AI Agents and the Fight for Customer Data

2026-06-05 · Source: The a16z Show · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cloud Computing & IT Infrastructure · Depth: Intermediate, extended

Summary

George Fraser, cofounder and CEO of Fivetran, discussed the future of data infrastructure with Martin Casado, emphasizing the critical role of centralized data foundations for AI agents. Fraser highlighted that while companies historically built data infrastructure for business intelligence, the new imperative is providing context for AI. He addressed concerns about AI agents threatening enterprise software, arguing that many companies, like SAP with its new API policy, overestimate this risk, comparing it to past debates over open APIs in the 1990s. Fraser asserted that "data gravity" is a misconception and advocated for open data access, citing Fivetran's open data infrastructure.com benchmark. He also shared Fivetran's internal use of AI for troubleshooting across its 750 connectors and discussed the company's acquisition strategy, including the dbt merger, positioning dbt as a major beneficiary of AI coding agents.

Key takeaway

For AI Architects and Data Engineers building agent-based systems, you must prioritize establishing a robust, centralized data foundation. Insist on full data access from SaaS vendors, leveraging contractual language and benchmarking tools like open data infrastructure.com to avoid "pre-internet ChatGPT" limitations. Your focus should be on enabling comprehensive context for AI, not just cost reduction, as open data access remains critical for agent efficacy and long-term business intelligence.

Key insights

Centralized, open data infrastructure is crucial for AI agents, despite vendor data access restrictions.

Principles

• AI agents need centralized, up-to-date data context.
• Data access restrictions by vendors are ultimately unsustainable.
• Data gravity, based on egress fees, is largely a misconception.

Method

CIOs should insist on controlling a complete data copy in a data lake, leveraging contractual language in MSAs. AI coding agents can act as "infinite junior engineers" for complex, long-tail troubleshooting.

In practice

• Benchmark vendor data access policies using open data infrastructure.com.
• Integrate data access guarantees into large vendor MSAs.
• Treat enterprise AI agents as employees for workflow integration.

Topics

• AI Agents
• Data Infrastructure
• Open Data Access
• Enterprise SaaS
• Data Integration
• Fivetran dbt Merger

Best for: CTO, VP of Engineering/Data, AI Product Manager, Director of AI/ML, AI Architect, Data Engineer

Related on AIssential

• The End of Software’s Monopoly on Work — AI agents shift enterprise work from applications to a trusted data context layer, demanding unified, governed data for reliable action.
• Snowflake's AI Bet & The Future of SaaS — Snowflake's existing architecture and business model are uniquely suited for the AI era, creating a strong competitive advantage.
• not much happened today — AI performance increasingly depends on context pipelines and open harnesses, not just model weights.
• D&B's database of 642 million businesses was built for humans, not AI agents. So they rebuilt it. — Enterprise data systems built for humans require fundamental redesign for AI agents.
• SaaS Is Not Dead Yet — AI agents enable individual tool creation but lack enterprise-grade collaboration and data sharing capabilities inherent in traditional SaaS.
• AI Applications: Tools, Use Cases, and Platforms — Effective AI adoption requires careful tool selection, robust integration, and strong governance, balancing cost, performance, and ethical considerations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The a16z Show.