In private equity, the most important research is also the most expensive. Before a fund commits several hundred million dollars to buy a company, it hires a firm like McKinsey, Bain, or BCG to go interview that company's customers, competitors, and industry experts, then package the findings into a 200-page report. The bill often lands between $500,000 and $1 million. It takes weeks. And because it costs that much, it is reserved for the biggest decisions - the deals large enough to justify the spend.
DiligenceSquared, a New York company from Y Combinator's Fall 2025 batch, is built around a single frustration with that arrangement: the insight is great, the price tag means you can rarely afford it. Its founders spent years on the buy-side and the consulting side of that exact transaction. Now they are using AI agents to run the interviews and assemble the report themselves, aiming to deliver the same class of research at roughly a tenth of the cost.
The pitch is not that AI writes a slick summary and you trust it. Anyone who has sat in an investment committee knows that is a non-starter. Every number in a diligence report gets challenged. So the company built the opposite of a black box: in a DiligenceSquared report, you can click any insight and read the exact transcript it came from.
01 / THE JOBWhat the company actually does
Commercial due diligence is the market-research leg of an acquisition. When a fund is weighing a target, it wants to know whether that company's customers actually like the product, whether they plan to keep paying, how they view the competition, and where the market is heading. The traditional way to find out is expert interviews - dozens of calls with customers, ex-employees, and specialists - synthesized against proprietary market data.
DiligenceSquared automates the slow parts of that loop. Its AI agents conduct moderated expert interviews at scale, transcribe them, and synthesize the findings into an interactive report. Human experts, many with consulting backgrounds, review the output before it reaches a client. The result is meant to hit the same investment-grade bar as a traditional engagement, delivered faster and cheaper.
The founders describe the deliverable less as a document and more as something you can interrogate. A partner can pull any thread - a claim about churn, a note on a competitor - and see who said it, when, and in what context. That traceability is the feature the whole product is organized around.
There is a reason the interview step is the one worth automating first. In a traditional engagement, the calls are the constraint: scheduling busy executives, waiting on their availability, running each conversation one at a time, then paying people to transcribe and tag what was said. It is the part of the process that scales worst and costs the most in human hours. Handing that to agents does not just save money - it removes the queue that makes the whole thing take weeks.
The synthesis step is where the company deliberately keeps people. Deciding what a set of interviews actually means - whether a pattern is signal or noise, whether a customer's grumble is a churn risk or a negotiating posture - is judgment, and judgment is what a fund is paying for. DiligenceSquared's stated division of labor is that the machine handles volume and the human handles meaning, with the transcript sitting underneath both so the reasoning can always be checked.
02 / THE MATHWhy $50,000 changes the behavior
The headline change is price. DiligenceSquared targets roughly $50,000 per project, against the $500,000 to $1 million a top-tier consultancy charges. That is not just a discount - it changes when a fund is willing to commission real research at all. When rigorous market work costs a million dollars, you run it only for the marquee deals. When it costs a fraction of that, you can run it earlier, and on more of the pipeline.
The business model mirrors consulting: project-based engagements, with some clients signing on for multiple projects over time. What is different is the unit cost, and the speed. Cutting a two-month process to something far shorter matters when a deal timeline is measured in weeks and a competing bidder is on the same clock.
Lowering the price also opens a category of work that barely exists today. A fund that would never spend a million dollars to research a smaller add-on acquisition might spend a fraction of that. A corporate weighing a new market entry might commission a study it would otherwise have skipped. The addressable moments multiply once the cost no longer forces a fund to ration its curiosity - and that expansion, not the discount on any single report, is the part of the story investors seem to be buying.
03 / THE PEOPLEOperators who sat on both sides of the table
The founding team is the credibility here. CEO Frederik Kofoed Hansen spent about six years as a Principal on Blackstone's private equity team across San Francisco, New York, and London, where he helped run multi-billion-dollar buyouts and commissioned the very diligence reports the company now automates. His co-founder, Soren Biltoft-Knudsen, ran the other side of that transaction as a Principal at BCG's private equity practice in New York, leading dozens of commercial diligence projects for major funds.
The third co-founder, Harshil Rastogi, is a repeat founder and former Google engineer who brings the production machine-learning side. The split is deliberate: one founder who bought the reports, one who built them, and one who can turn that judgment into working AI systems.
That biography also shapes the product philosophy. People who have defended a diligence report inside an investment committee know that the first question is always "how do you know that?" Building the interface around source transcripts is what you design when you have been on the receiving end of that question.
It is also a useful hedge against the standard objection to AI in high-stakes work. The worry with any language model is the confident, unsourced claim - the sentence that reads well and turns out to be invented. A diligence report is only as trustworthy as its worst-sourced line, and a single fabricated number can sink an entire engagement's credibility. Anchoring every insight to a transcript is the company's structural answer: the model can summarize, but it cannot quietly make things up, because the receipt is always attached.
04 / THE MARKETWhere it fits, and who else is here
Commercial due diligence is a multi-billion-dollar market, and it has been remarkably slow to change. The core workflow - find the right experts, interview them, synthesize - looks much as it did before modern AI existed. That inertia is precisely what makes it an opening: a large, high-stakes budget line that has not yet been rebuilt around automation.
DiligenceSquared is not alone in noticing. Its closest AI-native competitor, Bridgetown Research, raised a $19 million Series A in early 2026. The larger incumbents remain the consultancies themselves and the expert-network firms that broker interviews by hand. DiligenceSquared's wager is that the combination of AI scale plus human review, wrapped in source-level traceability, is the specific mix that investors will trust with a real decision.
The positioning is narrow on purpose. Rather than chase the flashy consumer demo, DiligenceSquared picked one of the least glamorous and most defensible corners of the market: the research that decides where large pools of capital go. It is a budget line that is high-stakes, recurring, and staffed today by expensive people doing repetitive work - the kind of profile that rewards a tool built by people who know the workflow from the inside rather than one guessing at it from the outside.
Backing that wager is a $5 million seed round led by Relentless, the firm founded by former Index Ventures partner Damir Becirovic, with participation from Y Combinator, Amino Capital, Founder Factor, Multimodal Ventures, Twenty Two Ventures, and angel investors. The company says it has already completed projects for several of the world's largest private equity firms and mid-market funds, with clients representing more than $2 trillion in combined assets under management.
Three things make commercial diligence slow and expensive. DiligenceSquared targets each:
05 / THE BETCan AI sit in the investment committee?
The open question for any tool in this category is trust. A fund is not buying a report; it is buying the confidence to move a large amount of money. That is why DiligenceSquared keeps humans in the loop and puts the transcript one click away from every claim. The design is a direct answer to the skepticism a partner brings into the room.
If the approach holds, the interesting effect is not that a few reports get cheaper. It is that rigorous market research stops being a luxury reserved for the biggest deals and becomes something a fund can afford to run routinely. That is a different market than the one the consultancies serve today - one defined less by the size of the check and more by how often investors are willing to actually check.