AI Agents and the Fight for Customer Data

The episode opens with a clear premise: modern AI agents can only act intelligently when they have real‑time access to a company’s unified data store. George Frazier explains that Fivetran’s long‑standing role—replicating data from SaaS tools like Salesforce, NetSuite, and custom databases into a single warehouse—has evolved from supporting traditional business intelligence to powering contextual AI workflows. This shift turns data platforms into the nervous system of enterprises, enabling agents to answer queries, automate processes, and augment human teams with up‑to‑date business knowledge.

A major tension emerges around SaaS vendors’ reactions to this new demand. Some, notably SAP, have begun restricting API access for AI agents, effectively creating walled gardens. Frazier argues that such lock‑downs hurt customers, who need unrestricted data to build accurate models and dashboards. He also dismantles the “data gravity” narrative, showing that change‑data‑capture pipelines move only incremental changes, keeping egress costs trivial. The discussion highlights Fivetran’s OpenDataInfrastructure.com benchmark, which scores vendors on openness, egress fees, and contractual restrictions, flagging the worst offenders and encouraging industry transparency.

For CIOs, the takeaway is actionable: insist on a copy of all corporate data in a controllable data lake, and embed explicit data‑access language in high‑value contracts—especially those exceeding $500 million. Leveraging model contract clauses from the OpenDataInfrastructure repository can shift the balance of power back to the buyer. As AI agents become standard components of enterprise workflows, open data pipelines will remain the competitive advantage, ensuring that organizations can harness AI without being trapped by proprietary API silos.