How Zapier gives its GTM teams and their agents reliable access to company data

AI makes it easy for anyone to query company data. But trustworthy answers depend on how that data was organized, defined, and curated first. A lot can go wrong, if GTM teams are just pointing LLMs at data systems and pulling numbers. A CRM runs transactions, not broad analysis. So, pointing an LLM directly at one can produce answers based on incomplete data, unvetted relationships, or unclear definitions. A data warehouse is a more traceable source for analysis. Call transcripts need similar preparation. Searching thousands of raw transcripts can over-weight a few vivid examples. Semantic tagging helps AI find evidence across the full collection, but teams must still measure how common a pattern is and which customers it represents. Zapier's data team manages roughly 7,600 production tables in Databricks. Embedded teams narrow that catalog into the fields, metrics, relationships, and context AI needs for specific questions. Validation remains human work. When Zapier built its Sales 360 Genie, team members checked 100 question-and-answer pairs against trusted sources. Transcript classifiers need the same review before teams can trust their conclusions and make important decisions. Join Alex Horner and Sergio Santos, two leaders from Zapier's data team, for a candid conversation about centralized data, embedded curation, transcript analysis, human validation, and governed AI access. They’ll talk all about how they’ve partnered with Zapier’s GTM and RevOps teams to democratize data.