Test management
Realistic document layouts for meaningful test cases.
Exercise OCR, extraction and validation with familiar structures; make unwanted edits visible.
01 / PRODUCT Documents for testing, AI & demos
The document stays.
The identity is replaced.
Turn invoices, medical letters, contracts, payslips and official notices into synthetic document twins. Personal fields are replaced; layout and task-relevant content are retained within the agreed scope.
Demo access on request. Designed for operation in the customer network. Document families and release criteria are assessed in the pilot.
SOURCESYNTHETIC TWIN
- PERSON
Erika Mustermann→ Lena Hartmann- BIRTH_DATE
13.02.1974→ 04.07.1972- TELEPHONE
+49 30 12345678→ +49 33 21253320
Crop of the existing product recording; no original customer data. Outlines explain the fields.
- On-premises in your network
- CPU-only · no GPU required
- No LLM · no external AI API
- Local OCR & NLP
In the agreed on-premises deployment, documents remain in the customer network. The provided demo uses a separate hosting environment.
02 / CONTEXT
Test data that fits your processes.
Real customer documents contain sensitive information. Redaction alone changes the test case. CoverTwin combines replacement values with controlled output verification.
Data & AI
Evaluation and training preparation with clear approvals.
New fictional labelled text for initial model tests. Source-derived training export requires separate acceptance and is currently blocked.
Privacy & information security
Technical release rules for controlled output.
Residual-data checks, roles and traceable verification results support your assessment of permitted use.
03 / APPLICATIONS
Industries & use cases
Potential applications; document families and permitted use require pilot acceptance.
- Insurance & health insurance
- Prepare invoices, claims documents and correspondence for document testing.
- Banking & financial services
- Assess contracts, account and application documents as potential pilot test families.
- Healthcare
- Review medical letters and billing documents from clinics, laboratories and billing centres in a controlled workflow.
- Public administration
- Scope applications, official notices and evidence documents for migration, UAT and training.
- Energy & telecommunications
- Assess invoices, contracts and customer correspondence in existing extraction workflows.
- HR & payroll
- Prepare payslips and personnel documents with agreed replacement and release rules.
- Law firms & legal
- Preserve contract and case-file structures for testing; assess context and permitted sharing separately.
- Software, IDP & input management vendors
- Select document families for QA, evaluation and product demos and assess them in a pilot.
- BPO & scanning services
- Test scanning and processing workflows with realistic copies; include external recipients in release decisions.
Cross-industry use cases
- Test data
- QA, migration and UAT with realistic document structures.
- Document AI
- Training and evaluation preparation; source-derived training export only after separate acceptance.
- Demos & sales
- Document examples with fictional personal fields for agreed demonstrations.
- External providers & offshore
- Controlled sharing only with an agreed purpose, recipient controls and the required legal basis.
- Training
- Practise document workflows with approved test copies.
04 / PRODUCT VIEW
Two views of the pilot.
An entirely fictional test document and newly generated fictional development text. The static comparison shows actual CoverTwin output; the recording shows the separate development workflow.
Static comparison of an entirely fictional test document. Personal fields, including telephone, were replaced. The highlighted fields and source/output values can be inspected in the interactive comparison above.
Inspect highlighted fieldsReview newly generated fictional text and labels. The recording shows the development workflow; availability and approval form part of acceptance.
Video text description
The interface generates new fictional text for PII development. Entity spans and roles are highlighted in the preview. An authorised person reviews the examples and development dataset. No uploaded source documents are used for this dataset. The recording does not establish currently approved source-derived training export.
05 / PROCESS
From source to a reviewable test copy.
Ingest
Associate a PDF, scan, image or text with a job.
Read
Extract text or read it with local OCR.
Identify
Recognise personal fields in document context.
Synthesise
Replace recognised identifiers with consistent fictional values.
Verify
Read output again and check residual identifiers, readability and domain content.
Deliver
Release for the intended use only after technical checks pass.
06 / SECURITY
Security starts at the deployment boundary.
In the on-premises model, intake, processing and output stay inside your infrastructure. Network rules, identity, key management and monitoring are agreed before the pilot.
- 01Your DMS / object store
- 02CoverTwin · local processing
- 03Verification gate & review
- 04Your test workflow
Encrypted jobs
The documented pilot uses AES-256-GCM for source, output and comparison. Key management and infrastructure encryption require deployment acceptance.
Retention & deletion
Documented demo state: original preview for 15 minutes, output job for up to one hour. Owners can delete jobs. Customer retention periods are agreed separately.
Roles & review
Jobs are assigned to a user and tenant. Original previews require a reviewer role. Enterprise SSO and the role model are scoped for the target system.
Traceable verification
Verification results and acceptance criteria support traceable decisions. Audit integration and immutable archiving are separate integration scope.
Technical verification does not replace a legal anonymity assessment. Context and health information may still be personal data.
07 / DEPLOYMENT
Operation under your control.
On-premises
Containers based on Docker Compose in your network. Integration with document storage and test workflows is scoped together.
Your private cloud
Customer-managed environment with agreed network and access rules. DMS adapters, SSO and queues form part of the agreed integration scope.
CPU and RAM requirements: on request, based on your document families and pilot measurements. Licensing, integration and support follow the confirmed scope.
08 / PILOT
A pilot. An informed decision.
Two document families and one downstream task provide a focused starting point. Duration and dates by agreement.
Initial discussion
Clarify objectives, data ownership and security requirements.
Select document families
Choose two families and a separately labelled acceptance set.
Test in your network
Measure leakage rate, automation rate, unwanted edits and review effort against agreed thresholds.
Assess the results
Review the findings, integration scope and next steps together.
09 / QUESTIONS
What to know before the pilot.
Are the results automatically anonymous?
No. Fictional replacement values alone do not establish legal anonymity. Context, rare facts and re-identification risk need assessment for the intended use.
Can I export source documents for model training?
Source-derived training export is currently blocked. It requires separate legal and technical acceptance, a labelling policy and release rules. Newly generated fictional development text is a separate workflow.
Do we need a GPU or an external AI service?
Local OCR/NLP processing is CPU-based and uses no LLM or external AI inference API. Capacity is measured with your documents.
What is integrated in the pilot?
Containers and HTTP endpoints provide the foundation. Document vocabulary, DMS integration, SSO, queues and output formats are agreed as a concrete scope.
How do we get demo access?
Request access through the contact section. The separate demo is protected by Cloudflare Access and intended for fictional test data.
10 / PROVIDER
Provider
11 / CONTACT
Let’s scope your pilot.
Tell us about your test workflow and document families. We will respond shortly and agree the next step with you.