Capsa AI

The product is described as an AI operating system for private equity and the capability set is entirely model work: extracting and consolidating financial information from company records, analysing performance, reviewing customer contracts, indexing historical deal and communication data, and surfacing traced answers to investment questions. Later development is explicitly agentic and indexing focused. Remove the models and the manual diligence process the company measures itself against is what remains.

The delegation framing is precise about what is handed over, targeting manual low-skill tasks so professionals can focus on higher-value strategic decisions, and outputs are traced answers and structured analysis rather than recommendations to act. The investment judgement is never described as automated. Held at B because no review requirement or checkpoint is published for extracted financial figures and contract terms that feed an investment committee decision.

Traceability is stated as a design property rather than an aspiration, with the platform surfacing and tracing answers to investment questions and maintaining full auditability across the investment process, which matters because an extracted figure entering a committee paper must be attributable to its source document.

Held at B because no accuracy, extraction error rate or validation approach is published for financial extraction and contract review, and the company describes building proprietary models without describing how they are evaluated.

The company was revenue-generating within eight months of founding and serves funds managing over 30 billion dollars across three countries, having co-developed sector-specific features with multiple leading firms. Outcomes are quantified and modest enough to be credible: early customers reported a 20 percent reduction in due diligence time alongside a 90 percent satisfaction score, and cited careful consideration of private equity specific workflows as the differentiator from general AI tools. Held at B because no fund is named and both figures come from the company rather than independent assessment.

No boundary statement was located, and the exposure is acute for this product specifically. The platform indexes historical deal and communication data for firms that bid against each other for the same assets, and its stated ambition is to help funds leverage that historical deal data for better decisions, which is precisely the material that would be most valuable if it crossed tenants. Nothing states how one firm's deal history is separated from another's.

No data protection agreement, retention schedule or subprocessor list was located. Data security is asserted at the highest levels without specification. The platform ingests company records, customer contracts and internal communications from target businesses that have not contracted with the vendor and may not know their material is being processed, which is a distinct exposure from customer data.

No attestation, certification, trust centre or enumerated control set was located, only assertions of the highest levels of data security and full auditability. That is a visible gap against a direct competitor in the same private markets niche that publishes three held certifications, and institutional investors conducting operational due diligence on their managers' vendors will ask for the documentation.

No regulator, statute or supervisory expectation is named. Early coverage described the product as automating compliance tasks, which on inspection means diligence process discipline rather than regulatory compliance, and private capital firms face valuation, marketing and investor reporting obligations that go unmapped.

No fairness testing or governance disclosure was located. The adapted exposure runs through capital allocation: a system that structures how diligence is conducted and which findings surface shapes which businesses clear investment committee, and a firm leveraging its own historical deal data to inform future decisions is by construction learning from what it has previously favoured. Nothing addresses whether that reinforcement is examined.

No guarantee, indemnity or correction process was located. The affected party is the target company whose records, contracts and performance are analysed without any relationship with the vendor, and which has no route to see how its financials were extracted or interpreted, or to correct a misread contract term that contributes to a valuation or a decision not to proceed.

No base model, provider or hosting arrangement is identified. The company states it is building proprietary models and is investing in agentic and indexing capability, which describes ambition rather than dependency, and one market data platform is named on the integration side. A fund cannot document whose models process its deal material from published information.

Integration targets where a deal team's material actually sits, connecting customer relationship systems, shared drives and a major market data platform into a unified source, with historical deal and communication data indexed from existing systems rather than requiring migration. One data platform is named individually. Held at B because no relationship management or portfolio system is named and no developer documentation was located.

No hosting provider, region, residency commitment or private deployment option was located. The company operates across the United Kingdom, United States and Germany, and German fund clients in particular sit under data expectations that make processing location a procurement question, which published material does not address.

No pricing, packaging or basis of charge was located. The company does publish the cost of the problem it replaces, at roughly 50,000 pounds annually per professional in repetitive diligence work, which gives a buyer a denominator without giving it a numerator.

Coverage is deliberately narrow and deep, aimed at private equity and wider private capital firms rather than spanning institution types, with the workflow reach extending across sourcing, due diligence and portfolio monitoring rather than a single stage. Presence spans the United Kingdom, United States and Germany with offices in London and New York. Held at B because one buyer segment is served and no fund count or size range is evidenced.