Claira

Granular structured data cannot be extracted from bespoke legal agreements at scale any other way. Without the models, analysts read structured credit documents by hand, which a bank executive quantifies at more than twenty minutes each. The company builds domain native models rather than applying general ones.

The same executive states that its specialized models and pretraining substantially surpass legacy language processing approaches, a comparison that comes from a customer rather than the vendor. Agentic workflows handle ingestion, classification, routing and analysis.

The division is clear from the product's own description: the platform assembles what a credit analyst needs to analyse and decide on a deal, delivering structured insight and workflow ready outputs, with the investment judgement left where it belongs. Automation covers ingestion, classification, routing and extraction rather than conclusions. Agentic bi-directional email workflows extend that without moving the decision. What is not described is any confidence indication on extracted terms, or what an analyst is expected to verify before relying on a structured value drawn from a bespoke agreement.

Traceability is built into the highest value workflow rather than promised generally, with executed documents received directly at loan closing producing a data feed the company describes as auditable and traceable, which means a structured value can be tied back to the executed agreement it came from. That is the correct control for extraction, where the failure mode is a term silently misread.

A named bank executive independently characterises the underlying models as materially better than legacy language processing. What is missing is measurement: no extraction accuracy, error rate or validation result is published for any asset class.

Two global banks jointly led the seed round alongside a technology fund, so the investors include customers, and one of them had already made a strategic investment through its spread products division three years earlier. A senior trading executive at that bank is quoted by name with a quantified baseline and result.

A major European private equity firm is named as a customer, and a partnership carries the capability to the 76 buy side firms and ten dealers on an institutional loan trading platform. Further case material describes analyzing over five years of a lender's historical investment memos and credit documents. The team is 20 people across two offices.

The missing data boundary statement is the profile's most consequential gap, and the product's stated purpose makes it acute. The platform systematically captures institutional investment knowledge and applies it to future transactions. It draws on a firm's proprietary research, historical memos and credit judgments, precisely the asset a private credit fund considers its edge. The same platform serves competing firms and a shared trading venue. Nothing states whether captured knowledge remains tenant isolated, who owns the resulting structured data, or what happens to five years of analyzed memos if a customer leaves.

There is no public data protection agreement, retention schedule, subprocessor list or deletion commitment. The material is commercially rather than personally sensitive: credit agreements, offering documents, investment memos and internal research. That lowers the consumer privacy exposure and raises a different one, since these are among the most confidential documents a fund holds. Ingestion runs through a dedicated secure email address, described without any statement about handling, retention or access.

Security is addressed only through the description of a secure dedicated ingestion address. No attestation, certification, trust center or enumerated framework is public. Two global banks have invested and one is a collaboration partner, so the underlying controls have been examined at the most demanding standard available. None of that assessment is published for other institutions to rely on.

No regulator, statute or rule is named. The asset classes involved carry substantial regulatory context, since municipal issuance operates under its own disclosure regime and structured credit under securitisation rules, and the platform's outputs feed pricing, valuation and risk decisions that supervised institutions must be able to defend. The auditable and traceable data feed is the property that would support such defence, and it is presented as a workflow benefit rather than tied to any obligation.

With no individual assessed, the bias exposure is the inverse of the product's main selling point. Systematically capturing past investment decisions and applying them to future transactions encodes a firm's accumulated judgment and reapplies it. That is valuable when the judgment was sound and self reinforcing when it was not.

Sectors previously avoided stay avoided, borrower profiles previously declined keep failing, and the blind spots in five years of historical memos become the defaults of the next five. Nothing describes how the system distinguishes a lesson from a habit, or whether a firm can examine which historical patterns are driving current recommendations.

No guarantee, indemnity or correction process is published. The institution is reasonably positioned, because traceability back to executed documents lets it reconstruct any extracted value and identify where an error entered. Nothing describes what the vendor owes when a misread covenant, rate or maturity feeds a pricing or valuation decision. Nor is there anything on how errors are notified once discovered, or how a correction propagates to analyses already built on the flawed extraction.

Models are described as domain native, specialised and pre-trained, which distinguishes them from a general purpose interface without identifying any provider, base model or version. No hosting arrangement or subprocessor list appears.

For a platform whose customers include banks that must document model provenance and change control, and whose competitive claim rests specifically on the models being purpose built, naming what they are built from is the disclosure a model risk function would ask for first.

The ingestion design removes the usual adoption barrier entirely by using email as the interface, with a dedicated secure address receiving documents that are automatically ingested, classified and routed with no uploads, portals or manual tagging, which matters because deal documents already arrive by email and analysts will not change that habit. The company states it fits into existing systems and feeds a customer's own underwriting and valuation tools. The named integration into an institutional loan trading platform reaching 76 buy side firms and ten dealers is the substantive one. No portfolio, order management or document system is named.

Hosting provider, region selection, residency commitment and private deployment options are all unpublished. Customers include banks operating across the United States and Europe, and the platform handles confidential deal documents before execution. Processing location is a routine question for those institutions, and published material does not answer it.

Claira publishes no pricing, packaging or basis of charge. The company states that customers scale as they grow with no fees. That describes a pricing philosophy rather than a price, and it gives no unit, whether per document, per deal, per seat or per firm. The value side is quantified precisely through the customer baseline, so a buyer can estimate the benefit without any sense of the cost.

Buyers include private credit funds, banks and investment firms, along with capital markets and trading desks, mid market managers and law firms. Together they cover most of the institutional side of private and structured credit. Asset class coverage is wide, and each class carries its own document conventions. It takes in collateralized loan obligations, municipal bonds and leveraged loans, plus commercial real estate and private credit generally.

Functional reach runs across underwriting, diligence and trading, and on through valuation, risk management and portfolio monitoring. Coverage is confined to institutional investing rather than extending to retail or corporate banking.