How AI is Quietly Rewriting the Risk Appetite of CRE Lenders
By Katie Quilligan, Investor at BankTech Ventures and Rachel Gammons, Client Engagement Director at Blooma
Simple Subscribe
Subscribe Now!
Artificial intelligence isn’t bulldozing traditional credit standards in commercial real estate lending. Yet it is quietly reshaping how CRE lenders evaluate risk, one deal at a time. In many ways, that’s exactly what the industry needed.
Lenders are under immense pressure. Portfolios are growing, markets are volatile and regulatory requirements are relentless. AI promises faster underwriting, smarter insights and fewer surprises. But in an industry built on relationships, judgment and local nuance, trust doesn’t come automatically.
Need to Know:
- AI isn’t rewriting CRE credit playbooks — it’s sharpening them. The biggest gains aren’t looser standards, but clearer visibility into risk, especially in gray-area deals that once stalled underwriting.
- The real hurdle isn’t technology, it’s trust. Algorithm aversion, regulatory scrutiny and past AI misfires keep lenders cautious — making transparency and human oversight non-negotiable.
- Portfolio intelligence is where AI delivers outsized value. From early warning signals and concentration risk to multi-factor stress testing, AI surfaces threats traditional models miss.
- The payoff is resilience, not just efficiency. Banks winning with AI aren’t chasing speed alone — they’re reducing downside risk in an environment where one bad loan can erase dozens of good ones.
The Trust Gap That’s Holding Banks Back
That hesitation is rational. Researchers call it algorithm aversion: people prefer human judgment, even when machines are demonstrably more accurate. In CRE lending, one misstep can cost millions.
Take Zillow’s iBuying collapse: a $500 million loss driven by AI models that couldn’t account for local knowledge, renovation realities and market timing that experienced investors instinctively weigh. For CRE lenders, it became proof that algorithms can’t replace judgment in markets where every property has unique characteristics.
The data tells the story. For example, one McKinsey study found that fewer than 25 % of banks had fully deployed generative AI use cases, even though over half said it was a strategic priority. Even well-resourced teams stumble when they treat AI like a magic wand instead of what it actually is: a tool that works only when used thoughtfully, with people in the loop.
When AI Changes the Game
AI’s influence operates invisibly in the background, surfacing patterns and outliers that might take humans days or weeks to notice. That doesn’t mean lenders approve riskier deals. It means they see risk with more clarity.
Consider the “gray zone” deals: loans too risky to approve outright, but not risky enough to reject. Historically, many were shelved because they didn’t fit neatly into a credit box. AI illuminates these overlooked opportunities through speed, structured data and pattern recognition. Risk appetite isn’t looser. It’s better informed.
As teams grow comfortable with AI, something subtle happens. Underwriters become willing to explore less familiar markets or atypical deal structures, all grounded in data rather than gut instinct. This isn’t recklessness. It’s what happened with spreadsheets and credit scoring models: technology didn’t replace judgment; it expanded it.
Portfolio Intelligence That Actually Matters
The real transformation isn’t happening at the deal level. It’s happening at the portfolio level.
Traditional stress testing has always been backward-looking: take historical scenarios, apply static assumptions, hope the results approximate reality. AI flips this. Advanced systems can now model dynamic, interconnected scenarios. What happens when remote work permanently reduces office demand in secondary markets while interest rates spike and credit availability tightens?
Stress-testing against cascading shocks. Not just “rates go up” scenarios, but multi-factor modeling that accounts for correlations across asset classes. One example: imagine a bank discovering their multifamily portfolio has hidden concentration risk, not to geography, but to a specific tenant income bracket vulnerable to employment disruptions in certain industries. Traditional stress testing would never surface this connection.
Modeling climate risk with precision. It’s no longer enough to flag FEMA flood zones. AI now integrates wildfire modeling, hurricane projections, heat stress impacts and insurance availability trends. Recent research using MSCI’s Geospatial Asset Intelligence dataset demonstrated that physical climate risk materially affects real estate valuations. The study examined firms impacted by hurricanes between 2022 and 2024 and found statistically significant underperformance, providing real-world evidence for why granular location data and vulnerability assessments are essential for forward-looking investment analysis. This validates what AI-enhanced climate modeling enables: moving beyond simple FEMA flood zones to sophisticated, multi-factor risk assessment that accounts for wildfire smoke risk, heat stress impacts and insurance availability trends across entire portfolios.
Catching portfolio drift early. As individual underwriters approve compliant deals, AI can detect when the aggregate portfolio is slowly shifting toward higher risk. For instance, a bank could spot sooner their average loan-to-value ratio creeping up because underwriters consistently landed near the high end of acceptable ranges. AI catches what quarterly reports can miss.
Detecting tenant risks through behavioral signals. Financial metrics like timely rent payment provide only a partial view of tenant health. Research on AI-powered business intelligence for commercial property shows that behavioral signals—declining work order responsiveness, reduced access card swipes, decreased facility usage—often precede formal exit notifications by months1. AI excels at aggregating these multidimensional signals across property management systems to identify tenant distress early. For CRE lenders, this creates portfolio-level intelligence that traditional underwriting misses.
Running scenarios in minutes instead of weeks. When Silicon Valley Bank collapsed in March 2023, banks using portfolio-level AI ran exposure scenarios in hours. Those relying on manual analysis took weeks and made decisions in the dark during the critical window.
The Federal Reserve now expects banks to demonstrate robust scenario analysis capabilities. AI is becoming the only practical way to meet them.
Why Transparency Wins
The most effective AI systems are traceable. Every recommendation can be traced back to underlying data. Logic is consistent. Humans remain in control, reviewing and contextualizing.
This matters as regulators scrutinize model risk management. The OCC, FDIC and Fed have made clear that “the model said so” isn’t an acceptable justification for credit decisions. The institutions succeeding with AI aren’t those with the fanciest algorithms. They’re the ones that designed transparent systems their own credit committees can interrogate and trust. When lenders conquer that trust gap, they begin to see AI change the game.
The Real Payoff
One regional bank using Blooma, a CRE intelligence platform, layered AI onto a $17 billion CRE portfolio. Manual data entry dropped 80%. Underwriters could focus on high-value tasks. Risk monitoring became proactive rather than reactive.
But here’s what matters most: the real value isn’t cost reduction. It’s risk reduction. AI-enabled lenders are catching problems earlier, spotting concentration risks faster and making more informed decisions in uncertain environments. In a business where a single bad loan can wipe out the profit from 50 good ones, that edge is existential.
Three Questions That Define Success
Can we explain our AI-driven insights to regulators, credit committees and ourselves? If your team can’t articulate why the AI flagged a deal or surfaced a portfolio risk, you have a transparency problem, not a capability.
Are we stress-testing against scenarios that could actually hurt us? Or just the ones that are easy to model? The difference between these two approaches is the difference between risk management and risk theater.
How is our risk appetite evolving and is our team aligned? AI will surface opportunities and risks that weren’t visible before. Your credit culture needs to evolve with that visibility, not against it.
Done right, AI doesn’t change credit policy overnight. It refines it. It expands the lens through which underwriters see the world, surfacing insights that humans might miss and giving teams the confidence to make data-backed decisions in the gray zones of risk.
The banks that figure this out won’t just be more efficient. They’ll be more resilient and more competitive.
