TRUE

Let's talk about how the conversations around AI has shifted...a year ago, the focus was almost entirely on capability: can you classify documents, extract data, calculate income? Now the conversation is shifting toward something much harder: Why hasn’t it meaningfully changed the operation yet? At The Gathering, I sat down with Deedra and Caleb from Fairway Independent Mortgage Corporation to talk about what it actually looks like to implement AI at an enterprise level. One point they made really stood out to me: Mortgage operations are still fundamentally built around repetitive manual validation across every touchpoint in the file. If that operating model doesn’t change, adding AI simply layers new technology onto the same workflow. That’s why I appreciated Fairway’s approach so much. They’re not treating AI as a headcount reduction exercise. They’re treating it as a way to scale intelligently, reduce friction, and handle more volume with the same core teams. That’s a very different mindset than most of the market. We covered: -why most AI projects stall after the pilot phase -why clean data at intake matters more than downstream automation -what “all in” on enterprise AI actually means operationally -and why adoption and workflow redesign matter just as much as the technology itself Full session here: https://lnkd.in/eUrTrxsj Again thanks for a great event HousingWire Zebulon Lowe Brena Nath Makenna Clay Sarah Wheeler Diego Sanchez Clayton Collins TRUE

Self-employed income is not hard because of math. It is hard because the system does not reflect reality clearly. You’re dealing with multiple documents that: – don’t follow consistent structure – represent income differently – sometimes contradict each other From a systems perspective, the challenge isn’t extracting values. It’s building a coherent representation of the borrower’s financial state. Most workflows don’t do that. They extract from each document independently, then rely on a human to reconcile the differences. That’s where the time goes. With newer architectures, especially when you combine structured models with LLM reasoning, you can start to handle this differently. Instead of: extract → hand off → reconcile You can: extract → compare across sources → reason about differences → update state That reduces how often a human needs to step in. Not because the system is perfect, but because it’s maintaining consistency. That’s a fundamentally different problem than extraction. Where does reconciliation happen in your current workflow?

The best sessions aren’t panels. They’re conversations. That’s exactly what we’re doing next week in Austin. After our final prep with Fairway Independent Mortgage Corporation , one thing stood out: This isn’t going to be a “talk about AI” session. It’s going to be a real conversation about how mortgage operations actually run today and why cost per loan hasn’t moved the way it should. We kept coming back to a few themes: – Most teams are still spending their time checking, not deciding – “Automation” often just shifts work instead of removing it – And the real opportunity isn’t speeding up underwriting… it’s starting with cleaner, decision-ready files That’s where the conversation gets interesting. Not features. Not hype. How lenders actually change the way work gets done. Really looking forward to unpacking this live with Deedra Massey Bosworth and Caleb Ondrusek, on the main stage: Wednesday, April 29 at 9:45 AM Also excited to meet up with some of the people pushing this industry forward like Zebulon Lowe , Clayton Collins , Diego Sanchez , Brooke Anderson Tompkins , Greg Sher , Ryan Edmonds , Michael Hammond JD, Michael Kelleher … should be a great few days. And we’ll have the TRUE team on site as well — Bill Hajjar, Alex Levi , and Liz Long Adkins — so come find us. If you’re going to be at The Gathering and want to connect, drop a comment or shoot me a message. See you in Austin.