Salesforce reported second-quarter revenue of $11.3 billion, up 11% year over year. ^[1] Salesforce stock closed the week of August 24 up 22.4%, marking its sharpest weekly gain in years. ^[2] The article argues that the hardest part of building an AI agent is often getting CRM data ready, because AI agents inherit existing data quality problems rather than cleaning them up. ^[3] Salesforce and Anthropic announced Claudeforce, embedding Claude models across enterprise workflows. ^[4] Salesforce management raised full-year guidance to a range of $46.1 billion to $46.4 billion. ^[5] The article defines AI-ready CRM data as data that is accurate, complete, consistent, unique, current, structured, and connected, meaning the information an agent depends on reliably supports the decisions it is designed to make. ^[6] The article's first cleanup step is to audit the CRM before deleting anything, identifying duplicate contacts and companies, missing fields, invalid emails, outdated opportunities, and inconsistent lifecycle stages, and prioritizing problems that could cause an AI agent to make wrong decisions. ^[7] The article's second step is to resolve duplicate records carefully by establishing matching rules (email, phone, or name plus company for contacts; domain for companies) and survivorship rules, because similarity does not automatically mean duplication and automatic merging can destroy valid customer history. ^[8]