Data analysts
Analysis that shows its work.
CSVs and PDFs in Drive, extraction across models, and every number verified by code.
A model's arithmetic is a guess wearing a confident tone.
Sandboxed code runs the actual computation: totals reconcile, dates parse, outliers get flagged.
Fifty documents to extract, one analyst, one week.
Subagents process documents in parallel and merge results into one dataset you can audit.
The monthly report is the same job every month.
An automation reruns the pipeline on schedule and files the output in Drive.
Load the sources
CSVs, exports, and PDFs land in a Drive folder; every chat and agent reads the same set.
More about DriveAnalyze and verify
The model drafts the analysis; sandboxed Python runs against the real files and proves every number.
More about Code ExecutionScale it with subagents
One subagent per file runs the same extraction; the orchestrator merges a clean dataset.
More about SubagentsSchedule the report
The pipeline reruns on cron and files the report in Drive with the numbers already checked.
More about AutomationsWhat this replaces
Built from these features
Drive
Folders, previews, and lightweight versioning for AI work.
Code Execution
Sandboxed Python, Node, and shell: agents and you, side-by-side.
Subagents
Parallel subagents, isolated context, full observability.
Multi-Model Chat
GPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.
Automations
Delegated runs with schedules, bounded execution, and full progress traces.