Traydstream

Models do the reading, combining optical character recognition, natural language processing and machine learning trained on the specific language and rules of trade finance to digitise structured and unstructured content across more than 150 document types and extract thousands of attributes from dense technical paperwork. That is not achievable by rules alone.

What holds this at B is that the decisive asset is the rule library itself, more than 250,000 trade and compliance rules supporting over 400,000 permutations, which is domain codification built by trade experts rather than a learned model, and it is what a bank is actually buying. The models make that library applicable to paper.

The division is explicit and correctly scoped. The system reads, checks against the rule library, flags discrepancies and routes exceptions for human review, so automation resolves the clean cases while anything anomalous reaches an experienced checker, and the headline speed claim is carefully qualified as applying to compliant transactions specifically rather than to all traffic.

That qualification is the substance: a document that fails a rule does not get processed faster, it gets escalated. Centralised auditability accompanies it, so every check and every discrepancy is recorded, which matters in a function where a wrongly accepted document creates legal exposure for the bank.

The architecture makes errors attributable, which is unusual and valuable. Reading is done by models and deciding is done by a deterministic rule library, so a wrong outcome is traceable either to a misread attribute or to a rule applied incorrectly, rather than disappearing into a single opaque judgement.

Fully centralised auditability produces a digital record of the whole trade business for compliance purposes, and exception routing means low confidence cases surface rather than resolve silently. What is missing is measurement: no extraction accuracy, false discrepancy rate or missed discrepancy figure is published, and the last of those is the number a bank's operational risk function would want most.

Two African banking groups are named with executives quoted, and both quotes describe renewal and expansion rather than initial adoption, which is the strongest form of customer evidence because it reflects a decision taken with experience of the product. The client base is stated to include major global banks across Europe, Asia and the Middle East.

A partnership exists with one of the largest United States banks for automated trade document processing, and the technology is embedded inside a leading trade finance software provider's collaborative platform. Credibility hires are exceptional for this market: a former global head of trade and managing director at that same major bank, who also chaired the industry's principal trade banking association, joined as an adviser in January 2026. An independent analyst firm has reviewed the product. Funding is reported between 43.5 and 66 million dollars.

No data boundary statement was located. Models trained on the specific language of trade finance improve with exposure to more real documents, and the platform processes traffic for competing banks financing the same corridors and often the same underlying trades, since a single shipment can involve an issuing bank and a confirming bank both running checks. Trade documents also reveal a corporate's suppliers, pricing and volumes. Nothing states what is retained, whether document learning is shared, or how one bank's trade book is separated from another's.

Structurally favourable because of what trade documents are. Letters of credit, bills of lading and certificates of origin describe transactions between companies, so the payload is commercial rather than personal and no consumer is the subject of any assessment. What is sensitive is competitive: shipment terms, pricing, counterparties and trade corridors are among the most closely held information a corporate has. Held at B because no data processing terms, retention schedule or subprocessor list was located, and the platform operates across multiple jurisdictions with differing rules on commercial data.

No attestation, certification, trust centre or enumerated framework was located. Major global banks across three continents have completed vendor assessment, a large United States bank entered a processing partnership, and an established software provider embedded the technology in its own platform, so assurance has been demonstrated repeatedly in private. Publishing the control set would be the natural complement to the auditability the product already emphasises.

The rule library is regulatory codification at scale rather than a claim about compliance, covering global and regional trade finance rules, sanctions lists and compliance requirements across more than 250,000 rules and hundreds of thousands of permutations, with sanctions screening and anti money laundering detection running alongside document checking.

The company also publishes substantively on trade based money laundering as one of the most significant methods for laundering illicit funds globally, which is the specific financial crime typology this function exists to catch. What holds it below the top grade is that no individual instrument is cited, so a reader cannot confirm which rule sets the library encodes.

No individual is assessed and the adapted exposure falls on exporters. A discrepancy flagged against a letter of credit delays or blocks payment, and the party waiting is frequently a smaller supplier in a developing market for whom that delay is materially more damaging than for the bank or the buyer.

Rule coverage is unlikely to be uniform, since the library will be deepest for the corridors and document conventions that generate most volume, which means exporters in less represented markets face more manual handling or more questionable discrepancies. Trade finance already has a well documented access gap concentrated among smaller firms in emerging economies, and no coverage or error breakdown by corridor, document type or exporter size is published.

No commercial guarantee or indemnity was located, and the auditability commitment is substantive and aimed at the right risk. Trade document checking creates direct legal exposure, since a bank that accepts a discrepant document may lose its right to reimbursement, so a fully digital auditable record of every check and every discrepancy is what the institution needs when a dispute arises months later. That protects the bank.

The exporter whose document is flagged has no described route to contest a discrepancy raised by the system, which is inherent to the underlying process rather than a vendor failing, and the vendor adds nothing to it.

The technical approach is described by component, covering optical character recognition, natural language processing and machine learning trained on trade finance language, and the rule library is the company's own built by trade practitioners, which shortens the chain at the decisive point.

What is not disclosed is any provider: no model supplier is named for the reading components, no sanctions or watchlist data source is identified despite screening being part of the offering, and no subprocessor list or hosting arrangement was located.

The technology reaches banks two ways and both are named. It integrates directly with banks' existing trade finance systems, and it is embedded inside a leading trade finance software provider's collaborative platform, where that provider's customers consume document checking, sanctions screening and anti money laundering detection as part of their existing workflow with integration through to their back office trade solutions.

A separate partnership with one of the largest United States banks covers automated trade document processing. For a product whose adoption depends on fitting into decades old bank trade infrastructure, being resold inside the incumbent platform is the decisive distribution fact.

No hosting provider, region selection, residency commitment or private deployment option was located. The question is live given deployments across Europe, Asia, the Middle East and Africa, several of which impose local requirements on where banking data may be processed, and given that trade documents carry commercially sensitive counterparty information for corporates in each of those markets.

No pricing, packaging or basis of charge was located. Implementation is described as adapting to trade workflows without complex implementation, which addresses effort rather than cost, and the availability of the technology through a third party platform introduces a second commercial route that is equally undescribed. Nothing indicates whether charge falls per document checked, per transaction, per user or as a licence.

Four buyer types are served, spanning banks, other financial institutions, corporates and exporters, and geographic reach is genuinely global with named or stated deployments across Europe, Asia, the Middle East and Africa. Document coverage is the substantive dimension at more than 150 common global trade document types, spanning letters of credit, open account and collections, which is the range that determines whether a bank can route all its traffic through one system. The limit is functional: this is trade documentation and its compliance checking, and the newer positioning around execution, financing and risk management is not yet evidenced in the same way.