In this episode of Connect, the California MBA podcast, host Paul Gigliotti sits down with Varant Herculian of JazzX AI to explore how mortgage lenders can move AI adoption beyond the demo stage and into real-world execution.
With 30 years in mortgage lending, technology implementation, and customer success, Varant breaks down what an end-to-end AI platform actually means, how to balance AI-driven automation with human judgment, and what separates lenders who achieve real adoption from those left with underutilized technology investments.
Read the transcript of Stop Whiteboarding and Start Building: How Lenders Actually Adopt AI with Varant Herculian:
PAUL GIGLIOTTI
Good afternoon and welcome to Connect, the podcast of the California MBA, where strategy, innovation, and leadership come together to help shape the future of mortgage finance. This is where the voices and ideas driving our industry forward meet real world execution. My name is Paul Gelotti, CEO of the California MBA, and if you’re here, you care about building better for your business, your clients, and for the future of housing.
Before we begin, I want to take a brief moment to thank our 2026 President Council sponsors whose support make these conversations as well as all of our advocacy and legislative initiatives that we work on possible. So huge shout out and thank you to AmeriHome, Consolidated Analytics, Loan DNA, FundingShield, CMG Finance, Guild Mortgage, Rocket Mortgage, Wiener, Brodzky Kider, and Western Alliance Bank. Thank you again for your participation. I love that it’s a long list and I appreciate all of you. So today I am very excited to bring as our guest a friend of mine Varant, who is now at Jazz X AI. Varant, thank you so much for being here.
VARANT HERCULIAN
Thanks for having me, Paul. Excited for the discussion.
I’ve got a little short bio on you. So I’m going to run through it. Everything about Varant that people in the industry have wanted to know. No, I’m just kidding. You spent 25 years in the mortgage industry. So clearly you started when you were like 10 years old, right?
VARANT HERCULIAN
Something like that. But yeah, I think it’s been 30 years when that cassette over there was still relevant.
PAUL GIGLIOTTI
That’s right. So you started originating loans, helping lenders navigate digital strategy, technology implementation, and operational transformation. So that brings you now to Jazz X AI, which is perfect for everything that you’ve done in your work experience.
Today’s conversation is not going to be simply about technology. We’re going to shift a little bit because of your background, and what I hope to uncover is ideas on how lenders decide to adopt transformational technology, how you move from an exciting demonstration to real world execution, and how you bring the employees and the team along for that journey, because there’s so much movement in our industry with artificial intelligence and technology. I thought that this would be a perfect line of discussion for you. So welcome to Connect, Varant. It’s great to have you here.
VARANT HERCULIAN
Great to be here. Excited for it.
PAUL GIGLIOTTI
Real quick, Varant, how long have you been at Jazz X now?
VARANT HERCULIAN
It’s been about four months now. It’s still pretty fresh.
PAUL GIGLIOTTI
So you know the product, you know the value proposition, you know how it works, and you’re integrating yourself into the customers that are utilizing the solution as well as the team. So you’ve got your thumb on it. You’ve got your finger on the pulse, so to speak, now, right?
VARANT HERCULIAN
Absolutely.
PAUL GIGLIOTTI
All right. So let’s get technical from a front row seat to mortgage transformation. You’ve often sat at the forefront in mortgage technology for, as we were talking about, more than two decades, right? Starting with originating loans, you worked across digital strategy, and now you’re helping lenders adopt AI at a time when the industry is moving incredibly quickly and almost to the point of where it’s stumbling upon itself. What drew you to Jazz X, and why did this feel like the right moment for you to step directly into AI transformation in the mortgage ecosystem?
VARANT HERCULIAN
It’s a great question, Paul. We’ve seen a lot happen and evolve over the last two to three decades, like you mentioned, right? That cassette is no longer as relevant. I used to have a pager when I started in the business. I’m sure everybody had their HP 12C, trusty handy dandy calculator, with them to do everything they had to do. But then comes along the internet and AUS and even the digital point of sale, right? And every one of these had a major impact in our industry, in our business, and the lenders that benefited the most were the ones that were most willing to adopt the new technology and transform their business around it. Clearly AI is that next major inflection point, and the opportunity to help lead that change at Jazz X with its customers is truly an opportunity to dent the universe, in my opinion, and this is exactly where I need to be right now today.
PAUL GIGLIOTTI
And so it was the solve that Jazz is bringing to the marketplace, right?
VARANT HERCULIAN
The solve that Jazz is bringing to the marketplace is absolutely amazing, right? The true end-to-end AI platform solution. But what was even more exciting was the challenge that presented to lenders on how they could even potentially reimagine that future state operating model, right? We just got stuck into analysis paralysis and design mode trying to whiteboard every single permutation on what that’s going to look like, and that’s where a lot of lenders are currently stumbling. So the opportunity to be part of that and help figure out the solution there, and how we can get there smarter, better, faster, is really what’s most exciting for me personally.
PAUL GIGLIOTTI
And that certainly adheres to your experience, right? I mean, you’ve had your finger on the pulse. So you have a clear lens as to, from intake to the secondary market or sale of the loan, right? So I’m really excited to dive into this conversation as we talk more about Jazz, your experience, and your outlook on the mortgage industry. You also brought up something that’s very interesting, a hot topic: adoption. And I think the reason why that’s a hot topic — well, there’s always this fear around adopting technology, because it’s going to take my job away, or I don’t understand it, I don’t know how to use it, the technology is going to be smarter than me. But I also think, in an age where so much information is coming at companies and team members, adopting a solution, or understanding the true purpose of a solution, really helps the adoption exercise, I would think.
So I hope we get into those concepts a little bit more in our conversation. But I wanted to jump through, looking beyond another solution, this topic or conversation around — there are so many tools. The industry is siloed. The mortgage loan life cycle is siloed. And up until the AI revolution, all the tools and technology were siloed. You had intake tools, you had processing tools, you had appraisal tools, you had underwriting tools, you had closing tools, secondary tools, post-closing tools, and it made it, from my perspective, a little bit more clunky. I saw that when I was CEO at Pinnacle Home Loans. Everything was clunky because everything was still siloed and the solutions were almost not talking to each other. From your perspective, what does an end-to-end AI solution actually mean in practice? And how is that different from deploying another point solution into one part of the mortgage process?
VARANT HERCULIAN
You’re absolutely right, Paul. You and I have seen that over the years, the Frankenstein ecosystem where we’re stacking point solution on point solution, and we deploy small armies to put the paper clips and duct tape together to hold it all together for us, right? That’s part of the problem and the challenge that we face here today. But to answer your question, end-to-end AI doesn’t mean one bot making every decision and automatically solving all your problems. It means creating a shared intelligence layer, a knowledge fabric so to speak, across the entire mortgage life cycle, so that the loan officers, processors, underwriters, and their digital workers are all working from the same loan context. A point solution improves one task, and end-to-end AI connects the work, the decisions, the handoffs, across the entire manufacturing chain, while working with — and not replacing — the lender’s existing legacy systems. That’s a big fear right now: do I have to rip and replace all this tooling that I put in and invested a ton of money in over time, or can I stack this intelligence layer on top of it and realize these efficiencies across the entire manufacturing process without that?
PAUL GIGLIOTTI
So from the perspective of the solution you’re working with now, Jazz, it’s an AI solution that overlays your current technology solution. Is that what you’re saying?
VARANT HERCULIAN
Absolutely. Think of it as an intelligence layer that sits on top of the current legacy stack, that serves as your source of knowledge, judgment, and intelligence, whereas your LOS is still the source of truth throughout the entire manufacturing process.
PAUL GIGLIOTTI
Very interesting. So, when we talk about adoption, it must be a little bit easier for the solution, for adoption of your solution, because you’re talking about it almost like turbocharging the already laid-in technology solutions. Right. So adoption should be a little bit easier from that perspective.
VARANT HERCULIAN
That’s exactly right. It’s a support system that’s going to help make your life better, easier, and faster, not just another point solution that you have to figure out how to swivel seat to and wedge into your existing tech stack.
PAUL GIGLIOTTI
That’s an interesting concept. All right, sounds good. So let’s move into an area that I know you’re steeped in. I’ve worked with you, as a lender or a provider, and I’ve been a customer of yours a handful of times now. And an area that I’ve always seen you shine in, as well as others, is customer success. So I want to move into that arena a little bit.
A lender can purchase powerful technology, but that does not automatically mean, as we were talking about adoption, employees will trust it, use it correctly, or allow it to change the way they work. When a lender begins implementing AI, what separates an organization that achieves real adoption from one that ends up with another underutilized technology investment? So from your perspective, Varant, I’m sure you’ve seen Jazz and other solution providers, you’ve seen some of the solutions that you provide being adopted quickly and you see the success, and then you’ve seen other solutions where it’s just not taking lift. What’s different about the organizations with those two outcomes?
VARANT HERCULIAN
Yeah, we’ve seen both flavors of that. And the formula is pretty distinct, right? So real adoption starts with the business outcomes, not the technology itself, right? Most people are driving towards an implementation of the technology versus using the north star metrics of what are we looking to achieve. The lenders that succeed, they choose a focused use case, right?
They start with a small number of loans and they involve their frontline employees early on to improve the solution in an iterative agile process as they learn in production, right? So what I’m saying there is part of my philosophy of an agile production launch, and to try to alleviate that stalling of a technology adoption initiative by trying to whiteboard out every permutation of what could happen in a conference room for five months, when we get that real high-fidelity learning in production on those first five loans much faster.
That’s where we uncover the true gaps and handoffs, the true future state workflow, how those roles are going to evolve, how the work is going to evolve. And that’s when you should finish the last mile of your deployment plan blueprint, of your adoption plan, of your training plan and your communication plan. And that’s where most of us tend to fail, right? That’s where we can really fortify that plan and make it effective and scalable at the end of the day. So people adopt the AI when they trust it. And we need to be able to capture those testimonials, create a champion network, do that in a real production environment, actually using the technology in and of itself for a real message that’s going to land and cascade as we try to scale past a control launch.
PAUL GIGLIOTTI
So I think what you’re saying is it depends on maybe how organized and structured each of those two scenario companies are. Is that kind of how you would wrap that up?
VARANT HERCULIAN
I was more so speaking about the approach and the framework on how to tackle the adoption than the actual outcome in and of itself. But that’s exactly right, it’s how you structure and organize that plan. Are we going to have a bunch of suits in a conference room try to whiteboard it from day one, or are you going to actually roll your sleeves up, have a more agile approach and learn high fidelity in production in real life, right?
PAUL GIGLIOTTI
So, is it kind of like a concept that you’re not looking to achieve 100% perfection in the rollout, but jump in and start to utilize it bit by bit to get buy-in? Is that where you’re coming from in that scenario?
VARANT HERCULIAN
That’s absolutely right.
PAUL GIGLIOTTI
Got it.
VARANT HERCULIAN
I was just going to say that you and I can finish — or anybody that’s been in the industry for several years and is trying to deploy technology — they can put together what they feel that future state looks like, what that operating model is, and that design and that workflow is probably about 80% of it relatively quickly, right? It’s that last mile, when we try to account for every million different permutations, that we get stuck in, that we shouldn’t do until you actually put it into production in a small controlled environment so that you can observe, learn, rinse and repeat, and apply those changes as you incrementally scale based on proof and evidence.
PAUL GIGLIOTTI
Got it. Okay, that makes sense. I have another question based on what our conversation has been around, where we’re talking about implementation, deployment, in an organization, a lender, a solution provider coming in with the solve. Can we also talk about the difference between an organization that — or what is the outcome — either talk about the difference, or what is the outcome, with an organization that is trying to bring in technology that’s got a workflow that’s disjointed, as opposed to an organization that has a workflow that’s tight but there’s just a lot of keystrokes and a lot of room for error, data error, and so deploying the solution on top of those two workflows — what have you seen in regard to a variance of outcome in those two scenarios?
VARANT HERCULIAN
Well, we see a lot of both, and every combination of the two, right? Every lender is a unique snowflake in terms of how tight their workflow is and how well they document it. And that’s kind of the beauty behind what Jazz X does, and what this AI platform can do, right? So you have one thing where you have these business rule models that are laid out, that have a more deterministic approach, that works very well with a tight and clearly defined workflow that might have a lot of steps and keystrokes in it, and automating that is relatively easy. But within that there’s decisions and judgment that need to be had, where you need to leverage the reasoning power of AI to make judgment calls. And that’s typically why people have the looser workflow in the front — to allow their people and their personnel to be a little bit more agile in their processes and use their best judgment accordingly. So the combination of the two, obviously, is the de facto home run, which is what we’re looking to achieve. But what we’re doing here solves a problem for both of those.
PAUL GIGLIOTTI
Okay, got it. So that makes sense. So because Jazz AI lays over your current system, it lays over your current workflow, so perhaps your solution has the ability to also enhance workflow, whereas a lot of solutions need a clear workflow to be able to provide the solve.
VARANT HERCULIAN
That’s exactly right. And that’s a key part of our secret sauce, right? Is to be able to automate those repetitive tasks, apply judgment and reasoning that’s auditable and tunable, tying it back to the source data, and also allowing for a feedback loop, so that you can have those subject matter experts determine whether or not that action, that result, that judgment by the AI was correct and sound, or if they’d like to tune it and adjust it. And then those decisions drop in as knowledge nuggets, get institutionalized in the knowledge fabric, and you can scale across it moving forward from there.
PAUL GIGLIOTTI
Oh, that’s brilliant. So let’s move on and talk about our industry as a whole. It’s not low risk and it’s not a low complexity business, right? We’ve got a ton of risk that you’re responsible for at Jazz, that lenders are responsible for — it’s a very complex business, because first and foremost, the commodity is actually money, and it’s usually an individual’s biggest investment, and there’s also a lot of passion and emotion behind it. So it’s very layered, and we’re making these decisions involving credit, compliance, investor guidelines, borrower data, and again, it’s the largest transaction of a consumer. How should lenders think about the balance between AI-driven execution and human judgment? From your perspective at Jazz, what do you talk about when you’re talking about the technology making all these decisions versus a human eye on it?
VARANT HERCULIAN
That’s an art and a science right there. And we kind of touched on it a little bit earlier, but it’s very easy for the AI and automation to handle more of the preparation, the comparison, and the routine execution, right? While the people remain accountable for the exceptions and the ambiguity and the consequential decisions. But AI has judgment and reasoning capabilities as well, and how much you let it do, how autonomous you let it operate, is going to vary based on risk aversion. And that balance is going to vary by lender, by task, by role, and even by the individual user. This is also going to evolve, I should say, too, as the trust and the evidence grows over time, that comfort level is going to grow over time and your risk aversion will subside. But that’s why the tunable autonomy is so important. Lenders can begin in a recommendation-only mode and gradually increase automation with clear guardrails, keeping everybody safe. And the goal is not to remove the human judgment but to focus it on what matters most, right? So that’s always going to be there.
PAUL GIGLIOTTI
So let’s move to ROI. Every lender — that’s like the second thing that they go to — how can you help bring business in, and what’s my immediate ROI, is now what I’m hearing. There’s absolutely no shortage of AI messaging in the marketplace right now, industry agnostic but inclusive of mortgage lending. Every company is talking about automation, efficiencies, being more productive, and this AI transformation, this AI age. Lenders are operating in an environment where every investment must be justified because of margin depression. Lenders are making less and less in order to just bring in some volume, and for servicing, if they happen to service. So when you sit down with a lender, what measurements should be established at the beginning so everyone can determine whether the implementation is truly creating value? How do you have that conversation with lenders, Varant?
VARANT HERCULIAN
That’s a mission-critical conversation we have to have with every lender. And that’s one of the things I love about Jazz X — the maniacal focus that we have on delivering value to our customers. And it’s not just something we talk about, it’s a key part of our discovery, implementation, and value realization model. And what you have to start with there is clearly defining the problem that you’re trying to solve, building that case for change, going through that exercise and agreeing on the desired outcomes. Then you can establish a baseline measure for throughput, cycle time, quality, the economics, such as loans per FTE, manual underwriter touches, clear to close, cycle times, defects — everything that we constantly measure in our BI tools today, that we all have some form of Tableau, Domo, or dashboard that we’re trying to look at. And training and usage matters, but the real test is where the operation becomes faster, more efficient, and more consistent. So we’ve got to tie back every single outcome, every single result, back to that baseline and show a measurable ROI as we deploy the technology in a controlled environment, in that agile launch, before you can scale. You need that proof point.
PAUL GIGLIOTTI
So that is a tricky conversation for you. I’m sure it’s a leading question from lenders. Do you find that implementation upfront costs — do you find that lenders are skittish around implementation upfront costs?
VARANT HERCULIAN
Not as much as you might think. I think that lenders are actually open to those implementation upfront costs if we can prove out the value with a POC or a POV, right? They’re open. Nobody’s reluctant if you can truly deliver on that value.
PAUL GIGLIOTTI
Oh, that’s awesome. Well, Varant, I want to thank you for your time. I want to thank you for your partnership over the years, and specifically at Jazz. And also thank you for this conversation. I appreciate you joining us, and more importantly, bringing the conversation back to execution. So thank you for being here and sharing your knowledge. From my perspective, AI might create extraordinary new capabilities, but the ultimate measure of success will be whether lenders can integrate those capabilities into the operation function, bringing their people along and producing better outcomes for the business and the borrowers.
VARANT HERCULIAN
It’s awesome. Thank you.
PAUL GIGLIOTTI
Yeah. I mean, we’re in this — we’re in this brilliant age, and I often equated — I heard this from John Chambers, he equated the AI revolution to the internet disruption, and I think that’s where we’re sitting. So it’s really exciting to be a part of it and live through it and see the transformation. So again, thank you for being here. To our audience, thank you for tuning in to Connect. This is exactly why we have these conversations, because the future of our industry will not be built by technology alone. It will be built by leaders who know how to connect, innovate, and strategize. Until next time, stay engaged, stay sharp, and keep building with purpose. Thanks everyone.
