Identity and onboarding document capture
Tavnit reads the documents collected during onboarding — identity documents, proof of address, incorporation papers — and returns names, numbers, dates and addresses as structured fields. Because the whole set is processed together, whether the details agree across documents becomes a check rather than an assumption.
Why this is painful
Onboarding a client means collecting documents, reading them, and confirming they describe the same person or entity. The reading is tedious; the confirming is the part that carries risk, and it is done by eye.
Volume makes it worse. A team that onboards steadily ends up with inconsistent standards between reviewers and no reliable record of what was checked when.
What to extract
| Field | Why it needs care |
|---|---|
| Full name as printed | Extract verbatim. Ordering and transliteration differ between documents and that difference is the check. |
| Date of birth | The single most useful cross-document consistency field. |
| Document number and type | Passport, national ID, licence — each with its own numbering conventions. |
| Issue and expiry dates | Expiry is a rule rather than a field: an expired document is a finding. |
| Address | From proof-of-address documents, which need to agree with what the client declared. |
| Entity registration details | For business onboarding — company number, registered office, directors. |
What makes identity & onboarding hard
Names do not match, and that is normal
The same person appears as 'Jose Garcia Lopez', 'J. García' and 'GARCIA LOPEZ, JOSE' across three documents. Extract names verbatim rather than normalising during extraction, then compare deliberately — normalising too early throws away exactly the difference you need to assess.
Photographed ID is the norm and quality varies
Clients photograph documents on phones, with glare across the machine-readable zone as often as not. Review on onboarding is proportionate given what a mistake costs, and the audit trail then records who accepted which document.
This is the most regulated data you will process
Identity documents carry retention limits, erasure rights and access obligations that ordinary business documents do not. Private Buckets and per-user grants keep onboarding files to the compliance team, and the append-only trail evidences who accessed what.
How the pipeline handles it
- Collections — Classifies whatever the client sent — passport, utility bill, incorporation certificate — and routes each to the right flow.
- Flows — One flow per document type, so a check can reference a named field rather than searching text.
- Human in the Loop — Sign-off on every onboarding, with a permanent record of who approved which document.
- User Roles — Restricts identity documents to the team that needs them rather than the whole organisation.
Common questions
- Does it verify that a document is genuine?
- No. Tavnit extracts and structures what the documents say, and lets you check consistency across a set. Authenticity verification against issuing authorities is a separate specialism, and this does not replace it.
- Can it check that details agree across documents?
- Yes. Process the set together and compare the extracted fields — name, date of birth, address — so a mismatch surfaces as a value rather than something a reviewer has to spot.
- How is identity data kept restricted?
- Private Buckets with per-user access grants limit visibility to the compliance team, and the append-only audit trail records every view, edit and approval.
Other document types
Try it on one document
Build a flow, send a real identity & onboarding through it, and see what comes back. Free credits to start.
Start free