How to Improve Underwriting Processes

I built an AI agent workflow that automatically underwrites real estate investment pitch decks. Here's how it works, the current pitfalls, and how I'd improve it: ➡️Trigger: PDF file is uploaded to Google Drive folder. ➡️Data Extraction: PDF file is passed to MistralOCR, which is one of the better AI tools for reading pdf files and returning accurate text. ➡️Analyzing: Markdown text is passed from Mistral to an openAI model, along with instructions on: -Who I want the openAI model to be (senior investment analyst) -What input I am giving them (pitch deck, formatted in markup text language) -What the ai's job is (read the text, underwrite the deal, and determine if it passes initial screening). -How I want it to output the data (High level summary in text format) ➡️Outputs: It identified the following takeaways: Key Strengths: (attractive price, well‐documented sources/uses of funds, strong projected returns) Key Risks: The absence of onsite management, aging physical assets with significant vacant units, and the need to overcome sub‑market rents ➡️How I'd improve the workflow: -Create a trigger and filter in Gmail, so this workflow runs automatically when I'm sent deals -Include a formatting step to organize the output from Mistral before sending this to openAI, to improve accuracy and lower usage costs -Plugging in a more powerful model (o3-mini vs 4o-mini), or adding on another AI to check the work of the first AI -Better prompting -Connect a database that the AI can refer to for context, and add future outputs to

Sitting with CTOs from 16 major lenders last week, I asked one question: "How well does your LOS handle complex decisioning?" Average score: Below 7. Not because their systems are broken. But because loan origination systems were never built to be decision engines. Here's what Rafi Goldberg from Sapiens explained on the Power House podcast that changed my perspective: AI decisioning isn't about replacing your underwriters. It's about competing on decisions. Think about what actually differentiates your lending: • That 20-year underwriter who knows when to make exceptions • The processor who catches patterns others miss • The branch manager with instincts you can't explain That institutional knowledge is your competitive advantage. Except it's trapped. The technical challenge isn't automation—it's translation. How do you convert decades of human pattern recognition into decision logic that scales? This is where the architecture matters: Traditional business rules approaches fail over time. They become brittle and inflexible, an albatross of technical debt unable to meet business needs. AI decisioning changes that paradigm. Combining declarative decision models with analytics and AI, your experts’ decision can now be converted to business assets at scale, with no loss in business intent and all the observability and adaptability you’ve come to need and expect. One CTO today said it perfectly: "Our LOS manages transactions. But our decisions happen in Excel sheets and email chains." That's the gap. While everyone races to perfect their point-of-sale experience, the real differentiator is decision velocity and precision. Your best people make hundreds of micro-decisions daily. Each one based on experience you can't hire off the street. When they retire, that knowledge disappears. Unless you capture it now. The mortgage industry keeps focusing on the wrong automation. We digitize applications. We automate verifications. We streamline workflows. But decisions? Those still happen in silos. What if your junior underwriter could access your senior team's pattern recognition? What if every loan officer could tap into your best performer's instincts? That's not replacing human judgment. It's amplifying it. The lenders who win the next decade won't have the slickest UI or the fastest application. They'll be the ones who turned their tribal knowledge into scalable, intelligent decision engines. Every lender in that room today knew their LOS wasn't built for this. The question is: Who's going to fix it first?

I started underwriting deals at 19 with Robert Kiyosaki's online cash flow calculator. Then I moved to a yellow legal pad. Those first deals were small multifamily properties in Madison 20+ years ago. I'd calculate cash flow and principal paydown to see if returns made sense. Simple, but it worked. Fast forward to today, and the process has evolved significantly. When we hired our VP of Acquisitions Evan Dillon a few years ago, he pointed out something that seems obvious in hindsight: Most people underwrite a deal, save it in a folder, and that information basically disappears forever. So we built a database. It's about 30 lines of code. You click a button after underwriting a deal and everything transfers over automatically. Now when we look at a new opportunity, we can instantly compare it against every deal we've analyzed in that market. In doing this, you start to see patterns. When something comes in that stabilizes at a meaningfully higher cap rate than everything else you've looked at, you know it's worth pursuing. It's not that we have some magic system in place — the value is just in that we don't let information disappear. Every deal you underwrite teaches you something about the market. You just need to capture it so the knowledge builds over time instead of evaporating.

Same issues we had in 2021. Gift letter unsigned. CD sent late. Case number pulled wrong. Everyone accepts this as "just mortgage." We shouldn’t. We’ve convinced ourselves that these problems are inevitable. That dealing with fires is just part of the job. But here's what we know: The goal isn't to manage problems better. The goal is to make each problem impossible to happen again. It’s easy to approach these problems through: • More training • Better communication • Stricter oversight All of these are band-aids. You train someone not to forget the gift letter signature. They forget anyway because humans forget when managing 40 loans with 800 moving pieces each. You communicate better about CD timing. Someone still misses it because they're drowning in urgent requests. This is what we tackled. Fix it systematically. Not through training. The goal is to build systems that make the mistake impossible: • Gift letters can't be uploaded without a signature verification step • CDs automatically trigger based on loan timeline, not human memory • Case numbers get validated at the point of entry, not weeks later in underwriting Fix it once. Never see it again. This isn't about being harder on people. This is about being smarter about systems. The truth is we’ve accepted dysfunction as normal for decades. We don’t have to anymore. Processors need to stop having to remember 800 tasks across 40 files. Loan officers need to stop chasing updates on problems that could have been caught weeks earlier. Underwriters need to stop cleaning up messes that were knowable and preventable. Right now, everyone is constantly fighting fires, which means no one has time to build the systems that would prevent the fires in the first place. That's what we're changing. We're not going to accept "that's just how mortgage works." We're going to ask: "How do we make this impossible to happen again?" The answer isn't more training. It's what we're committing to build together: Systems that catch problems before they become emergencies. Workflows that surface issues when they're calm and manageable, not when they're weekend disasters. Processes that work even when humans are tired, distracted, or overwhelmed. Here's what we know: Twenty years from now, if we don't change how we approach this, teams will still be dealing with unsigned gift letters, late CDs, and wrong case numbers. The same problems. Forever. We're going to fix them systematically. Once and for all.

I met with one of our recent customers, a software platform at $10mm+ ARR and offering embedded payments. They are scaling fast and moving to a multi-processor setup (e.g., using both Stripe and Adyen) to handle different regions and risk profiles. Here is the massive risk we identified in their "blind spots" 🫠 The Scenario: A fraudster they previously caught and banned on Processor A (Stripe) re-applies to the platform. Because the platform treats these processors as silos, the system doesn't recognize them and boards them onto Processor B (Adyen). They’d never know it was the same bad actor until the chargebacks hit 😭 Here is the advice I gave them to fix this visibility gap: 1. Unify your data into a single pane of glass. Stop logging into Stripe Dashboard, Adyen, and others separately. You need a system that ingests data from all your processors and normalizes it. You need to search for a merchant and see their history regardless of which rail they were boarded on ✅ 2. Play "connect-the-dots" across processors. You must be able to see hidden relationships across the entire portfolio. If a new applicant shares a bank account, email, or device fingerprint with a "bad" account you shut down on Stripe last year, you need to know before you board them on Adyen 🫡 (We've had many clients get saved by our feature, Account Graph, which does exactly this.) 3. Stop the manual "detective work." Support teams are still manually validating onboards - Googling addresses, checking social, etc. It doesn't matter which processor handles the payment; the underwriting bottleneck is the same 😔 My advice: Automate it 🤖 1. Auto-pull Google Street View + AI to verify if it's a real business 📍 2. Auto-scrape websites and social profiles to verify identity 💻 3. Auto-check adverse media to spot reputational risks 📉 4. Automate the 90%, manually review the 1%. Create global rules that sit above the processor level. If an applicant is linked to a previously banned account on any processor, auto-reject them. Free your team from the noise so they can focus on the complex cases 🔥 Their goal was to replace fragmented tools with one source of truth. That's the right move, and Coris is happy to help 🙌

AI-driven underwriting is reshaping lending economics, and surprisingly few have caught on yet. I've been reflecting on why credit decisioning, especially to SMBs, remains so manually intensive. At first glance, you'd think regulation or tech limitations hold things back, but the core bottleneck is actually human labor. Banks still rely heavily on manual processing—reviewing outdated financial statements and Dun & Bradstreet reports, and depending heavily on human judgment to catch subtle risk signals. This problem feels familiar to one I worked on at Nauto (AI software for driver safety). Our models had to detect every crash perfectly (zero misses). But if we optimized strictly for perfect recall, precision plummeted. We flagged too many false positives, slowing down our human reviewers. So we built a human-in-the-loop system where AI pre-highlighted events, shrinking human review time down to just five minutes. The hybrid AI-with-human-oversight solution was key to managing scale and efficiency without sacrificing accuracy. Banks face the same recall-precision dilemma with underwriting. Traditional financial metrics, which are manually prepared, months old, and often incomplete, mean underwriters either miss important signals or drown in excessive manual reviews. At Slope, our hunch was that raw bank transactions could tell us more than quarterly financial statements ever could. So we built specialized LLMs trained on bank transaction data. With AI, we now construct credit-grade financials that are: ➡️granular (transaction-level) ➡️fresh (refresh daily) ➡️instantly verifiable (cannot be falsified) Then we layered on real-time signals from customer reviews and employee headcount changes that let us detect critical business shifts weeks or months before official reports. Our model dramatically cuts risk and cost. It opens up entirely new lending markets, segments previously labeled "too risky" or "not worth it." And this isn’t theoretical. Our models are assisting banks in underwriting millions of $ to real businesses, today. It reminds me of cloud computing replacing on prem services — a structural economic change, rather than a marginal improvement. If you're exploring similar shifts, reach out — I'd love to compare notes.

🔥 Underwriting is entering a new era and it’s not just automated, it’s becoming truly intelligent. The insurers who embrace this shift aren’t just upgrading technology… They’re elevating trust, accuracy, and the customer experience. Here’s what’s changing and why it matters. 👇 🌫️ The Reality Today Underwriters bring deep expertise but the system around them often slows them down: • Endless PDFs, handwritten forms, medical reports • Manual data entry across disconnected systems • Legacy platforms that don’t integrate • Delays in risk evaluation and pricing This isn’t a people problem. It’s a technology and workflow problem. 🤖 What Intelligent Underwriting Really Means This evolution isn’t about replacing humans. It’s about empowering them. With AI-led workflows, underwriting becomes: ✨ Faster — OCR + NLP extract data instantly from documents ✨ Smarter — ML models highlight risks humans might miss ✨ Consistent — explainable decisions through XAI ✨ More strategic — underwriters focus on complex, high-value cases Technologies like Azure Cognitive Services, Google Document AI, AWS Textract, HuggingFace NLP, Snowflake, and Microsoft Dynamics 365 make this possible. AI handles the repetitive tasks. Humans bring the judgment, empathy, and nuance. 🎯 The Strategy Behind Successful Transformation Insurers who get this right don’t start with tools.They start with a vision:- 1️⃣ Unified data foundation — Snowflake, Databricks, MDM 2️⃣ Intelligent Document Processing (IDP) — UiPath, ABBYY, Hyperscience 3️⃣ Predictive underwriting models — Vertex AI, Azure ML, SageMaker 4️⃣ Explainable AI — Responsible AI frameworks 5️⃣ Human-in-the-loop decisions — smart routing + case escalation 6️⃣ Incremental rollouts — one product line at a time, measurable results This is how insurers modernize underwriting without losing its core principles. 🚀 The Impact Insurers adopting intelligent underwriting are seeing: • Accelerated quote-to-bind cycles • Lower operational and processing costs • Stronger fraud detection • Better segmentation and pricing accuracy • A more satisfied, empowered underwriting team It’s the perfect blend of technology, transparency, and trust. 🌟 The Bigger Picture Intelligent underwriting isn’t a trend ,It’s the foundation of the next decade of insurance. The future belongs to companies that integrate AI + data + human expertise responsibly to deliver faster, fairer, and more personalized coverage. #InsuranceInnovation #AIInInsurance #UnderwritingTransformation #IntelligentUnderwriting #InsurTech #MachineLearning #ArtificialIntelligence #DataScience #Automation #DigitalTransformation #InsuranceTechnology #AIStrategy #IDP #Snowflake #AzureAI #GoogleCloudAI #ResponsibleAI #FutureOfInsurance

Even the best underwriters can't read 50 documents at once, scan the entire open web for risks, and build a financial model simultaneously. But a dedicated team of AI agents can. This is the core of our new agentic approach to underwriting. Instead of a single, rigid model, we give your team a workforce of specialists: • A Document Extraction Agent to turn PDFs into structured data. • A Web Search Agent to act as a diligent researcher. • A Financial Risk Agent to provide an instant summary. The result is turning hours of chaotic, manual work into seconds of structured insight, letting your team focus on the final judgment. In this preview, my co-founder, Maximilian Eber, shows this AI workforce in action. This isn't theory - leading lenders are already processing 100+ applications daily with these agents. He also reveals the game-changer: how you can customize every agent's logic and A/B test your underwriting strategies in minutes, not months. My question for you: What's the one underwriting task you wish you could delegate to a specialist AI agent right now? (See comments for how to get a personalized demo of these agents for your team) #AIAgents #UnderwritingInnovation #AgenticAI #B2BLending

Day 179: 📋 Term Loan Checklist — A Structured Approach for Faster Credit Decisions Most people treat a term loan checklist as routine documentation. But in reality, it’s a critical control point in credit assessment. Here’s the shift 👇 🧩 Generic Approach 👉Standard template 👉No structure 👉Repeated follow-ups 🚀 Structured Approach ✔ Clear segmentation of documents ✔ Logical flow for assessment ✔ Designed for both client clarity & internal efficiency 💡 Why a Term Loan Checklist matters more than you think In credit, delays rarely come from analysis — they come from incomplete or poorly sequenced information. A strong checklist: 🌀Reduces turnaround time (TAT) 🌀Improves data reliability 🌀Minimizes underwriting gaps 🌀Strengthens overall credit quality ⚠️ Hidden Risk Most Overlook Weak checklist → 🔺Inconsistent data collection 🔺Higher dependency on borrower explanations 🔺Increased probability of credit blind spots The checklist is not admin work. It’s input-level risk control. 🛠 Practical Framework 1️⃣ Segment documents (Transaction / Financial / Compliance) 2️⃣ Define purpose for each requirement 3️⃣ Ensure completeness before appraisal 4️⃣ Standardize across cases 5️⃣ Continuously refine based on rejection patterns 📌 Bottom Line A Term Loan Checklist is not just a formality. It’s the foundation of disciplined credit underwriting. Control the inputs → improve the outcomes. #TermLoanChecklist #CreditRisk #SMELending #BankingOperations #LoanProcessing #FinancialAnalysis #CreditDiscipline #ProcessImprovement #BankingInsights #Underwriting #TermLoan #Credit #Appraisal