From Raw Notes to Research Conviction with DV Assist
How GEM Investments transformed a manual, time-consuming documentation process into a fast, consistent, and trusted workflow with DV Assist.
A team doing serious work held back by the work around the work
GEM Investments is an experienced institutional asset allocator. Their ODD team, which is also their Compliance team, is rigorous, experienced, and deeply committed to getting manager diligence right. But like most allocator teams operating at scale, they faced a persistent and familiar tension: the time required to document and synthesize manager ODD research was eating into the time available to actually do it.
GEM wasn't looking for shortcuts. They were looking for infrastructure that could handle the mechanical burden of documentation without compromising the integrity of their process. They were also early believers in AI's potential to reshape how institutional investment work gets done. What they needed was a platform that met that belief with substance.
They found it in DiligenceVault.
Three friction points that compounded across every manager review
Before DV Assist, GEM's ODD team ran a workflow that any experienced ODD professional would recognize immediately. It was thorough. It was consistent in judgment. But it was also slow, effortful, and prone to the kinds of variability that come from manual, individual work at scale.
Meeting notes that didn't write themselves
After every manager meeting, the ODD team faced the task of converting rough, bullet-point notes captured in the moment, across varying formats into polished, structured internal write-ups. It was time-consuming, and the quality depended heavily on individual effort and style. The institutional knowledge was there. Systematizing consistently was the problem.
Complex documents, manual extraction
Manager documents such as pitch decks, DDQs, offering memoranda, policy documents, and audit reports are dense by design. Manually reading through lengthy materials to pull out the specific data points relevant to an ODD review is skilled, necessary work. But it's also work that consumes hours better spent on analysis and judgment, not search and extraction.
Assembling the full picture was a process in itself
Bringing it all together across notes, document extracts, context, and prior history into a cohesive internal memo required significant assembly effort. Even with an experienced team and clear templates, the process introduced formatting variability and took real time before the actual review could begin.
An end-to-end workflow without the end-to-end effort
GEM implemented DV Assist, DiligenceVault's integrated AI, to address each friction point directly. The approach wasn't to replace the ODD team's judgment it was to stop making that judgment wait.
Document Analysis: Insight extraction at the speed of search
DV Assist's Document Analysis capability extracts the most relevant information from complex fund materials in minutes, not hours. The ODD team can pull targeted data with confidence rather than reading cover to cover and highlighting manually. Key to trust: every extracted insight is paired with source citations and a confidence score, so team members know exactly where information came from and how confident the AI system is. Nothing is taken on faith.
AI Writing Assistant: From rough notes to professional narrative
Team members input their meeting notes, however rough and fragmented, and DV Assist can transform them into a structured, professional first draft. The voice stays theirs. The institutional knowledge stays intact. The formatting variability disappears. What used to take a few hours of careful editing now potentially takes minutes of review.
Compilation into a structured first draft memo inside DiligenceVault
With digital DDQs, document insights, and written narrative ready, team members compile everything into a structured first draft directly within the platform. The memo doesn't have to be assembled from multiple windows across different tools. It comes together in one place, from one workflow, building directly on the manager's existing knowledge base in DiligenceVault.
Confidence in the output isn't assumed it's built in
For an ODD team, trust isn't a nice-to-have, it is a key part of the value proposition. Team members who don't trust the output don't use it, and rightly so. GEM's team was direct about this from the start.
What made DV Assist trustworthy in practice
GEM's team singled out two specific features as decisive for adoption not because they were impressive, but because they solved the real objection: how do I know this AI output is right?
These aren't incidental product features of DV Assist. They are the design decisions that make an AI application usable in a high-stakes institutional context where a wrong answer that looks right is more dangerous than no answer at all.
GEM's experience isn't an outlier it's a signal
When we asked managers and allocators what drives their primary AI tool choice, ease of use came out on top ahead of other important factors such as performance, integrations, security, and cost. GEM's team called DV Assist "very easy to use" within weeks of rollout. The data and the experience point in the same direction.
Primary reason for AI tool choice DiligenceVault poll - May 2026
of managers and allocators cite ease of use as the primary reason they choose an AI tool.
GEM's experience is evidence that DV Assist is built for how this industry actually adopts tools not just capability-first, but usability-first.
Results that showed up immediately and keep compounding
GEM began their rollout in Q4 2025. The impact was not gradual. It was immediate, visible, and measurable within the first reviews.
1–3 hours saved per manager review
The range reflects the volume of materials. The time is real reclaimed from formatting, reading, and assembly, and redirected toward analysis.
Key insights surfaced in minutes
What previously required reading through entire fund documents is now part of a structured blueprint. Team members get the information they need, sourced and scored, before they would have finished page three of the PDF.
Standardized starting point across the team
AI-generated drafts don't carry individual formatting habits or stylistic variation. The team works from a shared, professional baseline and then applies their judgment from there.
Team members on analysis, not administration
The shift is qualitative as much as quantitative. Team members spend their cognitive energy on the decisions that matter not on search, copy-paste, and document assembly.
There is also a compounding effect that grows with time. Every insight extracted through DV Assist doesn't disappear after the memo is written. It becomes part of the manager's centralized knowledge base within DiligenceVault available for future reviews, LP reporting, and network intelligence. The work done today makes tomorrow's work faster and richer.
A partnership building toward what's next
GEM is an active partner in shaping where DV Assist goes. The success within the ODD team has opened the conversation about expansion and the roadmap reflects that collaboration directly.
Expansion within the firm
Other diligence teams process large volumes of offering documents, legal documents, manager letters, and performance materials. The same workflow efficiency that transformed ODD will be considered for other business areas.
Context mapping at scale
As context grows, so does the need to balance AI output with accuracy. Advanced context capabilities are on the roadmap developed with direct input from GEM's team.
External data connectors
The goal is seamless access to manager materials wherever they live. External data integrations aims to eliminate the friction of document transfer and make DV Assist's analysis available across the full scope of the manager research process.
Deeper network intelligence
Every interaction on DiligenceVault, every document, every review, every insight strengthen the data layer. GEM's continued engagement contributes to their knowledge base that becomes more valuable with every manager on the platform.
The path from raw research to investment conviction should not be a documentation problem.
DV Assist exists to clear that path so that teams like GEM can spend their expertise on the decisions that actually matter.