Pull data from anywhere
Every analysis starts with getting data in and today that means CSV exports, retyped PDF numbers, and copy-pasted tables that break the moment the source changes. Extract removes that step entirely. You connect a source once, and Arkx pulls the data into a clean, structured table in your database. The table versioned, traceable, and ready for the next step.
What it does
Our Extract feature handles the two realities of analyst data. A structured sources you can connect to directly, and the messy files that don't have an API.
For structured sources, native connectors pull straight from your accounting system, bank feeds, and tools. e-conomic, Dinero, Billy, Shopify, Stripe, Google Analytics, SharePoint and more. Authenticate once and the connection stays live.
For everything else, AI extraction reads structured data out of PDFs, scans, emails, and the kind of irregular Excel exports that break normal importers. You point Arkx at the file and it returns a clean table.
How a table gets saved
Every extraction lands the same way, so the pipeline behaves predictably no matter where the data came from:
- Connect — pick a source or drop in a file.
- Map — Arkx detects columns and types; you confirm or adjust.
- Land — the result is written to a versioned table in your database (e.g. raw.economic), not a file on someone's desktop.
- Track — every pull is logged: what was pulled, when, and from where. Re-run it later and the old version is still there to compare against.
That last point is the main difference. A CSV export or a copy pasted table is a dead snapshot the moment it leaves the source. An Extract table is a living, audited record. You can always prove when and where a number came from, because the lineage starts here, at ingestion.