According to @AndrewYNg, rapidly improving AI agents make cross-system data access critical, turning vendor-created silos into costly barriers for value creation in enterprise AI workflows, which he says is why buyers should favor software that lets them control and route their own data to agents for analysis, source: Andrew Ng on X, Nov 6, 2025. He reports one SaaS vendor sought over $20,000 for an API key to access his team’s own customer data, which he characterizes as a switching-cost tactic that blocks agentic workflows and slows decision automation, source: Andrew Ng on X, Nov 6, 2025. He advises businesses to organize unstructured data such as PDFs and to prioritize data portability, citing LandingAI’s Agentic Document Extraction and his advisory work at AI Aspire promoting data control in tooling decisions, source: Andrew Ng on X, Nov 6, 2025; source: deeplearning.ai The Batch, issue 326. He adds that individual data ownership enables agent workflows, noting he uses Obsidian because notes are Markdown files he can read and write with his own agents, source: Andrew Ng on X, Nov 6, 2025. Ng does not mention cryptocurrencies, but the enterprise pain points he documents — costly data access and portability needs — align with open, permissionless data access and indexing architectures that AI agents can consume, which is relevant context for crypto investors focused on AI-data infrastructure, source: Andrew Ng on X, Nov 6, 2025.