The Pulse
D2 Intelligence Launches DocuMind for Auditable Document Decisions
Wenge AI subsidiary D2 Intelligence has launched DocuMind, a document intelligence platform for financial services, insurance and public-sector workflows. The company says the system links extracted fields and AI conclusions to source evide
AI.info Team ·
D2 Intelligence, a subsidiary of Wenge AI, launched DocuMind on September 23 as a document intelligence and decision platform for financial services, insurance and public-sector organizations. The system combines document recognition, information extraction, cross-document checks and risk review in one workflow, with each conclusion tied to supporting evidence.
The launch was announced in a PR Newswire release issued by Wenge AI. The company positions DocuMind for situations where decisions must be reviewed later by compliance teams, auditors or regulators, including trade finance, credit documentation, KYC and KYB checks, insurance claims and government applications.
From scanned pages to linked evidence
DocuMind is built on Wenge AI’s DOMA architecture, which stands for Data, Ontology, Models and Agents. The platform uses optical character recognition and vision-language models to turn multilingual and multimodal documents into structured data, then applies domain ontologies, knowledge graphs and business rules to interpret relationships, constraints and transaction logic across documents.
The design goes beyond extracting text from a single file. The release describes workflows that compare information across bills of lading, invoices, contracts, identity records and other submissions. Large language models and multi-agent systems assess document completeness, credibility, transaction logic and related risks using knowledge bases and external data.
DocuMind records the source location for extracted fields and preserves the rationale, evidence and audit trail behind each conclusion. Organizations can configure fields, thresholds, rules and review paths. Low-risk cases can run automatically, medium-risk cases can be sent to reviewers and high-risk cases can be escalated.
HKMA sandbox test sets the performance claims
Wenge AI says DocuMind processed a trade-finance document of up to 10 pages in under 30 seconds in the Hong Kong Monetary Authority’s GenA.I. Sandbox. The company compares that result with a manual review time of 10 to 20 minutes.
The same release says extraction accuracy for core fields exceeded 95% and that the review automation rate reached approximately 85%. Those figures are company-reported results from the described sandbox use case, rather than a third-party benchmark published with the launch.
For trade-finance reviews, the platform extracts parties, amounts, ports, routes and goods descriptions from source documents. It can then compare those details with sanctions lists, dual-use databases and institutional rules to flag inconsistencies and possible compliance risks.
Insurance and public-sector workflows
DocuMind also targets insurance underwriting and claims. The company says it can classify and link identity documents, medical records, reports, invoices and prescriptions, while detecting blurred, duplicated, recaptured or tampered images. It can reconstruct timelines when a claim arrives as a fragmented set of records.
Public-sector uses include application reviews, approvals, onboarding, due diligence and compliance evidence management. The platform is intended to consolidate scattered records into review files that show how an approval, risk assessment or other outcome was reached.
Controls for regulated deployments
Privacy and deployment controls form part of the product pitch. DocuMind supports data masking, isolated environments and on-premises deployment, features aimed at institutions that cannot place sensitive documents in a shared cloud service.
A Wenge AI spokesperson said in the release that enterprise AI needs to operate inside business processes with rules, traceable evidence and human oversight rather than focus only on content generation. The statement frames DocuMind as an application of Wenge AI’s broader data, model and agent stack to document-heavy decisions.
D2 Intelligence is headquartered in Hong Kong and has research and development teams in major cities across mainland China, according to the release. The company says it serves organizations including Haier, BYD, Kantar, China Construction Bank, China Merchants Bank, Guotai Junan Securities and CMG Asia-Pacific.
DocuMind’s stated proposition is narrower than a general-purpose AI assistant: read heterogeneous records, connect facts across them, apply configured rules and preserve the evidence trail behind the result. Its initial proof point is the company’s reported HKMA sandbox test—up to 10 pages processed in under 30 seconds, with core-field extraction above 95%.