This is your weekly update on what’s breaking (and fixing) modern lending.

Three insights. Three minutes. Zero fluff. 

Why Detection Needs to Go Beyond Onboarding

Fraud isn’t standing still - it’s evolving with generative AI.

👻 The scary part? 
Generative AI has turned synthetic identity fraud into a $1B+ problem - and Deloitte projects it will hit $5B by year-end 2025. Deepfakes and AI-doctored IDs are slipping past KYC tools, fueling a 25% rise in fraud losses last year.

🚨 The real problem? 
Fraudsters can fake documents, but not behaviors. A legitimate applicant takes time, uses consistent devices, and shows steady income flows. Synthetics race through apps, switch devices mid-process, or cycle money in and out. Yet most financial institutions still only check documents at onboarding.

⚡ Action step:
Move from one-time checks to continuous monitoring. Track behavioral red flags - upload speed, device consistency, deposit reversals - alongside biometrics. These lightweight signals catch AI fakes in real time, without adding friction for genuine customers.

Hyper-Personalization Misses the Human “Why”

Financial Institutions collect mountains of transaction data, yet often miss the why behind customer and member behavior. Emotional drivers like loss aversion or financial anxiety shape financial decisions more than transaction data ever shows.

💡 The psychology? 
As Chris Nichols of SouthState Bank recently wrote, behavioral scientists call this the “Ikea Effect”: when customers put effort into shaping something, they value it more. In banking, that means a applicant, customer, or member who sets loan goals, chooses account perks, or names a savings fund feels greater ownership and loyalty than one who just receives a generic product.

🚨 The real problem? 
Most personalization engines still run on data alone - “you spent on travel, here’s a travel card.” Customers tune out, loyalty slips, and engagement stalls. Data without behavioral context feels generic.

⚡ Action step:
Blend behavioral cues with data analytics. Let customers and members “assemble” parts of their journey - progress trackers in onboarding, goal-setting in savings or loan repayment, perk selection in accounts. Each micro-effort builds attachment and turns personalization from a data exercise into a loyalty engine.

The AI Implementation Divide

MIT’s State of AI in Business 2025 report delivered a sobering stat: 95% of custom enterprise AI tools never make it past pilot. At the same time, OpenAI’s Sam Altman compared today’s AI hype to the dot-com bubble, warning investors are “overexcited.”

Yet the picture isn't uniformly bleak. The data reveals a critical divide in AI implementation success.

🥇The success stories?
Bank of America earmarked $4B of its $13B tech budget for AI this year - logging 3B interactions with Erica and enabling 90% of employees to use AI daily. JPMorgan Chase is following suit, rolling out an in-house LLM Suite to hundreds of thousands of staff.

And just last week, Grasshopper Bank rolled out a new AI tool for small business banking that goes beyond the usual balance alerts. Instead of static messages like “your account dropped below $500,” their system looks at payroll, invoices, and cash flow trends - then warns: “You’ll be short $22,000 in nine days if nothing changes.” It’s a shift from dashboards that only report what happened, to tools that actually coach businesses on what’s coming.

Ron Shevlin, Chief Research Officer at Cornerstone Advisors, called the launch a “big deal.” His take: most platforms already send alerts, but Grasshopper’s Model Context Protocol (MCP) makes alerts smarter - predicting what’s next, not just reporting the past - and lays the foundation for AI agents that can guide decisions directly inside financial workflows - a step toward next-gen digital experiences.

💡The psychology of success?
Winning financial institutions don’t chase “AI transformation.” They focus small-solving one painful workflow at a time. Bank of America built tools that save junior bankers hours on meeting prep. When features failed - like voice AI in Erica when customers preferred text for financial queries - they were cut fast.

🚨 The real problem? 
Most institutions stay trapped in pilot purgatory because projects launch with vague goals. “Test AI for lending” isn’t a plan. “Cut 3 hours of manual prep from each loan package” is. MIT found external AI partnerships succeed 2x more often than in-house builds - yet most financial institutions still try to build everything themselves.

⚡ Action Step:
Before your next pilot, write down the exact manual task the tool will eliminate. If the answer is fuzzy, you’re on the road to pilot purgatory.

🔍 Want to go deeper?
As Sophie Heller, Chief Transformation Officer at BNP Paribas CPBS, told Jim Marous on Banking Transformed: behavioral sciences are now essential tools for transformation leaders. Winning financial institutions aren’t just plugging in AI - they’re redesigning journeys around how people actually behave.

🙌 Help Us Grow

Know someone in lending who’d benefit from these insights? Forward this to them - and hit reply if there’s a topic you want us to cover.

📅 Next newsletter drops Tuesday, 09/02. Happy Labor Day!