Last week, I asked Apollo Co-President Scott Kleinman how his firm is leveraging AI in private credit. Without hesitation, he told me that Apollo is going all in.
That moment stuck with me鈥攂ecause ten years ago, I was in the trenches doing the kind of work that AI is now poised to automate. Even at the world鈥檚 top investors like Apollo.
In my private equity days, I spent sleepless nights combing through Excel files, building models from scratch, and meticulously analyzing data rooms鈥攎ostly for deals that would never close. It was important work, but much of it was low-leverage. My fund was paying top-tier talent to move numbers from one spreadsheet to another.
It wasn鈥檛 just inefficient. It was a barrier to scaling the asset class.
Today, that barrier is falling fast.
AI is rapidly upending the private markets. Particularly in private credit, where AI adoption is inflecting. Not in some hypothetical future鈥攂ut right now, in workflows that have long resisted automation.
Lenders are using AI to qualify borrowers faster, automate diligence, and generate detailed financial memos that used to take teams of analysts days to complete. Borrowers are streamlining capital raises and reporting with AI-powered tools. And the entire process is becoming more efficient, more scalable, and more precise.
But here鈥檚 the catch: generic AI models won鈥檛 get us there. The stakes are too high, and the workflows too nuanced, for shallow reasoning or hallucinated answers.
Tools like OpenAI鈥檚 GPT-4o, Google鈥檚 Gemini, and Anthropic鈥檚 Claude are incredibly powerful at processing natural language. But when it comes to complex financial workflows鈥攔easoning over raw Excel files, calculating unlevered free cash flow, mapping covenants鈥攖hese general-purpose LLMs consistently fall short.
We saw this firsthand. My company, Arc, benchmarked GPT-4o against our internal AI system on a series of tasks in private credit. Finding net debt? GPT got it wrong more often than not. Calculating leverage ratios? Missed. Computing customer cohort retention over a 24-month span? Failed.
Financial analysis is a different beast. It鈥檚 multi-step, math-heavy, and intolerant to hallucination.
That鈥檚 why the future of AI in finance isn鈥檛 horizontal鈥攊t鈥檚 vertical.
At Arc, we鈥檝e spent the last two years building Arc Intelligence: a vertically integrated, agentic AI system trained specifically on the workflows of private credit lenders. It鈥檚 model-agnostic, multi-modal, and capable of digesting unstructured data like Excel files with a level of precision that far exceeds generic models (~3x higher accuracy than foundational LLMs like OpenAI, as of this column).
More importantly, it doesn鈥檛 just answer questions鈥攊t completes workflows. Our system pulls, cleans, and analyzes data, then writes diligence reports grounded in verifiable source material.
Think of it as the financial analyst that never sleeps. (sidebar: I tried not to sleep on those deals either鈥攂ut even I would burn out by day seven of chasing a dead process.)
This is personal for me. I remember what it felt like to lose entire weekends to data entry, formatting, and memo writing. I built what I wish I had in the bullpen.
And it鈥檚 not just me. We鈥檙e seeing demand surge from across the private credit market. Over 300 lenders have joined the waitlist to use Arc Intelligence. The LP dollars are flowing to the funds that have real AI strategies. And deal teams are looking for every edge they can get in a highly competitive market.
AI is no longer a luxury. It鈥檚 table stakes.
This wave of adoption won鈥檛 just make private credit faster. It will expand the market鈥攅nabling lenders to serve more companies, originate better deals, and redeploy their teams toward higher-leverage work.
We鈥檙e still early in this shift. But one thing is clear: the private credit funds that embrace vertical AI will define the next era of the market. The rest will be left behind.
And for those of us who spent years buried in spreadsheets at 3 AM, that future can鈥檛 come soon enough.