Managing asset growth and rising diligence complexity with AI
How a leading Client RFP Team handled an 86% rise in question volume and increasing product complexity across SMAs, Models, and ETFs, with a single system for every request format, AI-assisted drafting, and audit-ready governance throughout.
More assets. More products. More questions. Same team.
The firm's US RFP Team works where investor relations and due diligence overlap. The team supports completion of RFPs, RFIs, and DDQs across a widening product set, and has spent years putting technology to work inside those workflows.
Over a five-year span, from YE20 to YE25, the operating environment changed sharply. The number of questions the team handled rose 86%, the number of requests rose 35%, and new business demand grew across separately managed accounts (SMAs), Models, and ETFs. Volume and complexity climbed together. New product lines meant new content, new disclosures, new regulatory requirements and new positioning to keep current. More sophisticated allocators asked harder questions. And through all of it, the team set out to meet that demand without growing headcount in proportion to it.
In a regulated environment, speed on its own is not the objective. Every response has to be versioned, approved, and traceable to an approved source. For this team, diligence is not support work; it is the throughput the business runs on, and how fast each request clears, without bending the controls, is what keeps clients on schedule. The challenge was straightforward to state and hard to solve: handle more work, hold the controls, and do both with the same team.
Good AI output starts with governed content. That groundwork was laid in 2018.
Since 2018, the team has run its diligence on DiligenceVault: one place for its approved answer library, the requests clients send in, the tracking and handoffs between contributors, and the records an audit later draws on. None of it was set up with AI in mind. It was set up to run a regulated operation cleanly.
That earlier decision is what made AI practical years later. The approved answers, the source materials, and the workflow were already sitting inside one governed system, so AI-assisted drafting had something trustworthy to work from on day one instead of trawling inboxes and shared drives. When AI projects stall, the cause is usually the data rather than the model: the content underneath was never organized well enough to use. The firm never hit that wall. The team had run its diligence on DiligenceVault since 2018, keeping its content centralized, approved, and governed all along, so by the time AI arrived the foundation was already in place.
One bottleneck stayed in place regardless. Even with everything centralized, writers and internal subject matter experts (SMEs) still spent real time drafting by hand, fielding the same questions over and over, and digging through documents to assemble each response.
Rising volume and complexity across RFPs, RFIs, and DDQs
The 86% growth in questions and 35% growth in requests did not arrive as more of the same work. Requests grew more complex in parallel, spanning a broader product range and more demanding diligence standards, so each additional request carried more drafting effort than the last.
Requests arriving in every format, with no single place to manage them
Not every allocator sends a digital request. A significant share of the team's intake arrived as Word documents and Excel files, each requiring manual reading, question extraction, and entry into the workflow before drafting could even begin. Running two separate tracks, one for DiligenceVault requests and one for offline documents, created duplication, version risk, and gaps in the audit trail. The team needed one system to cover both, not a workaround for each.
Recurring requests with small variations consuming disproportionate effort
Many allocators send the same DDQ or assessment template quarter after quarter, updating a handful of questions each cycle. Without a way to detect that a new document is a variant of one already answered, the team treated each submission as a fresh intake, reassembling and re-drafting content that was largely unchanged.
Significant new business demand for SMAs, Models, and ETFs
Growth in separately managed accounts, model portfolios, and ETFs meant new product facts, disclosures, and positioning to maintain and reuse, and more first drafts to produce against tighter deadlines.
No clear weekly view of what needed attention and when
With volume spread across concurrent requests, contributors needed a reliable way to know what was open, what was due, and what was waiting on them, without having to check in across multiple threads or track down status manually.
From manual drafting to an AI-augmented review model
The shift was deliberate and governed-first. The firm did not adopt AI to replace judgment; it adopted AI to remove the manual drafting that stood between its experts and the judgment work only they could do. DV Assist was introduced as an internal drafting aid, scoped tightly to approved content, with human ownership preserved at every step.
Centralized on DiligenceVault as the system of record
Instead of spreading diligence across disconnected tools, the team standardized the whole process on DiligenceVault: one approved-content library, structured intake, shared workflow tracking, and audit-ready records. It was an operations call made with no AI agenda, but it quietly set up everything that came afterward.
Volume climbed 86%; turnaround still improved 40%
As questions rose 86% and requests rose 35%, the team kept tightening its process and average document turnaround time fell 40% over the same period. Process maturity carried the team a long way, but manual first-draft creation remained the hard ceiling on how much further productivity could go without adding people.
Adopted DV Assist under the firm's AI governance standards
The case for AI-assisted drafting only held if it matched the firm's governance posture exactly, and it did. Use stays internal, with no AI ever facing a client. Access is permissioned by role inside DV. Drafts draw only on approved internal sources, and every line traces back to where it came from. People review at more than one checkpoint, and nothing executes, decides, or communicates on its own. That set of constraints, not the time saved, is what made the tool adoptable in a regulated setting.
An AI-augmented review model, end to end
The old sequence had writers and SMEs drafting by hand and digging through documents for each answer. The new one opens with a first draft DV Assist produces in seconds, pulled only from approved internal materials such as prospectuses, fact sheets, SAIs, and audited financials, each line tied back to its source. From there the team checks the facts, sharpens the wording, and makes the calls that need judgment. The mechanical work moved to the platform. The thinking stayed with people.
Internal use only; AI never interacts with a client
Access permissioned by role inside DiligenceVault
Drafts drawn only from approved internal sources
Human review required at more than one checkpoint
A complete trail of drafts, edits, and approvals
No autonomous execution, decisioning, or messaging
What DV Assist actually does
Every capability below connects directly to a problem the team was already living with. None of it requires replacing the people who know the business. It removes the mechanical work that was standing between them and the answers their clients need.
Autofill DDQs & RFPs
Drafts responses from existing materials such as fund DDQs, PPMs, policies, presentations, and LPAs, turning hours of manual effort into minutes of review. Each suggestion carries source attribution and a confidence score, so writers can see why it was made. For recurring requests, autofill draws on previously approved answers to the same questions, so the second time a familiar question arrives, the first draft is already strong.
One system for every request format
Requests arrive on DiligenceVault and in Word and Excel files, and both tracks run inside the same system. Offline documents are ingested and digitized automatically, so the team manages one workflow, one audit trail, and one content library regardless of how an allocator chooses to send their questionnaire. No separate handling. No parallel process. No gaps in governance.
Template detection for recurring questionnaires
When an allocator sends a quarterly assessment in Word and changes a handful of questions from one reporting cycle to the next, DiligenceVault detects that the document is a variant of a template it has already processed. It maps the edits against the prior version, so the team digitizes and populates only what changed rather than starting from scratch each time. The more often a template recurs, the faster it moves.
Writing and Content Assistant
Sharpens precision and tone with grammar correction, summarization, and readability edits. Predefined prompts such as Summarize, Elaborate, Trim, or Make More Readable let a writer refine a draft in place.
Smart Suggestion Engine
Surfaces the best answers from the approved content library and past responses so strong language gets reused consistently, and it learns from the firm's institutional knowledge to make those recommendations more relevant over time.
MyWork and weekly digest
Every contributor starts the week with a consolidated view of what is open, what is due, and what is waiting on them. MyWork surfaces the right tasks to the right people at the right time, and a weekly digest email means the whole team has what it needs to prioritize without chasing updates across inboxes or checking in manually.
The clock was the obvious win. The headroom mattered more.
The first thing anyone noticed was speed: sourced first drafts show up immediately, so deadlines stay reachable even as volume climbs. But for a team designed to take on more work without taking on more people, the result that counts is what the reclaimed hours free the team to do.
Faster first drafts, tighter deadlines met
DV Assist generates high-quality first drafts instantly from approved content. Drafting effort that used to gate every response now happens up front, so the team comfortably meets deadlines even as request volume climbs.
Expert time spent on judgment, not search
With source-backed drafts ready to review, SMEs face fewer repetitive, manual asks. Their time shifts to validating key facts, nuanced editing, and the high-judgment questions that genuinely need their expertise.
Higher throughput without proportional headcount
The team improved throughput and messaging consistency at scale, meeting its core objective: handling rising demand across SMAs, Models, and ETFs without growing the team in lockstep with the workload.
Quality and controls held, by design
Throughput gains did not come at the expense of control. Mandatory human review and a full audit trail of drafts, edits, and approvals keep quality and governance standards intact on every response.
The gains compound as the library stays current. Because DV Assist reuses approved content, its output improves the more current that content is. Keep product facts, disclosures, and positioning fresh and every later draft pulls the right language on its own; keeping responses inside DV rather than email holds onto the routing, traceability, and audit trail the team depends on.
"DV AI Assist is an internal-only, assistive capability that generates first-draft responses using pre-approved content already stored in DV, enabling writers to focus on review, refinement, and judgment-based decisions rather than manual drafting."
- Client RFP TeamThe economics of shifting hours from drafting to review
For the firm's internal ROI exercise, DiligenceVault put numbers to the shift from a search-and-paste way of working to an AI-assisted, governed one. Moving from manual drafting to source-backed review reduces the cost per response while raising institutional accuracy. The economics break down as follows.
| Metric | Current state | With DV Assist | Improvement |
|---|---|---|---|
| Total annual questions | 70,000 | 70,000 | - |
| Time per question | 18 minutes | 6.5 minutes | ~64% reduction |
| Annual labor hours | 21,000 hours | 7,583 hours | 13,417 hours saved |
| Capacity equivalent | ~11.5 FTEs | ~4.2 FTEs | 7.3 FTEs reclaimed |
Time per question is a blended value across factual or data-based questions, standard narrative, and customized narrative, estimated across content search, adaptation, SME clarification, four-eye review, and formatting. Reclaimed capacity equates to roughly $810K annually assuming a blended cost of $110K per FTE; substitute your firm's own FTE cost. Efficiency gains apply across both digital requests received through DiligenceVault and offline Word or Excel requests handled via AI-powered ingestion and autofill.
Productivity now, and a partnership that keeps going
Bringing DV Assist into the workflow was a real step forward in how the firm runs RFPs and DDQs, scaling the work and tightening governance in the same move. With that productivity behind them, the team is widening how it uses DiligenceVault.
Building on the productivity gains
With hand-drafting off the critical path, the team has room to push throughput and consistency further across a growing product set, while human review and audit trails keep every response governed.
Collaborating on what comes next
More capability is on the way this year, and the team plans to put it to work, extending a partnership built around running operations well with governance designed in from the start.
The complexity grew. The team handled it. The judgment stayed with the people.
More assets, more products, more questions, and more demanding diligence standards. What made it manageable was not more headcount; it was a system built to absorb complexity and give the people who know the business back the time to use that knowledge well.