Nine months ago, F2 didn't exist as a standalone company. Now the startup has $24 million in equity capital and a pitch that sounds almost quaint until you think about it: What if the future of AI in finance runs through Microsoft Excel?
The New York-based firm announced a $14 million seed round on June 10, 2026, led by HighlandX, with follow-on backing from Left Lane Capital, NFX, Y Combinator, and Torch Capital. The funding arrives as F2 emerges from its parent company Arc Technologies, carrying with it a thesis that feels both retro and prescient—that the spreadsheet, that most venerable of business tools, might be the killer interface for artificial intelligence in banking.
Their target market helps explain the enthusiasm. F2 builds what it describes as AI infrastructure for private credit funds and commercial banks, sectors where Morgan Stanley projects assets under management will hit $5 trillion by 2029, and which still do much of their analytical heavy lifting in, yes, Excel. It's an industry managing staggering sums with tools that would look familiar to a banker from 1995.
The Operating System Bet
Rather than trying to replace Excel, F2 has built what amounts to an AI layer that lives inside it. The platform pulls in multiple large language models—Opus 4.8, ChatGPT 5.5, Gemini 3.5 Flash—and channels their capabilities into workflows for deal screening, underwriting, and portfolio monitoring. The architecture generates native spreadsheet outputs with per-cell audit trails, a design choice that matters more than it might sound. In an industry where investment committees demand to see your work, black-box AI recommendations are about as useful as a chocolate teapot.
"Audit Mode," which launched in February, writes every calculation into source-linked Excel databooks built for exactly that kind of scrutiny. May brought "Institutional Knowledge," converting years of deal memos, financial models, and CRM data into a queryable knowledge base. The latest addition, announced with the seed round, is "Adam"—an AI agent orchestrating sub-agents to automate what F2 describes as investment team workflows.
The company claims its platform scored 95.25% on SpreadsheetBench Verified, a dataset of authentic spreadsheet questions from a 2024 academic paper. Independent confirmation of F2 topping those rankings proved elusive, though the figure appears in company materials.
Deals in the Thousands
Perhaps what caught investors' attention was the traction. F2 says deals analyzed on its platform jumped tenfold to more than 15,000 since January 2026, while monthly active users climbed 650% over the same stretch. The customer base now includes over 100 private credit funds and commercial banks. Among the named clients: Bain Capital and Live Oak Bank.
Craig Driscoll, who led HighlandX's investment, positioned the opportunity in almost evolutionary terms. "The firms that embed AI in sourcing, underwriting, and monitoring will build structural advantages," he said.
A managing director at Live Oak Bank, quoted in F2's announcement materials, pointed to "efficiency and standardization" in credit analysis workflows—the kind of quote that sounds anodyne until you consider how much of this work still happens manually, analyst by analyst, deal by deal.
Google Wants In
The seed round came bundled with another development: a partnership with Google Gemini Enterprise that will embed F2's platform directly into Google's financial services customer base. It's a distribution play, essentially, slotting F2 into Google Cloud's enterprise AI ecosystem at a moment when the tech giant has been aggressively adding vertical partners throughout 2026.
The capital itself will flow toward customized agentic workflow deployments, expanding the product suite (including a newly announced Portfolio Monitoring feature), and building out engineering and enterprise teams across New York, San Francisco, and London.
Timing Meets Velocity
That F2 went from spinout to $14 million seed round in under a year says something about both founder execution and market timing. For context, median seed rounds hovered around $2.4 million in 2025, according to Osler's venture deal data. Meanwhile, Morgan Stanley projects private credit assets under management will hit $5 trillion by 2029—creating what amounts to a wide runway for any infrastructure play that can bring standardization to an asset class that remains stubbornly manual.
There's a security angle worth noting. F2 maintains it operates with zero-day LLM data retention and doesn't train models on client data—table-stakes assurances when you're underwriting billions in loan commitments. Whether that defensive architecture holds as the platform scales across jurisdictions and regulatory regimes is, of course, a question that only time answers.
What's clearer is this: Nobody is replacing Excel. The spreadsheet has survived decades of would-be successors. F2's wager is that the next chapter doesn't require its replacement—just a rewiring of what runs underneath. In a world rushing toward AI agents and autonomous systems, there's something oddly reassuring about a bet that keeps the interface humans already trust. Even if, perhaps especially if, those humans are managing trillions.
