DiligenceVault

Models have worked in the product since 2019, with a generative module added in 2024, across questionnaire autofill, document processing and memo drafting. Without them, the two sided manager network, workflow and scorecards built over a decade would still stand.

Every AI assisted output is reviewed, edited and approved by the client's team, with source citations, version history and a library of approved prompts, and uncertainty is flagged rather than guessed. Nothing says at what threshold uncertainty gets flagged.

An audit trail covering every action, not only outcomes, and a named approver on each machine drafted output make the process reconstructable. Accuracy figures for autofill or extraction, error analysis and model documentation are unpublished.

Aggregate outcomes lead: 82 percent higher efficiency, 113 hours saved per diligence cycle and up to 60 percent shorter cycles, across 100,000 users and more than 20,000 managers. GEM Investments is named in a case study, while the headline testimonial comes from an unnamed $500 billion asset manager.

Isolation and encryption cover client data, which is never sold, shared or used to train shared models. That wording leaves open whether it trains anything else, and the lifecycle terms of the enterprise model provider are not stated.

Clients are the controller and DiligenceVault the processor, with data visible only to the client firm's own users under internal permissions and deletion available on request. Manager answers expose operations and terms to allocators who may be weighing competitors, and no retention schedule is published.

ISO 27001 and SOC 2 Type II are named, and TX RAMP and Data Privacy Framework certification sit beside them. No audit period, auditor or certificate is published, and operational due diligence teams, its own users, will ask for them.

A fund adviser's fiduciary duty over sub advisers is the stated reason the oversight module exists, and ADV filing monitoring is built in. Its Pulse Growth plan adds an AI compliance analyst that checks materials against the SEC marketing rule, though no formal admission stands behind the position.

Standardized questionnaires reward managers with the operational staff to answer them well, which tracks size more than investment quality. Emerging managers and small boutiques carry the same burden with fewer people, and nothing published addresses that effect.

The allocator stays accountable for a manager selection under its own fiduciary duty, and DiligenceVault publishes no accuracy guarantee or remediation term. A manager whose questionnaire answers are summarized wrongly in an allocator's memo never sees how its submission was rendered.

Input provenance is precise: managers respond and share profiles directly rather than being scraped. Behind the AI features sits an enterprise language model with no provider named, and no subprocessors are published.

Managers already on the platform for one allocator can answer another without onboarding again, which makes the network itself the integration. Power BI comes with every allocator plan and interface access starts at Spark Pro, though no portfolio or customer relationship system is named and no public API documentation is linked.

Microsoft Azure hosts a single global service across paired regions and availability zones, with no on premises option. Users in 150 countries send manager and allocator data across borders routinely, and no region choice or residency commitment is offered.

Prices are on the page: DV Spark runs $1,000 or $3,000 a month by managed entities and DV Pulse $300 or $750 a month by projects, each with unlimited users and implementation included. AI credits beyond the included allowance carry no published rate.

Both sides of diligence get their own product, DV Spark for allocators from private banks to family offices and DV Pulse for asset managers and fund service providers. Coverage runs from the United States to Australia and South Africa, and from mutual funds to hedge funds and private markets.