Channel Configuration — Header Fields
On the Header Fields page you define which header data Workist extracts from an incoming document and how each field behaves in further processing:
- whether it is a required field (so that missing values route the document into assistance),
- from which confidence the field is considered reliable enough to be passed through automatically,
- whether the field may block automation when its confidence is below the threshold,
- in which order the fields appear in the output file.
The header fields define the blueprint for the header row of your output file. The list on the left determines the order in the output file.
This guide describes editing the existing standard header fields (order type, customer, order number, …). For the full workflow for custom fields — creating them via "Custom Field", data type, restricting options, data label name, and AI instructions — see the guide on agentic extraction under Custom fields as the target of agentic extraction. Custom header fields are usually populated via LLM extraction; the creation process therefore belongs there thematically.
Prerequisites
- A Workbench account with access to the channel.
- You know which standard header fields your target system expects in the header row of the output file.
- Before making structural changes, a brief coordination with your contact person at Workist is recommended.
Step 1 — Layout of the page
In the Workbench, switch to your channel and open the Header Fields entry under Document Processing.
The page is two-column:
-
Left — "Configured Fields". A list of all header fields that are enabled for this channel. Each entry shows the field name and, below it, the status:
Status Meaning Required Required field. If the value is missing, the document goes into assistance. Optional Not a required field; a missing value does not block automation. Number (e.g. "3", "7") The field has sub-fields (e.g. delivery address → company, name, street and no.). The number indicates how many sub-fields are configured. The currently selected field is highlighted. Use the "+" symbol at the top right of this column to add another field, and the "×" symbol in a row to remove the respective field.
-
Right — "Configuration options for
<field name>". The detailed settings of the field selected on the left. Which options appear depends on the field type — not every field supports every option.
At the bottom right of the page you will find the central Save button. It saves all changes to all fields in one step; do not switch between fields without saving first, otherwise edits are lost.
The order of the field list on the left is the order in which the fields appear in the output file and in assistance. It can be changed via drag and drop (hover over an entry with the mouse, click the grab handle on the left, and move it).
Step 2 — Edit an existing field
In the left column, click the field you want to change — in our example "Order type". The right-hand side automatically switches to "Configuration options for Order type".
2.1 — Required
At the very top of the options you will find the "Required" toggle.
- Enabled — the field must be filled. If extraction finds no value, the document goes into assistance so that an agent can set the value manually.
- Disabled — the field is optional. An empty value does not block automation.
As soon as you flip the toggle, the status of the entry in the left field list also changes between Required and Optional.
2.2 — Further options
For the prepared standard fields, all further properties are already sensibly preset — usually you do not need to change anything here. The only thing typically adjustable is the field label name (the name shown in the Workbench assistance; changeable at any time without affecting the output file).
Further properties such as data type, data label name, a restricted value list, or AI instructions are mainly relevant for custom fields. You can find the full workflow for this under Custom fields as the target of agentic extraction.
2.3 — Confidence threshold and automation behavior
Some fields additionally allow you to control the automation behavior per field.
| Option | Meaning |
|---|---|
| Confidence threshold | A per-field threshold (0–100 %) from which this specific field is considered reliable enough to be passed through automatically. It overrides the global threshold from Document Processing → Document → Automation for this field only. |
| Field should not block automation when below the threshold | When active, the document does not go into assistance, even if this field's confidence is below the threshold. Useful for fields whose value is of interest to you but not critical. |
Not every header field supports its own confidence threshold. For standard fields such as delivery address, you typically find only the "Field should not block automation …" toggle — in this case the confidence threshold is taken from the global settings. A finer per-field confidence is available in particular for custom fields as well as for some special fields.
If you set, for example, 80 % here, then 80 % applies to this field, even if the global threshold in Document → Automation is 95 %. Set a per-field threshold only if you are aware of the difference.
Step 3 — Add a standard field
If you want to add another header field to the channel, click the "+" symbol at the top right of the left column.
The dropdown lists two kinds of entries:
- "Custom Field" (the topmost entry, separated by a divider) — creates a custom field that you can name freely. You can find the full workflow for this under Custom fields as the target of agentic extraction.
- Prepared standard fields (e.g. "(expected) goods issue date", "carrier provider", "order type", "carrier account number", "document type", …). These are already supported by Workist's standard logic, and you only need to configure their options.
Select a standard field. It appears immediately in the field list on the left — selected — and the right-hand side shows its (preset) configuration options.
For a standard field, the important options are preset — in most cases you can click Save directly. If you want to mark the field as optional, switch off "Required" beforehand.
Until you click Save, the change is not applied. If you leave the page before saving, the Workbench discards the new entry.
Step 4 — Remove a field
If you no longer want to extract a field, hover over the corresponding entry in the left list — a "×" symbol then appears at the right edge.
Click the "×" to remove the field. For fields with sub-fields (e.g. delivery address), all sub-fields are removed as well.
If a field is removed, Workist no longer extracts it from the next processing onward and no longer delivers it in the output file. If your target system still expects the field, the downstream imports will fail. Therefore remove fields only in coordination with those responsible for your ERP / target system.
Removal also only takes effect upon Save.
Step 5 — Save
Click Save at the bottom right to apply all changes.
- All changes to all fields (adding, removing, editing, ordering) are applied in a single save operation.
- If Workist finds an error during saving (e.g. a duplicate data label name, an empty required field), the Workbench marks the affected field in red — the save operation is aborted in this case.
- After a successful save, the new configuration takes effect from the next incoming document. Already processed documents are not re-evaluated retroactively.
Which setting when?
| Your situation | Recommended setting |
|---|---|
| Your target system requires the field; otherwise the import breaks. | Required = on. Leave "Field should not block automation …" off. |
| The field is nice to have, but an empty value is okay. | Required = off, "… not block …" on, and optionally set a custom confidence threshold somewhat lower than the global one. |
| Only a fixed value range is allowed (e.g. only 4 permitted order types). | Enter all permitted values under "Restrict options" — see Agentic header extraction → Custom fields. |
| You want to see a custom field in the output that is not included in the standard. | „+" → Custom Field — workflow in the LLM header guide. |
| This field's confidence should be allowed to be lower than the global threshold. | Set the field's confidence threshold specifically lower, and enable "… not block …" depending on the risk or not. |
Frequently asked questions
Why can I no longer edit the data label name? After the first save, the name is fixed because it is used as a key in the output file. If it really needs to change, create a new field with the desired name and remove the old one — and coordinate this in advance with the team that manages the target system.
What happens if I leave "Restrict options" empty? Workist then accepts any values of the configured data type. The list is therefore purely optional.
Why do some fields not show a "Confidence threshold" option? Not every standard field supports a per-field confidence — for these fields the global threshold from Document Processing → Document → Automation applies automatically. You can set per-field confidence values in particular for custom and some special fields.
How does the order of the field list on the left take effect? The order determines both the order in the output file and the display order in assistance. It is best to sort the fields so that they match the order in your target system — this simplifies later mappings.
What is the difference between "Required" and a high confidence threshold? "Required" acts on the value (is the value missing? → assistance). The confidence threshold acts on the certainty of the extraction (the value is there, but Workist is not sure? → assistance). Both combine: a required field with low confidence goes into assistance in any case.
Where do I see which header fields the standard logic already supports? In the dropdown of the "+" symbol. Everything above the divider is Custom Field; everything below it is a prepared standard field.
Help and support
If a field is not extracted or output as expected after a change, contact your contact person at Workist and have the following information ready:
- the channel name and the tab (Header Fields),
- the field name in the Workbench as well as the data label name in the output file,
- a specific example document that shows the problem,
- for required-field topics: whether the field is set to Required and whether the value is clearly recognizable in the source,
- for agent-based topics: whether the field is additionally referenced in Agentic header extraction.