Rebuild a cold calling lead score with Claude and no ML background

Owner.com sells to independent restaurants, and its old lead scoring had stopped running. Managers were stacking about 30 Salesforce filters by hand each week, and nobody could explain the old point values. YG Hong, who had no machine learning background, spent two weeks with Claude building a new score that predicts which leads will connect a rep with the person who decides. Claude read the old system's docs in Slack, Notion and Google Docs and explored the data in Snowflake over MCP. In a test, three reps given only top scored leads reached decision makers twice as often as the rest. The score now runs daily, so each rep's list is refreshed every morning.