One of the blessings of being a founder is that you can reach out to folks at the top of their game and have them speak candidly about their experiences. There's something about being an "outsider" that makes you less intimidating.
Over the past few months, I’ve had what felt like 100+ chats with institutional investors. I spoke to pensions, endowments, sovereign wealth funds, family offices. I've found that LPs are often juggling more managers, meetings, and memos than any human could reasonably track, and sometimes (emphasis on sometimes) they’ll slow down enough to talk about the job one CIO called “organized chaos.”
One theme that kept popping up is the tension between directional correctness and absolute precision. LPs know that manager projections, benchmark comparisons, and even their own portfolio models will never be 100% “right.” Markets change, strategies shift, and no spreadsheet survives contact with reality. Yet paradoxically, many LPs are hesitant to lean on AI if they can’t be absolutely certain it’s pulling the exact figure from a PPM, LPA, or quarterly report.
So why does it have to be 100% or nothing? I have two theories. First, LPs live in a world of compounding uncertainty. If you can’t control the outcomes, at least you want to control the data inputs. Second, there’s accountability. You can share the blame with colleagues for a missed diligence angle, but if you’re leaning on AI and it slips, that mistake feels like it sits squarely on your shoulders.
So how do you build an AI company for LPs with these concerns in mind? For starters, you have to be even more paranoid about accuracy than your users. That means validating the numbers before they’re surfaced. For example, if a manager says DPI is 1.4x, don’t just take it at face value, tie it back to the cash flow statements. Next, cite, cite, cite. Don’t just tell me the IRR is 18%; point me to the exact line in the quarterly letter or audited financials where that number lives. And finally, measure accuracy systematically. Just because output “feels” right in a memo draft doesn’t mean it is. LPs are allocating billions on these decisions—vibes don’t cut it.
At Sonar, we’re chasing the dream of 100% accuracy, but we know it’s a Herculean effort. What matters is getting as close as possible while being grounded, transparent, and humble.