From hours of document search to 20-minute reviews
How a leading Investment Management Administrator built the data foundation for AI, and what happened when they turned it on.
A business built on diligence, where speed and accuracy both matter
The firm administers bespoke Separate Managed Accounts (SMAs) as well as Private Funds (Funds) across a highly regulated marketplace. In just a few years, the firm has grown to over $4.5 billion in assets under administration, serving more than 85 RIA partners alongside major national and global institutional clients.
Their business model depends on one thing working well: bringing new subadvisors to market quickly and compliantly. Every new SMA program or Fund launch begins with structured subadvisor evaluation. That evaluation has to satisfy both the firm's internal standards and the specific compliance requirements of whichever client will onboard the subadvisor. A single onboarding can touch an RIA, a subadvisor, an institutional client, a custodian, and a broker-dealer, each with distinct documentation standards.
In that environment, due diligence is not a back-office function. It is the core operational throughput of the business. How fast the firm can evaluate and clear a subadvisor directly determines how fast the firm can bring an RIA's program to market, and is a key contributor as to whether the firm wins that business in the first place.
AI works when the data is ready. The firm made sure it was.
When the firm launched, they made a deliberate infrastructure decision. Before worrying about scale or automation, they needed their diligence process to be centralized and structured from the start. They adopted DiligenceVault as their operating system for subadvisor due diligence. Not as a document repository, but as the system of record for every evaluation, from day one.
That decision turned out to matter more than anyone anticipated. The firm centralized subadvisor data, structured how documents were collected, and ran every evaluation through a consistent framework. When AI capabilities became available, the data was already organized. There was no sprawl to clean up. No documents scattered across inboxes or shared drives. The foundation had been laid, and when DV Assist launched, the client could layer AI directly on top of it.
Firms that struggle with AI adoption often discover the problem isn't the AI. It's that their underlying data wasn't structured enough for it to work. This administrator avoided that problem entirely by investing in centralization first.
Hundreds of pages, scattered across multiple documents
Every subadvisor evaluation required manually reviewing offering memorandums, prospectuses, Form ADVs, compliance manuals, and supplemental materials. Relevant data points (ownership structure, liquidity terms, conflicts of interest, regulatory disclosures, key personnel) were buried across different sections of different files, with no consistent structure between managers.
1.5 to 3 hours of manual effort per evaluation
Finding, extracting, and entering data into the DDQ framework was painstaking. The floor was never less than ninety minutes. Complex evaluations consumed a full half day, time that was not available to spare as the firm's client pipeline grew.
Skilled analysts spending time on clerical work
The administrator's diligence process requires genuine domain expertise. A compliance disclosure here carries different implications than in a traditional fund structure. Client-specific requirements vary. Yet the mechanics of finding and transcribing that information were fundamentally clerical: experienced professionals searching and copying rather than evaluating risk and exercising judgment.
They didn't flip a switch. They ran a proof of concept first.
The client's path to AI-assisted diligence was deliberate. When DiligenceVault introduced DV Assist Document Intelligence, the firm engaged early, but on their terms. Before any deployment decision, they ran a guided proof of concept: their real documents, their actual DDQ framework, evaluated against their own standards for accuracy and source fidelity.
Centralized on DiligenceVault from launch
Rather than building on fragmented systems and retrofitting later, the firm structured their diligence process on DiligenceVault from day one. Every subadvisor evaluation ran through a consistent framework. Documents were collected in one place. This wasn't an AI decision. It was an operational one. But it became the prerequisite that made everything that followed possible.
Evaluated AI options, including building their own
As AI document tools proliferated, the administrator assessed what was available, including whether to build internal tooling or use standalone document parsers. The consideration was serious. The conclusion was that disconnected AI tools would create a different problem: outputs with no traceability back to source material, living outside the platform their team used every day. An integrated solution, where the AI lived inside the same system as the data, was the only model that made operational sense in a highly regulated environment.
Ran a guided POC before committing to deployment
The firm did not adopt DV Assist based on a demo. They ran a structured proof of concept with DiligenceVault: their documents, their DDQ framework, evaluated against a specific question, does the AI extract the right answer, from the right document, and can we verify it? When the answer came back yes, the case for adoption was clear. Critically, the POC also built internal confidence. Analysts who had been skeptical could see exactly where each prefilled answer came from before they were asked to trust it. That traceability, not the speed, was what got the team on board.
Integrated into the standard evaluation workflow
With the POC validated, the client integrated DV Assist into their standard subadvisor evaluation process. Subadvisor documentation arrives through DiligenceVault. DV Assist reads the documents against the firm's specific review framework, extracts the relevant data points, then prefills the DDQ. Every answer is linked to the source passage it came from. Analysts step in to verify, refine, and apply judgment. The clerical portion of the work had been eliminated. The analytical portion remained fully in human hands.
Time savings were the headline. Growth capacity was the point.
The most visible result is the one that's easy to measure: DDQ population time dropped from 1.5 to 3 hours per evaluation to roughly 20 minutes. That's an 80 to 90% reduction in manual effort. But for a firm whose revenue model is built on onboarding subadvisors at scale, the business impact extends well beyond the time saved per file.
More programs, without adding headcount
The firm's business grows by onboarding more subadvisors onto institutional client platforms for new RIA programs. Diligence was the rate-limiting step. Compressing the evaluation cycle means the same team can clear more managers per quarter, directly enabling more programs and more client revenue without a proportional increase in operational cost.
Faster time-to-market for RIA partners
RIA partners choose administrators partly on execution speed. When an RIA wants to launch a custom SMA program, the time from engagement to institutional client approval is a competitive differentiator. The firm can now move faster. Their RIA partners get to market sooner, and the relationship is stronger for it. Furthermore, RIAs can spend less time and resources on the diligence process themselves, allowing them to focus on what they do best, invest on behalf of their clients.
Every answer traceable to its source
In a highly regulated environment, auditability is not optional. DV Assist links every prefilled DDQ response to the specific passage and document it came from. When an evaluation is revisited for a new program, the sourcing is already structured and on file. No reconstruction. No uncertainty about provenance.
Judgment applied where it actually matters
Analysts now begin each evaluation with a substantive, sourced draft rather than a blank framework. Time previously spent on document search and transcription shifts entirely to verification, exception review, and the risk assessment that experienced diligence professionals are hired to perform.
"The platform gave us a way to govern AI responsibly. Every prefilled answer has a source, and the information remains within our DiligenceVault database. That's what made it workable in a highly regulated environment, and what got our team comfortable with it. AI Assist has been a major reason as to how we have completely digitized our due diligence process."
Executive, Leading Investment Management AdministratorFrom document extraction to automated risk screening
Document Intelligence is step one. The administrator is now integrating the Review & Risk Agent, a new functionality which is part of DV Assist 2.0. It is designed to screen completed Subadvisor DDQ responses against the firm's proprietary compliance parameters before a human analyst opens the file.
Proprietary risk frameworks, applied automatically
The Risk Agent is designed to flag policy deviations, outlier responses, and threshold breaches calibrated to the client's exact standards and the regulatory requirements of the marketplace, before review begins.
End-to-end: ingestion to screened review
Document Intelligence populates the framework. The Subadvisor returns it. The Risk Agent screens responses. Analysts begin with a structured, pre-screened file, focused entirely on the exceptions and judgments the platform cannot make for them.
The firm didn't automate their diligence. They made it possible to do more of it.
The analysts still own every judgment call. They just reach those judgments faster, with less manual work behind them and a stronger audit trail supporting every conclusion.