BlueFlame AI

The removal test leaves nothing, since every capability is generative: summarising confidential information memoranda, analysing board decks, extracting provisions from partnership agreements and confidentiality agreements, completing diligence questionnaires, drafting investment committee memos and producing investor updates.

Retrieval augmented generation and natural language search over unstructured deal material underpin it, and a knowledge base add-on converts board decks, memoranda and contracts into structured, searchable form.

Output is draft material a professional completes rather than decisions taken, covering investment committee memos, investor updates and diligence responses, with more than fifty pre-built workflows the firm configures and task templates users build themselves. Research integration is designed to be citable so an analyst can follow a claim to its source. Held at B because no review requirement, approval step or accuracy checkpoint is described for documents that inform investment committee decisions and reach limited partners.

Two structural choices support verification. Automations are model-agnostic, so a firm is not locked to one provider's behaviour and can change it, and integrated research is specifically described as producing citable insights, meaning a claim in a draft can be traced to its source rather than accepted.

Held at B because no accuracy, extraction error rate or evaluation method is published for a system reading partnership agreements and drafting committee material, and an independent comparison notes it is not built to reason inside spreadsheets or produce auditable financial output, which bounds where its analysis can be relied on.

Clients are described as global managers running hundreds of billions in assets, spanning private equity, hedge funds, wealth managers, private credit and funds of funds, and the company reached a 50 million dollar valuation on an oversubscribed round raised from industry leaders and management rather than institutional venture capital, which is itself a signal about who believed the product worked.

Acquisition in 2025 by a major deal intelligence and virtual data room company is independent validation of a different kind. Held at B because no client is named and no adoption count, retention figure or outcome measure is published.

No boundary statement was located, and the exposure is acute for this product specifically. The platform holds confidential information memoranda, partnership agreements and diligence material for managers who bid against each other for the same assets and raise from the same investors, and following acquisition it sits inside a group whose core business is hosting other parties' transaction documents. Role-based access controls who sees what within a firm; nothing states what separates one firm's deal material from another's, or whether any of it informs the models.

Access and location controls are published rather than described in principle, covering single sign-on with multi-factor authentication, role-based access and data residency controls, backed by an independent audit attestation. Personal data exposure is limited since the subjects are firms and transactions rather than consumers. Held at B because no data processing agreement, retention schedule or subprocessor register was located, which matters where several external model providers are in the chain.

This is the most complete published security position in the index. An independent service organisation audit attestation of the more demanding operating-effectiveness type is held rather than merely targeted, hosting is on a named major cloud provider, and the access control set is enumerated: single sign-on with multi-factor authentication, role-based access, and data residency controls. The founding team came from cybersecurity and governance backgrounds, which shows in a posture built for a buyer whose own investors conduct operational due diligence on its vendors.

No regulator, statute or rule is named. The company states its security, privacy and compliance features address investor and regulatory mandates, which describes an intent rather than a framework, and the documents the platform drafts, particularly investor updates and diligence responses, sit within marketing and disclosure obligations for regulated fund managers that are nowhere mapped.

No individual is assessed and the adapted exposure runs through investment reasoning rather than people. Investment committee memos drafted from shared models and common templates across many managers tend toward a common analytical shape, so the diversity of judgement that distinguishes one firm's underwriting from another's can erode without anyone deciding it should. Nothing addresses that, and no analysis of output consistency or model behaviour is published.

No guarantee, indemnity or correction process was located. The exposure is concentrated in generated documents that travel outward: an investor update or diligence response containing a model-introduced error reaches limited partners and counterparties as the firm's own statement, and nothing describes what the vendor stands behind or how an error traced back to extraction rather than authorship would be handled.

This is the first disclosure of its kind in the index and it answers the question every other vendor leaves open. The architecture is stated as model-agnostic and three language model providers are named directly, so a buyer knows exactly whose models process its confidential deal material and that the dependency can be changed rather than being locked in.

A research provider is separately named as an integrated capability, hosting infrastructure is identified, and the acquired data sources including a private company intelligence database and two private markets data platforms are named as well. Supervisory guidance expects institutions to understand third-party model provenance; here they can.

Integration is named, numerous and aimed at the systems this buyer actually uses. Native connections cover the dominant private markets relationship management platform alongside enterprise email and a major customer platform, clients can connect to more than twenty system providers, and over fifty pre-built workflows ship with the product.

Post-acquisition it reaches the acquirer's virtual data rooms and a company intelligence database covering more than nineteen million private companies, plus two leading private markets data platforms. That is the full working environment of a deal team rather than a single connection.

The hosting provider is named rather than left implicit, and data residency controls are offered as a product feature, which together answer both where processing occurs and whether the customer can constrain it. Held at B because no specific regions are listed, no private or single-tenant deployment option is described, and the location of the several external model providers in the chain is not addressed.

No pricing, packaging or basis of charge was located. An independent comparison notes that following acquisition the pricing and packaging are likely to move toward the acquirer's enterprise customer base, which is a material uncertainty for a prospective mid-market buyer and is nowhere addressed by the company.

Buyer coverage spans private equity, private credit, venture capital, hedge funds, wealth managers, funds of funds and investment banks across New York and London, and the platform addresses several functions within each firm including deal teams, sourcing and origination, investor relations, portfolio operations, legal and compliance, and knowledge management.

Independent assessment tempers this usefully, observing it is primarily a deal team tool that does not cover legal operations, fund administration or compliance workflows at equivalent depth, and does not extend to portfolio management or valuation analysis.