What is AI underwriting? AI underwriting is the use of artificial intelligence to read deal documents, normalize financials, and model returns so an analyst can reach a credit or investment decision faster. The pressing question for 2026 is not whether AI underwriting is fast, because it clearly is, but whether commercial real estate investors can trust it with decisions that move millions of dollars. For the full picture on how these tools fit into deal financing, see our guide to AI in CRE finance and capital markets.
Key Takeaways
- AI underwriting now reads rent rolls, leases, and a T12 in minutes, with banks reporting time to decision cut by 50% to 75% on commercial loans.
- A record refinancing wave is forcing the issue: the MBA expects about $805.5 billion in 2026 commercial originations and $875 billion of loans maturing this year.
- AI earns trust on speed and data extraction, but not yet on final judgment, where a human must own the credit and capital decision.
- Regulations like ECOA and CRA require lending decisions to be explainable and free of bias, which limits how far a black box model can go.
- The safe pattern is AI for the first draft and a human for the final call, backed by a documented quality control workflow that verifies every output.
AI Underwriting Explained
AI underwriting applies large language models and machine learning to the grunt work of deal analysis: extracting numbers from a rent roll, normalizing operating expenses, recomputing net operating income (NOI), and stress testing assumptions against market data. Done well, it compresses days of analyst time into minutes while leaving the investor in control of the decision.
The mechanics matter. An AI tool can scan a trailing twelve months (T12) statement, separate operating expenses from capital items, and recompute NOI without the financing distortions that trip up a rushed analyst. It can then flag that a 1.20x debt service coverage ratio (DSCR), calculated as NOI divided by annual debt service, leaves a thin cushion if rents soften, or that a 6.0% cap rate looks aggressive for a weakening submarket. The tools have become genuinely useful. The open question is how much of the final call you hand over to them.
Why 2026 Is Forcing the Trust Question
2026 is forcing the AI underwriting trust question because the volume of deals that need underwriting is surging at the same moment the tools became good enough to help. The Mortgage Bankers Association forecasts total commercial mortgage originations of about $805.5 billion in 2026, up roughly 27% from 2025, with multifamily originations near $399.2 billion.
Behind that surge sits a refinancing wall. The MBA estimates that 17%, or roughly $875 billion, of the $5.0 trillion in outstanding commercial mortgages matures in 2026, with the 10 year Treasury expected to average 4.2%. Every one of those maturities needs fresh underwriting under tighter conditions, and lenders cannot staff their way through the backlog fast enough. That is why banks are turning to AI to cut the time between application and credit committee, the most expensive bottleneck in the origination stack. At the CRE Finance Council mid year conference in June 2026, panelists stressed that future value gains must come from NOI growth rather than cap rate compression, which puts even more weight on accurate, fast underwriting.
Where AI Underwriting Earns Trust
AI underwriting earns trust in the parts of the job that are mechanical, repetitive, and easy to verify. This is where the time savings are real and the downside of an error is small, because a human reviews the output before it drives a decision. Marcus & Millichap CEO Hessam Nadji has noted that the first wave of AI value in CRE shows up exactly here, in underwriting processes, analytics, and data parsing.
Concrete wins include extracting and standardizing rent rolls and leases, reconciling a seller's pro forma against the actual T12, summarizing loan documents, and generating a first pass model an analyst can check. Banks deploying purpose built AI underwriting report time to decision reductions of 50% to 75% on commercial loans, according to industry surveys, and Morgan Stanley has estimated that AI could automate roughly 37% of tasks across the sector over time. Used this way, AI does not replace the underwriter; it removes the drudgery so the underwriter can spend time on judgment. If you want a repeatable process for catching AI mistakes, our CRE quality control workflow for verifying AI underwriting outputs lays out the checks step by step.
Where AI Underwriting Should Not Be Trusted Yet
AI underwriting should not be trusted to make the final credit or investment decision on its own. The reason is accountability: when a model is wrong, it is often wrong with total confidence, and no lender wants to explain to investors or regulators that a machine approved a loan that later defaulted. As one underwriter put it, few would hand an AI a blank check to deploy $20 million.
Three limits stand out. First, models can hallucinate or misread a nonstandard document, and a single wrong NOI or DSCR can flip a decision. Second, the Equal Credit Opportunity Act (ECOA) and the Community Reinvestment Act (CRA) require lending decisions to be explainable and free from discriminatory patterns, even unintended ones, which a black box model cannot guarantee on its own. Third, judgment about sponsor quality, business plan risk, and local market dynamics still rests on experience the model does not have. This caution is rational rather than anti technology: surveys show that while 92% of corporate occupiers have initiated AI programs, only about 5% report achieving most of their AI goals, a gap that mirrors the distance between running a pilot and trusting AI with real money. JLL likewise stresses that model transparency and data quality are prerequisites, not afterthoughts. Investors who want help drawing this line for their own shop can connect with The AI Consulting Network.
How CRE Investors Can Use AI Underwriting Safely
The safe pattern is straightforward: let AI produce the first draft, then make a human own the final decision through a documented quality control process. The investors getting real value from AI underwriting are not the ones who trust it blindly or reject it outright, but the ones who built a workflow around it.
- Verify every critical number: Recheck AI computed NOI, DSCR, cap rate, and loan to value (LTV) against source documents before they reach a decision memo.
- Keep a human in the loop: Require analyst and committee sign off on the credit decision itself, no matter how clean the AI output looks.
- Stress test the inputs: Use AI to challenge a sponsor's assumptions, not just to repeat them. Our guide on AI tools to stress test a sponsor's underwriting shows how an LP can do this on an offering memorandum.
- Match the tool to the task: AI lender matching platforms like the one covered in our look at AI commercial mortgage platforms can speed sourcing, but the credit analysis still needs human review.
- Document for auditability: Keep records of what the AI produced, what a human changed, and why, so the decision survives regulator and investor scrutiny.
CRE investors who want to build this kind of disciplined AI underwriting process, rather than bolt a chatbot onto an old workflow, can reach out to Avi Hacker, J.D. at The AI Consulting Network.
Frequently Asked Questions
Q: Can AI underwrite a commercial real estate deal on its own?
A: Not responsibly in 2026. AI can extract data, recompute NOI and DSCR, and produce a first pass model in minutes, but the final credit and investment decision should stay with a human. Regulation and accountability both require a person to own and explain the call.
Q: How accurate is AI underwriting?
A: AI is highly accurate at structured tasks like reading a rent roll or reconciling a T12, but it can misread nonstandard documents and occasionally fabricate details. That is why every critical figure should be verified against source documents before it informs a decision.
Q: Why are CRE lenders adopting AI underwriting so quickly in 2026?
A: A record refinancing wave is driving it. With roughly $875 billion in commercial mortgages maturing in 2026 and originations forecast near $805.5 billion, lenders are using AI to cut time to decision by 50% to 75% and clear the backlog without sacrificing discipline.
Q: What regulations affect AI in CRE lending?
A: The Equal Credit Opportunity Act and the Community Reinvestment Act require lending decisions to be explainable and free of discriminatory patterns. Because many AI models are hard to interpret, lenders must keep humans accountable and document how each decision was reached.