The interesting part is when the agent stops being a toy bolted onto one table and starts working inside your actual data stack.
Google just shipped exactly that with Data Agent Kit — not a chat window over a spreadsheet, but a working environment for your agent, right inside whatever IDE you already use (antigravity, Cloud Code, Codex, Gemini CLI, doesn't matter which). One click enables secure MCP connectors to BigQuery, Cloud SQL, AlloyDB, Spanner, and Knowledge Catalog — the agent gets access to your whole data estate, not one table in one chat.
There's a second piece people confuse with the connectors: skills. Not data access — a playbook. Markdown files with best practices for how to actually write correct BigQuery or Spark queries. MCP is the hands. Skills is the instruction manual for using them right.
I mentor data analysts, and this is exactly the shift I keep telling them to get ahead of. In real companies, data never lives in one clean place. Some in the warehouse, some in a transactional database, some in files sitting in Cloud Storage. The question isn't "can AI answer questions about data" anymore. It's "can you give it a real working environment where it actually does a piece of real analytical work."
And looks like the answer is yes.
Want a stronger data analyst role or a raise? Grab the FREE Product Analyst Playbook and get the exact roadmap to your next offer.
