Kolena

The deal might look good, but do the numbers hold up?  That’s where underwriting comes in. From digging through leases and rent rolls to analyzing cash flow, debt, risk, and returns, there’s a lot that happens before a deal gets the green light. We broke down 7 key steps + 8 metrics every real estate team should know, plus how AI is helping take some of the most manual work off their plates. Full breakdown in the comments.

NEW BLOG ALERT 🚨 Introducing the Kolena AI Engineering Blog. At Kolena, we’re deeply committed to the technology behind what we build. We started this blog to share more of that work openly, from the research and testing our engineering team is doing to the technical decisions that shape our product. First up: Benchmarking LLMs on Large Corporate Documents, written by Kolena’s Mark Chen. The article takes a closer look at: • How we benchmark LLMs on large, complex corporate documents • Which models come out on top and where they fall short • Why standard benchmarks don’t tell the whole story for real world document workflows Take a look at what our engineering team has been working on. First article in the comments.

Most AI evaluations fall short for one simple reason. Teams get sold on the demo instead of testing how the workflow actually holds up. You don’t need another AI experiment. You need systems that are auditable, repeatable, and ready for the realities of production. That’s exactly what Kolena’s AI Agents are built for.

Jeremiah Garza set his accuracy threshold at 98% before he chose an AI tool. 98% accuracy is the expectation at Longhorn Investments, so it is what they expected from an agent doing the same step. He validated against fax-backs, skewed scans, and loan files photographed on a phone, not a clean sample. Most AI pilots run that order backwards. Build first, measure later, then argue about whether the number is good enough. The September issue of The Kolena Brief walks through the five decisions that get made before an agent goes live. Where the threshold comes from. Which documents you validate on. Who owns the error afterward. Most teams do not have a Jeremiah in-house. The back half of the issue is about who does this work when nobody internally owns it. Write the five down before your next pilot and put a name beside each one. A list with no names on it is a pilot that has already finished. #AI #CommercialRealEstate #Lending #Insurance

What’s actually happening with AI adoption in CRE? Conner Longoria breaks down what we’ve learned from talking to CRE companies over the last year. One of the craziest stats: only 14% have seen an AI demo using their own documents. That gap between a polished demo and real-world performance is exactly where so many AI pilots start to fall apart. Want to see where your team stacks up? Check out the full report and watch the full video in the comments.

Thank you to Kolena for supporting the 2026 NRTA Conference as a Silver Sponsor! We're proud to partner with organizations that are helping advance innovation and education throughout the commercial real estate industry. https://loom.ly/PB6o2uk

AI is everywhere in commercial real estate. Knowing how to actually put it to work is a different story. That’s why we’re launching Kolena’s AI Adoption for CRE Executives certificate program. Three live, hands-on sessions built for CRE professionals: Session 1: Learn the AI fundamentals and how AI agents work across real estate workflows. Session 2: Identify high-value use cases across your organization, measure potential ROI, and build the business case for AI adoption. Session 3: Put it into practice by creating a production-ready AI agent and building an AI adoption roadmap for your organization. No coding or technical background required. Complete the program and capstone and you’ll earn a verifiable AI Adoption for Executives certificate. Spots are limited. Sign up in the comments.

Things are getting competitive at Kolena. Our first fantasy football league is officially underway, and today is draft day. Being fully remote means we’re taking the draft room to Zoom, getting the whole team together to make our picks live. The strategies are set, the draft boards are ready, and a few people are taking this much more seriously than they’d like to admit. We spend a lot of time building AI Agents to take work off your plate. Unfortunately, setting your fantasy lineup is still on you. May the best team win.

September is taking Kolena on the road. Two conferences. Two cities. Plenty of document talk. - Blueprint | Las Vegas | Sept. 22–24 | Booth #606 - NRTA | Tucson | Sept. 27–30 | Booth #112 Come meet the Kolena team and see what happens when AI takes the manual document work off your plate. Link to book with us in the comments.

The documents a private lender actually receives are fax-backs from title companies, skewed scans, speckled pages, and loan files photographed on a phone. That is the input an AI agent has to read. Jeremiah L. Garza, COO of Longhorn Investments, set a 98% accuracy threshold before AI touched a single loan file. In this clip he explains what happened when his team ran their real documents through the tools they had already adopted. Longhorn originates about $40M a month across 11 states. The accuracy number that governs a decision like that is the one measured on the documents you actually get. New episode of The Pragmatic AI Podcast in the links below.