Job Groups
Process one mixed batch end-to-end: split, route, OCR, extract, and export per document type.
Scan a stack that holds several different document types — invoices from three suppliers, a couple of receipts, a delivery note — and a Job Group sorts it out in a single pass. Instead of running each document type as its own job, you point the whole batch at one group and Scanix Desktop routes every page to the right template automatically. This page explains the idea and how the pieces fit together; the how-to pages walk you through building and running one.
What a Job Group is
A Job Group is a container that ties several capture templates together and decides, page by page, which template each document belongs to. It exists to solve the "many documents, many types, one scanner pass" problem: a single batch carries documents that match different templates, and the group sorts each one to its template without you separating the stack by hand first.
Think of it as a switchboard sitting in front of your templates. The templates you add to a group are its members, and their order is a priority list. When a batch comes in, the group offers each page to its members in turn and lets the first one that recognises the page claim it.
A plain single-template job is just the simplest case of this — a group with one member. You don't have to think about groups at all until your batches start mixing document types; once they do, a Job Group is what keeps the run hands-off.
Screenshot
The Job Groups list page — a grid of group cards, each showing the group name, description, member count, and AI service. — shot job-groups-overview-01
How a batch flows through a group
Everything happens in one pass, in a fixed order. Understanding the order is the whole mental model.
Split
The batch is broken into individual documents. Each member template carries its own splitting strategy — for example, a barcode separator or a patch-code sheet — and the group uses those rules to find where one document ends and the next begins.
Route — first match wins
Each document is offered to the members in priority order, and the first member that recognises it claims it. Order matters: members near the top get first refusal, so a broad catch-all placed high will swallow documents a more specific template below it should have handled. Anything no member recognises is set aside rather than forced into the wrong template.
How routing works covers priority, first-match-wins, and the set-aside documents in detail.
OCR
Each document is read with the OCR settings of the template that claimed it — its zones, its language, its options. Different document types in the same batch can therefore be read in completely different ways.
Extract
Each template pulls its index fields from the document it owns — the metadata you export. When the group has an AI service configured, documents are also run through AI extraction so fields can be filled from the AI's reading of the page, not only from fixed OCR zones.
Export per document type
Finally, each document is exported according to its owning template's outputs. Because every document kept the template that claimed it, the invoices land where invoices go and the receipts land where receipts go — all from the one batch you started with.
Screenshot
A processed Job Group run showing documents sorted under their templates, with a separate Unrecognized review pile for documents no member recognised. — shot job-groups-overview-02
AI-only groups
A group doesn't have to lean on member templates to recognise document types. If you give a group an AI service instead, the AI reads each document and drives the extraction directly — useful when your incoming documents vary too much for fixed templates, or when you simply want the AI to do the classifying and field-filling.
When an AI service is set, every routed document is sent through it to produce a structured extraction, and your index fields can be mapped to the values the AI returns. Browse the full set in the AI Services catalogue.
Bring your own AI
AI extraction runs on the engine you've configured. Cloud calls go straight from your machine to your own provider account (Anthropic or OpenAI) — Scanix is never the account holder and never sees your key or your documents. For groups that must never leave the machine, a local-only privacy setting hard-blocks any cloud call regardless of how routing is configured.
Next steps
Build a job group
Create a group, add member templates, set their order, and pick an AI service.
How routing works
Priority order, first-match-wins, and what happens to documents no member claims.
Run a job group
Point a mixed batch at a group and process it end to end.