A prospective LP sends a 120-question DDQ. You've answered most of these questions before — in last quarter's StepStone questionnaire, in the ILPA template you filled for the fund-of-funds, in your ADV. But not in this format, not in this order, and not in a way anyone can find in under an hour.

So the answers get written again. This is the actual DDQ problem: not that the questions are hard, but that your firm's approved answers don't accumulate anywhere reusable.

The overlap is the asset

Across fundraises, DDQs repeat heavily — the same operational, compliance, team, and process questions in slightly different wording. Which means the economics of DDQ work are supposed to improve every time you complete one. Your tenth DDQ should cost a fraction of your first.

For most funds it doesn't, because the reuse mechanism is a folder of old Word documents and one person's memory. A folder is not a knowledge base. The difference is three properties:

Answers are stored as approved units, not buried in documents. When the concentration-limit question shows up in new wording, the system matches it to the answer your firm already approved — it doesn't ask a model to improvise.
Every answer carries a citation. Which document, which section. An LP's diligence team cross-references your DDQ against your ADV and your fund docs; an answer that can't point to its source isn't a draft, it's a risk.
Every answer carries a confidence score. Green: your documents clearly support this — review is a skim. Yellow: partial grounding — read it properly. Red: your documents don't cover this — write it yourself, and note the gap, because the gap itself is information. A cluster of red answers before a fundraise tells you what to document now, not mid-diligence.

None of this requires a frontier-scale model. It requires structure: match against approved answers first, generate only what's missing, cite everything, score everything, and put every answer — green included — into a review queue where a human approves it before it goes anywhere. Fundry never sends anything on its own.

What it looks like in practice

Upload the questionnaire. The DDQ agent extracts the questions — including out of the table-heavy Word and PDF formats LPs love — classifies them, drafts each answer from your firm's knowledge base, and returns the batch with citations and confidence scores. Across our design partners, DDQ drafts matched their approved answers 85–94% of the time. Treat that range as what it is — a design-partner result, not a promise. The honest test is your own questionnaire.

The review is where your time goes now: minutes confirming green answers against their citations, real attention on the yellows and reds. What used to be a two-week internal project becomes a same-day draft and a focused review session.

And the next DDQ is faster than this one, because every answer you approve becomes part of the knowledge base the following questionnaire draws from. The asset compounds. That's the point.

Try it on your hardest one

We keep this test free because it's the fastest way to know whether any of the above is true for your fund: fundryai.com/ddq/free. No signup to start. Upload the DDQ you dread — the 150-question institutional one, not the softball — and look at three things in the output: how many greens need no real editing, whether the citations point where you'd expect, and what the reds tell you about your document library.

If the drafts aren't good against your documents, you'll know in twenty minutes. That's the evaluation standard we think every fund should hold every AI vendor to.