Document Collection asks a candidate for a document, uses AI to confirm the upload matches what you asked for, and reads the fields you select. Anything it isn't sure about goes to a review queue so a person can decide. Each ask can accept several document types, and each type is judged against checks written from your own sample documents, so you can see why an upload passed or was held back.
How Document Collection Works
Everything is managed in Templates > Document Hub. These are the building blocks you'll work with:
Term | What it means |
Ask | The request a candidate answers, such as "Identity document." One ask can accept multiple document types. |
Document type | A specific document that can answer an ask, such as a passport or a national ID. Each type has its own samples, checks, and fields. |
Sample | One to five example documents you upload for a document type. The checks are written from them. |
Checks | The lines an upload must satisfy (What a document must show) and the lines that cause it to be refused (Refuse if). |
Fields | The data read off every upload of a document type, optionally filed onto a candidate attribute. |
When a candidate uploads a document, it's compared against the checks for each type the ask accepts. A clear match is accepted automatically, a clear mismatch is marked unrecognized, and anything uncertain is held as Needs review for a recruiter to verify or reject.
Note: Documents that arrive through an integration, such as a signed contract returned from e-signature, aren't checked.
Important: Document Collection confirms that the uploaded document matches the type you requested and is clear enough to read. It doesn't detect forgery, check government databases, or confirm the document belongs to the candidate who sent it. To match a candidate's face, see Face Verification on AI Calls. |
Turning On Document Collection
Three company-level settings control Document Collection. Contact your Account Manager to turn them on or change them.
Setting | What it does |
Document processing | Turns Document Collection on for your company. When it's off, it doesn't check uploads. |
Real-time check on Landing Pages | Checks the upload as soon as the candidate picks the file, so they can retake a bad photo before submitting. It can hold a document back early but never accepts one on its own. The final check still runs when the document is attached. |
Second opinion | Requires a second AI model to agree before it automatically accepts an upload. If the two disagree, the upload goes to review. It's off by default, so one model's answer stands. |
Creating a Candidate Document
A candidate document defines what candidates need to upload during their application. You can specify the document name, provide instructions, and choose who can access and review the submitted files.
For example, you can create a document called "Proof of Identity" and allow candidates to submit a passport or national ID.
Go to Templates > Document Hub.
Click Create document and enter the name of the document you want candidates to submit, such as "Proof of Identity."
Select the Country of issue, if applicable.
Under Instructions for the candidate, explain what candidates should upload and any requirements, such as submitting a clear photo with all details visible.
Choose who can view and download the submitted documents.
Optionally, enable Check uploads automatically to have uploaded documents checked automatically. You can also enable Real-time validation once automatic checking is turned on.
Click Create document to save.
Training a Document Type
Each document type learns what to look for from your samples.
Open the document type from its ask in Document Hub.
Upload one to five sample documents in the Samples section. The first sample is the one candidates see as an example.
Wait about a minute. The system automatically writes the checks for that type.
There's no Retrain button. Adding or removing a sample is the only thing that makes the checks out of date, so the checks rewrite when you change the samples.
Important: Don't use a real candidate's document as a sample. Samples are visible to everyone in your workspace, and anything printed on a sample becomes part of what later uploads are judged against. Use a specimen or test document instead. The screen warns you if a sample appears to contain real personal data. |
Marking Points on a Sample
In Points on the sample, click the picture to add a numbered marker on a feature, such as the photograph or the document number. Each marker with AI check turned on becomes a line in What a document must show.
Reviewing and Editing the Checks
Each document type has two lists of checks:
What a document must show – What every accepted upload has to include.
Refuse if – Anything that should cause an upload to be refused.
You can edit any line, switch it off without deleting it, or remove it. You can also add your own lines to either list.
A document type needs at least three active checks. With fewer, or with every check switched off, uploads are held for review instead of being judged.
Checks go live as soon as they're written. To undo a change, open Version control and restore an earlier version in one click.
Extracting Data From Documents
In Data extraction, name the fields you want read from every document upload, such as the document number or date of birth. You can optionally map each field to the candidate attribute it should be filed against. The system always shows what it read on the upload itself.
Note:
Accepted documents only – Values are filed onto the candidate only when the document is accepted. A document held for review writes nothing to the candidate profile.
Unclear values stay empty – If a value can't be read cleanly, such as an ambiguous date, it's recorded as unresolved, and the attribute stays empty rather than being filled with a guess.
Using Document Results in Autoflow
Document Collection keeps two candidate attributes up to date for each ask, named after the ask. For an ask called "Identity document," they appear as Identity document Verification Status and Identity document Document Type.
[Ask] Verification Status – The document's status in words. This is the attribute Autoflow reads, so use it in conditions to move or message candidates based on the result.
[Ask] Document Type – Which document type answered the ask. It's only set for asks that accept more than one type.
To build rules on these attributes, see Autoflow Sets.
Testing a Document Type
Use the test bench to see how a document would be judged before candidates upload it. Nothing you test is stored, and no candidate is affected.
Open the document type in Document Hub.
Go to Try a document against these checks.
Drop a file into the test area.
Review the verdict the document would receive.
Stopping or Deleting a Document Type
Stopping a document type removes it from what the ask accepts but keeps its history. Deleting removes it completely and is only possible for a type that has never judged a real upload and isn't the last type on the ask. If a type has already been used, stop it instead.
Q: What happens to asks I set up before Document Collection was turned on? A: An ask that hasn't been trained yet keeps working the way it did before. It moves to Document Collection automatically as soon as its first set of checks exists. There's no second switch to turn on.
Q: Can one ask accept more than one kind of document? A: Yes. Add each document type that should answer the ask. For example, a passport and a national ID can both answer "Identity document."
Q: Do I need to retrain a document type after editing its checks? A: No. There's no retrain step. The checks are only written again when you change the samples.
Q: Where do recruiters review documents that are held back? A: In Templates > Document Hub > To review. Recruiters see held documents from applications in campaigns they can access, with a separate count for the ones assigned to them. |
