← Daniel Bloom/02 · AI Innovation
It's not enough to be 'good at AI.' To see organizational gains we need to change culture to have an AI-first bias-to-action, without diluting ownership of product and engineering fundamentals.
01
Problem
AI-assisted coding pushed the bottleneck downstream — QA couldn't keep up with the new merge rate.
Solution
A Claude Code portal delegates to multiple testing agents which read the Jira ticket, testing acceptance criteria using Playwright, validate performance and errors, and posts results back to the ticket — for cents per run.
02
Problem
Reviews were fragmented across Trustpilot and App store platforms, making it hard to identify trends and generate insights.
Solution
Claude + BrightData MCP scrape and structure reviews from every source, then surface trends and outliers in a single dashboard the team can actually act on.
03
Problem
Hundreds of key content articles had accumulated errors and branding issues, and no one had time to comb through them.
Solution
Claude crawls the blog, scores each post against a quality rubric, and proposes edits for human review — automating a multi-month manual project into hours.