Hugh Khan

Predicting whether a borrower will pay their next EMI is a math problem that directly affects daily targets of our clients. We recently have rolled out a framework for Early Warning Signals (EWS) that moves beyond just looking at past payment history. The core of my approach uses random forests to find the base probability of a default in the next EMI. However, internal data doesn't tell the whole story. To get a clearer picture, we integrated macroeconomic signals like inflation rates and industrial production indices. The technical trick lies in using lagged variables. For example, a spike in fuel prices two months ago might only start affecting a retail borrower's wallet today. By layering these external shifts onto borrower-level data, the model provides a more realistic risk score. Aisa model kaam asaan kar deta hai, especially when you have to decide which accounts to prioritize for follow-ups. Aggregating these risk indicators to different cohorts, helps build risk profile for different cohorts, which can drive business decisions. Do you prefer getting an automated risk score on your dashboard, or do you trust your own intuition more when calling a customer?

46

7 Comments

Community bank and credit union executives… If you are serious about running fintech programs the right way, Modules 13 through 24 are now live. We're going deep! Architecture. Multi program, multi rail orchestration. Model risk that examiners actually respect. Capital, liquidity, and real program economics. Horizontal reviews, MRAs, remediation. Evidence automation and continuous exam readiness. War gaming the ugly stuff before it shows up at your door. If you are a Braid customer, it is waiting for you in the Knowledge Center. If you are not a customer, you can access it through participating payments associations and bankers associations. #braided Authored and led by Kevin Johnston He bought a bank, served as COO and board member. Launched Braid. Now teaching others to do the same. 🚀 🤝 #CommunityBanks #CreditUnions #CommunityFI #FinTech #Payments

22

"If we bring in commercial lending infrastructure, will it work with our existing systems?" This is the first question most Community Financial Institutions ask. Smart question. The wrong approach is infrastructure that requires you to: - Replace your core system - Rebuild all your existing processes - Retrain your entire team on completely new workflows - Operate in disconnected silos The right approach is infrastructure that integrates with what you already have. What Integration Actually Means: System Integration Your existing core banking system stays. Your current platforms stay. Your familiar tools stay. Infrastructure should connect TO your systems, not replace them. Data flows where it needs to go. Teams work in familiar environments. Nothing gets ripped out and rebuilt. Workflow Integration Your team shouldn't have to abandon everything they know. Good infrastructure adapts to your institution: - Customized workflows that match your approval process - Reporting that delivers what leadership actually needs - Documentation that aligns with your existing standards Team Integration Your internal team doesn't disappear. They stay involved in: - Relationship management - Final approvals - Client communication - Strategic decisions What changes: They're not buried in manual processing anymore. The Onboarding Reality: Week 1-2: Discovery - Understand your current workflows - Map your approval processes - Identify integration points - Define success metrics Week 3-4: Configuration - Set up institution-specific workflows - Configure reporting - Connect systems - Prepare documentation Week 5-6: Training - Train your team on new workflows - Practice with test scenarios - Answer questions - Build confidence Week 7-8: Launch - Process first loans through new infrastructure - Monitor closely - Adjust as needed - Support your team This isn't theoretical. Institutions with different cores, different sizes, different processes have all integrated infrastructure successfully. Because the right infrastructure is designed to adapt, not to force you into a one-size-fits-all approach. What Your Team Experiences: Before: Manual everything, constant context-switching, overwhelmed During Integration: Training, questions answered, gradual transition After: Clarity, efficiency, confidence The transition isn't about abandoning what works. It's about removing what doesn't. The Real Integration Challenge: The technical integration isn't the hard part. The hard part is mindset. Shifting from "we have to build everything ourselves" to "we can bring in proven infrastructure." Institutions that make this shift successfully don't do it because they've given up control. They do it because they've realized infrastructure isn't the core competency they need to build internally. Your core competency is serving your clients. Infrastructure should enable that, not distract from it.

4

LoanPro has joined Plaid's growing partner ecosystem. Together, we're giving lenders access to real-time cash flow data. These proactive insights into income, savings balances, and debt trends empower credit providers to adapt their servicing and collections strategies with borrowers' evolving financial situations, both preventing delinquencies and improving the customer experience. Read more on how we're driving faster decisions and more inclusive lending👇 : https://lnkd.in/gEX5v5Gd

27

1 Comment

Real-time cash flow data? Already in our toolbox. GDS Link is now part of Plaid's partner ecosystem—bringing direct access to real-time cash flow insights into the decisioning workflows our clients use every day. That means faster application reviews, better visibility into borrower behavior, and smarter approvals. We're making it easier for lenders to say "yes" with confidence. Read more below:

13

Ask Ailo has been working incredibly well for our team, and all our partners subscribed to it. Building on the same architecture and accuracy metrics, we're now working on rolling it out for mortgage lender policies as well. The initial beta phase is looking really promising. We started experimenting with RAG architecture as soon as it became available. While the concept was exciting and it worked well sometimes, there were plenty of challenges that made it tough to rely on. We tried a bunch of different methods and technologies, everything from graph databases to using multiple specialised LLMs and eventually landed on our current setup, which has been working great. You can check out a demo below. If you're interested in joining the waitlist, just send me a DM. It's available as a Slack app, but soon we'll be launching a sleek, full-screen chat UI, along with a whitelabel chatbot UI that can be embedded into other systems. Stay tuned. P.S. The policies used are outdated. This is purely for demo purposes.

38

2 Comments