Build a job group
Use the six-tab designer to add members, AI services, field mappings, and export settings.
A Job Group lets one scanned batch carry documents from several different templates — say, invoices from three suppliers plus a couple of receipts — and routes each page to the right template automatically. This guide walks you through building one in the designer: adding member templates, picking AI services, wiring up where each field's value comes from, and setting where the output lands. You fill out six tabs, roughly in order, and save.
Bring your own AI
Job groups that use AI services run on the AI 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. You can also run entirely on-device. See the AI Services catalogue.
Create the group
Open Job Groups
In the left navigation, open Job Groups. Click New Group at the top right (or the New Job Group tile at the end of the grid).
Name it
In the Create Job Group dialog, give it a name — for example, AP Invoices (mixed suppliers) — and click Create. You'll land in the designer on the General tab. You can add member templates and splitting rules on the next screen, so a name is all you need here.
Screenshot
The Job Groups list page with the New Group button highlighted at the top right and the Create Job Group dialog open, showing the name field and the Create button. — shot job-groups-build-a-job-group-01
The six tabs
The designer has six tabs across the top. They're meant to be filled out roughly left to right:
| Tab | What you do here |
|---|---|
| General | Set the Group name and Description. The At a glance panel summarises members, AI services, and privacy. |
| Members | Add the templates that belong to this group and drag to set priority order. |
| AI Services | Pick one or more AI services and choose how Scanix routes providers and handles privacy. |
| Mapped Fields | For each member, choose where every index field gets its value from. |
| Jobs Matching | A read-only health check that flags routing problems before you run a batch. |
| Export | Set where the group's output is filed, the file format, naming, and collision handling. |
Only the General tab requires anything — a name. You can save partially and come back. The header Save button turns active when there are unsaved changes (you'll also see a · unsaved changes marker next to the group name).
Add member templates
Open the Members tab. Members are the templates a batch in this group might contain, and their order is their priority.
Add templates
Use the search box to find a template from your library, then click it to add. Repeat for every template that could appear in this batch type.
Set priority order
Drag the rows to reorder them. When a page comes in, the splitter walks the Members list top to bottom and the first member whose splitting rule matches the page claims it — first match wins. Put your most specific matchers at the top.
A page that no member claims goes to the unrecognized bucket, which you can review after the run. How routing works covers the matching order, catch-alls, and the unrecognized bucket in depth.
Screenshot
The Members tab showing a numbered, drag-to-reorder list of member templates on the left and the search-as-you-type add panel on the right. — shot job-groups-build-a-job-group-02
AI-classify unrecognized documents
The Members tab also has an AI-classify unrecognized documents toggle. When a document matches no member's splitting rule, Scanix can use AI to pick the best-matching member template (for example, Invoice vs. Receipt). It uses AI tokens per unrecognized document; leave it off to rely only on your configured matchers. See How routing works for when this kicks in.
Choose AI services
Open the AI Services tab if you want Scanix to extract typed fields — invoice totals, IBANs, contract dates, and so on. A group can run several services at once: each document segment the splitter produces is passed through every service you select.
Pick one or more services
Each service is a pre-built extraction schema. Select every service this group should run; click a card to see its full field schema in the right-hand panel. Choose No AI extraction to use template zones and OCR only, with no tokens consumed. Browse the full set in the AI Services catalogue.
Set provider routing
Under Provider routing, choose how Scanix picks a provider for each document:
| Policy | What it does |
|---|---|
| Local first, cloud fallback | Recommended. Runs the local model first and escalates to cloud only when cross-verification fails. |
| Local only | Never calls cloud. Fails the run if local can't satisfy the request. |
| Cloud only | Always calls the configured cloud provider. No local fallback. |
| Pinned to one provider | Forces a specific provider and model. No fallback — useful for reproducible audit runs. |
Depending on the policy, a provider sub-picker appears: Cloud fallback provider for Local first, or Cloud provider / Pinned provider for the cloud-bound policies. Your API key lives in Settings → AI Providers.
Screenshot
The AI Services tab: selectable service cards on the left with the Provider routing policy radios below, and the live field-schema preview on the right. — shot job-groups-build-a-job-group-03
Decide on privacy
The Privacy toggle switches between Privacy mode: allow cloud egress and Privacy mode: local only. When set to local only, no documents leave the machine even if local extraction fails — it overrides the routing policy. When cloud egress is allowed, you must tick I acknowledge cloud egress before the group can make cloud calls.
Cloud egress sends document text
With cloud egress allowed, document text — including any names, IBANs, or other PII it contains — is sent to your configured cloud AI provider under your own account. If documents must not leave the machine, set Privacy mode: local only.
Map fields to their source
Open the Mapped Fields tab. One AI call per document fans out across many index fields here, so you don't re-run extraction for every field you want to fill.
In group mode, the left sidebar lists every member template — pick one to edit its fields. Each index field has a single dropdown (Index fields → source) with three kinds of source:
| Source | When to use |
|---|---|
| Use OCR zone (unmapped) | The default. The field is read from the template's own OCR zone, exactly as if no mapping existed. |
| AI: <service> | Pull the value from an AI service's output. Options are grouped per service, so a group running several services gets one optgroup each. |
| From: <sibling template> | Pull the value from a different member's document in the same batch — for example, a cover sheet's order number populating every invoice that follows. Sibling sources appear in group mode only. |
The N / N mapped counter shows how many of the current template's fields are wired up. Mapped rows get an accent border so progress reads at a glance. Click Save mappings, then pick the next member from the sidebar.
Mappings are scoped to this group
Field mappings are scoped to (template, group) — the same template can map a field to AI in one group and to a sibling document in another. Wiring up this group won't affect any other group the template belongs to.
Screenshot
The Mapped Fields tab in group mode: the member sidebar on the left, the index-field rows with their source dropdowns in the centre, and the N / N mapped counter beside Save mappings. — shot job-groups-build-a-job-group-04
Check the Jobs Matching tab
Open Jobs Matching — a read-only health check that runs a static analysis over your group's member splitting rules and flags problems before you process a batch. The top row shows four stat tiles (members, clean, warnings, errors), and below them a group-level findings panel plus one analysis card per member. Fixes happen back in the Members tab or the Template Designer — this tab only reports.
How routing works explains each finding (shadowed members, unwired matchers, missing fallbacks) and how to resolve it.
Set up export
Open the Export tab to decide where this group's output is filed and in what shape.
When Use a group-level export config is on, every document the group produces is filed here, overriding each member template's own export settings. When it's off, each document exports per its matched member template's settings instead.
Choose formats and destination
Set the Document format (Searchable PDF, PDF (image), TIFF, Original file, or None) and the Metadata format (XML, CSV, JSON, Excel (XLSX), or None). Then set the Destination folder — type a path or click Browse.
Set the file naming pattern
In File naming pattern, build the output filename from tokens: ${filename} (the source file name), ${date}, ${timestamp}, ${counter}, or any extracted index field by name, e.g. ${InvoiceId}.
Turn on per-type subfolders (optional)
Enable Per-type subfolders to file each document type under its own subfolder (for example, …/Invoices, …/Receipts, …/Unrecognized). With it off, every document is filed directly under the destination.
Decide how to handle collisions
Under If a file already exists, choose what happens when an output name is already taken:
- Add a number — file (1).pdf
- Overwrite the existing file
- Stop with an error
Screenshot
The Export tab: Document format and Metadata format selectors, the Destination field with Browse, the File naming pattern input, the Per-type subfolders toggle, and the If a file already exists dropdown. — shot job-groups-build-a-job-group-05
AI-only groups need a destination
If a group has AI services but no member templates, it has no member export settings to fall back on, so a group-level export Destination is required — the Save button blocks and sends you to the Export tab until you set one.
Save
Click Save in the header. The group is now ready to process batches. To run one, point a source at the group — see Run a job group for the full end-to-end walkthrough.
Troubleshooting
Next steps
How routing works
Priority order, member matching, catch-alls, and the unrecognized bucket — explained.
AI Services catalogue
Browse every pre-built extraction service and the fields each one returns.
Run a job group
Point a mixed batch at your group, review the classified buckets, and export with Process.