A data scientist just demonstrated what we've been waiting for: AI that handles complete workflows, not just code snippets. Using Codex and Model Context Protocol (MCP), they processed 1.85GB of Apple Health XML data—from Google Drive download through BigQuery analysis—in 30 minutes. The AI located files, referenced six-year-old GitHub code, wrote Python parsers, uploaded datasets, ran SQL queries, and generated a structured report. What would have been "at least a full day" of manual work became a guided conversation with an AI agent.
This matters because it represents the shift from AI as a coding assistant to AI as a workflow participant. While tools like Cursor and Claude help write individual functions, this approach connects Google Drive, GitHub, BigQuery, and analysis tools through MCP—essentially creating AI agents that understand your entire data infrastructure. The developer explicitly chose Codex over their usual Claude to test the tooling, suggesting these workflows are becoming standardized across providers.
No other coverage exists yet, which is telling. The data science community tends to be skeptical of AI hype, but this feels different—it's a working example with specific timelines, file sizes, and tool integrations. The author admits this was a "simple example" with personal health data, not enterprise-scale complexity with governance requirements, data quality issues, or regulatory constraints.
For developers building AI workflows, this validates the MCP approach for connecting disparate systems. The real test isn't whether AI can analyze clean datasets—it's whether it can handle the messy, real-world data engineering that usually kills side projects. Based on this example, we're getting close.