings vs. manual analysis workflows
- 2:00 – 2:31
Using Claude to generate evals for AI product quality
Beyond analytics, Lisa describes using Claude to create evaluation sets (evals) for AI systems and products. Claude can expand a few example cases into a broad set of test scenarios quickly.
- •Evals are used to assess AI systems and AI products
- •Provide the target experience and a few sample test cases
- •Claude expands the option space of test cases
- •Scale from 1–2 examples to ~50 test cases rapidly
- 2:31 – 2:56
Strategic impact: More time on decisions, less on coordination
Lisa closes by emphasizing that Claude shifts PM time toward strategy and customer insight rather than operational overhead. She frames the benefit as capability expansion, not just automation.
- •Rebalance PM time toward strategy and customer conversations
- •Reduce coordination and operational busywork
- •Claude enables work PMs may not do independently otherwise
- •Viewed as capability extension rather than simple automation
