CarbonAI

The product

What CarbonAI is building

CarbonAI builds agents that do the analytical work of a commercial real estate loan end to end, under the tagline "Autopilot, Not Copilot." There are four workflows. Underwriting agents take a deal from raw source documents (rent rolls, leases, operating statements, legal docs) to a populated, review-ready model in the lender's own template and against its own criteria. Due diligence agents track every document a deal requires and keep watching in the background without a new prompt. Asset surveillance runs quarterly across the whole loan book and outputs reports in the firm's format. A validation layer ties every figure back to the highlighted line in its source document. The design rule is that agents finish the job and only stop for the decisions a human owns, like an assumption on rent bumps or a reserves question. Every step is logged. The site says more than 40,000 documents have been processed, and the product is SOC 2 Type II with no training on customer data. Jeff Saul's pitch is a direct shot at horizontal AI: "Lenders don't need another chat window that they have to coach through every step." Greystone is the named customer, using CarbonAI to automate due diligence collection. The company also says it serves debt funds, mortgage REITs, agency multifamily lenders and publicly traded financial institutions, and claims up to 90% time savings on tasks like lease review, financial statement analysis and loan onboarding (self-reported).

The money

Why this matters

The refinancing wall is real. The Mortgage Bankers Association figure cited in coverage is $875B of commercial mortgages maturing in 2026, and every one of those loans needs to be re-underwritten, extended or worked out. Lenders' constraint is review capacity: the same analysts reading the same rent rolls. CarbonAI's mission statement says it outright, it wants to "break the link between deal volume and headcount." CRE lending is also a bad fit for generic copilots. The inputs are messy scanned PDFs, every lender has its own model template, and an extraction error flows straight into a credit decision. That's why the audit trail is the product here, not a feature. A vertical player built by former lenders, with a working wedge into a name like Greystone, has a clearer path than a general-purpose document AI that has to learn what a DSCR covenant is.

Investors: Cowboy Ventures (Caroline Duffy; lead), Timber Grove Ventures, Bling Capital, Max Ventures