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 current remaining performance obligations climbed 14% in constant currency to $33.5 billion. ^[5] The author asserts that the answer to the question of AI agents writing to CRMs without human oversight is not a simple yes or no. ^[6] The author states that the question of whether AI agents should write to CRMs without human checking is framed incorrectly. ^[7] 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. ^[8] The article states the relationship 'clean data → reliable automation → an AI agent you can trust with real work,' and that breaking the first link makes everything after it harder to trust. ^[9] 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. ^[10] The article suggests that the issue of AI automation in CRM writing requires a nuanced approach rather than a binary decision. ^[11]