On June 17 the Federal Reserve held the federal funds target range at 3.50% to 3.75%, by a 12 to 0 vote.¹ The decision itself surprised no one. The statement did. It ran to about 130 words, down from 341 in April, and it dropped the language markets had read as a lean toward future cuts.² The closing line was blunt: “The Committee will deliver price stability.”
For anyone raising a fund this summer, that matters less as a rates call and more as a mood. Nobody is promising cheaper money soon, and allocators are taking their time. When capital is patient and selective, diligence gets deeper, not faster.

That is the backdrop for a question we hear from almost every manager we work with: can AI take the pain out of due diligence? Our answer is yes, with one condition that most tools on the market ignore.
The Burden Is Real, and LPs Have Said So
This is not a new complaint. When ILPA updated its Due Diligence Questionnaire in late 2021, the overview said plainly that the level of detail LPs require in upfront diligence “has increased,” and that the spread of customized questionnaires had created “an extraordinary administrative burden on all interested parties, including LPs, GPs and placement agents.”³
The DDQ also asks something many managers skim past. In the sections on deal sourcing and on screening and due diligence, ILPA asks managers to identify the technology tools they use, with a dedicated appendix for third parties and technology.³ If AI is part of how you source or diligence deals, a careful LP is going to ask how it works. “We use AI” is not an answer that survives an operational due diligence call.
Speed Is a Commodity Now
Drafting is the easy part. Any modern language model can turn a data room into a first draft of a 20-section DDQ response in minutes. Some of those drafts will even read well.
The trouble starts when an LP’s analyst picks one sentence and asks, “Where does this number come from?” If the honest answer is “the model wrote it,” you have a problem. A confident, unsourced sentence about fund terms or track record is worse than a blank, because now someone has to check everything else too.
We learned this the hard way building our own internal AI systems. Early versions produced fluent summaries that were mostly right. Mostly right is fine for an internal brainstorm. It is not fine for anything an investor will rely on.
Provenance Is the Product
The rule we now hold every diligence workflow to is simple: no claim without a source record. Every figure, date and statement the system produces carries a pointer back to the document, email or data-room file it came from. A reviewer can click from the answer to the evidence in one step.
Three design choices make that rule work in practice.
Label what is incomplete. One of the quieter failures we hit internally was a list read that stopped at a fixed page size without saying so. The output looked complete. It wasn’t. Our systems now state coverage explicitly, and a section built on partial data is marked partial rather than presented as finished.
Flag stale inputs. A figure pulled from a document that is months old should say so. In our own reporting, when an underlying source was out of date, the system learned to label the item stale and move on instead of quietly presenting old numbers as current.
Keep judgment with people. The AI assembles, cross-references and flags. A named person signs off. That is the only arrangement under which an investment committee can actually rely on the output.
What This Means for Managers Answering DDQs
If you are an emerging manager, you will answer the same core questions over and over, each time in slightly different wording. AI earns its keep here by maintaining one source-linked answer library, matching new questions to approved answers, and showing exactly what changed since the last version you sent.
The payoff is consistency. Inconsistent answers across LPs are a red flag in operational diligence, and they are usually caused by version sprawl rather than bad intent. A source-linked library fixes the cause.
What This Means for Allocators and Deal Teams
On the other side of the table, the same principles apply to reviewing a manager or a direct deal. AI can read the data room overnight, check figures against each other, and produce a list of gaps and inconsistencies by morning. That is genuinely valuable. But the list is only useful if each flag links to the page that triggered it, so the analyst spends time judging issues instead of hunting for them.
Regulators Are Buying Time. LPs Won’t.
On June 29 the Council of the EU gave final approval to a package that pushes back the EU AI Act’s high-risk obligations to December 2, 2027 for stand-alone systems and August 2, 2028 for systems embedded in products.⁴ Some managers will read that as breathing room.
We read it differently. Regulatory deadlines move. LP expectations don’t wait for them. The allocator asking how your AI produced an answer is not checking a compliance box; they are deciding whether to trust you with capital. Build provenance in now, while it is cheap to do, rather than retrofitting it after an LP asks the question you can’t answer.
Five Questions to Ask Any Diligence AI
- Can every output sentence be traced to a specific source document?
- How does the system mark data that is incomplete or out of date?
- Who signs off before anything leaves the firm, and is that recorded?
- What happens when two source documents disagree?
- How would you describe this tool in Appendix D of the ILPA DDQ?
If a vendor struggles with any of these, keep looking.
This is the thinking behind DDC, the due diligence use case within our Business Process AI (BPAI) practice. It is built to make diligence faster without asking anyone to take an answer on faith.
Make Your Diligence Faster Without Making It Riskier
Whether you are answering DDQs or reviewing them, we can show you what a source-linked diligence workflow looks like on your own documents.
References
- Board of Governors of the Federal Reserve System. “Federal Reserve issues FOMC statement.” June 17, 2026. https://www.federalreserve.gov/newsevents/pressreleases/monetary20260617a.htm
- CNBC. “Fed holds rates steady” (June 2026 decision coverage). June 17, 2026. https://www.cnbc.com/2026/06/17/fed-interest-rate-decision-june-2026.html
- Institutional Limited Partners Association. ILPA Due Diligence Questionnaire 2.0. Updated November 2021. https://ilpa.org/wp-content/uploads/2021/11/ILPA-DDQ-2.0.pdf
- Council of the European Union. “Artificial Intelligence: Council gives final green light to simplify and streamline rules.” June 29, 2026. https://www.consilium.europa.eu/en/press/press-releases/2026/06/29/artificial-intelligence-council-gives-final-green-light-to-simplify-and-streamline-rules/
For the supporting workflow, see how to create and manage an investor data room. To assess where AI belongs in your own diligence process, start with BPAI’s AI business process assessment.