AI in mortgage is everywhere—but moving from experimentation to real, measurable impact is where most lenders get stuck. In this webinar, we’re joined by Tony Grothouse, CEO of Revolution Mortgage and Corban Wells, VP of Technology Product Management, at PRMI, to share how they’re cutting through the noise and actually deploying AI in production. From reducing cost per loan to improving productivity and borrower experience, this conversation focuses on what’s working today—and what isn’t—based on real-world implementation.
You’ll hear how Tony and Corban are prioritizing use cases, building internal alignment, and navigating challenges like change management, trust, and compliance. The discussion also explores what it takes to move beyond fragmented point solutions toward a more integrated, end-to-end approach—one that drives faster cycle times, clearer ROI, and scalable impact across the mortgage lifecycle.
Read the transcript of AI in Mortgage: From Pilot to Production — How Leading Lenders Are Deploying AI Today:
Siddhartha Agarwal (JazzX AI):
Everyone, really excited to have you here. The level of interest in this webinar has been incredible and I think that’s because this is about moving from AI pilot to production, and hearing real experiences from the field.
We’re going to talk about where AI is actually delivering value in mortgage and where it isn’t. How lenders are prioritizing use cases across customer experience, cost, and growth. What it really takes to drive adoption internally, from change management to org design. And most importantly, the lessons learned from real deployments, including how teams are thinking about ROI.
Quick intro on me – I’m the CEO of JazzX AI. I’ve previously been at Google, Oracle, and Databricks, and got deeply involved in AI there. When I saw the opportunity in mortgage, I got excited about bringing that kind of transformation to this industry.
Very briefly on JazzX: we’re an AI-native mortgage platform that helps lenders move from pilots to real, measurable production outcomes. We sit on top of your existing LOS and work across the full loan file, applying guidelines like Fannie, Freddie, and FHA to identify what’s missing, why it matters, and what needs to happen next with full auditability.
We’ve built purpose-driven assistants across the lifecycle, from intake through underwriting and post-close, and give lenders the flexibility to configure workflows for their business. The goal is simple: instead of stitching together point tools, you can automate and coordinate the full lifecycle so files move continuously, cycle times drop, and ROI shows up quickly.
With that, let’s kick things off. No slides, this is a real conversation.
Tony, let’s start with you, then we’ll go to Corban.
Tony Grothouse (Revolution Mortgage):
Siddhartha, thank you, great to be here. I’ve really enjoyed our conversations and am looking forward to digging into what AI can do for this industry.
I’m Tony Grothouse, CEO of Revolution Mortgage. I founded the company in 2018 and we’re now licensed in 49 states and doing about $3.5 to $4 billion in volume, with plans to grow that significantly.
A big focus for us is making sure our people are spending time on the highest-value activities, what they do best in each role. And I see AI as a key enabler of that.
Excited to be here and share more about how we’re thinking about it.
Corban Wells (PRMI):
Yeah, my name is Corbin Wells, and I’m excited to be here. Siddhartha, thanks for inviting me.
I’ve been in the mortgage industry since 2001. I’ve tried to leave a couple of times, but it keeps pulling me back. I’m not sure if anyone else can relate.
I’ve been with Primary Residential Mortgage for about 12 years. We’re a little over a $3 billion lender and licensed in what we like to say is 49 states. We usually joke that New York is its own country, no offense to anyone there.
Looking forward to sharing how we’re using AI and where we’re headed.
Siddhartha Agarwal (JazzX AI): :
That’s great. Tony, Corban, we’ve learned a lot from both of you.
For everyone on the webinar, we’d love your questions, so please drop them in the chat. Emily, our Head of Marketing, is helping moderate and will make sure we get to as many as we can. If we run out of time, we’ll follow up afterward.
Tony, let’s start with you. From your perspective, where is AI delivering real, measurable value in mortgage today, and where is it still falling short? Any concrete examples would be helpful.
Tony Grothouse (Revolution Mortgage):
Yeah, absolutely. From a value standpoint today, I think the biggest impact of AI for us has been leveling people up. It’s about getting the right information into people’s hands quickly, helping them think better, analyze faster, and make decisions without always relying on someone else.
Instead of picking up the phone to call an underwriter, people can go find answers themselves. That on-demand expectation is real now. Everyone wants information immediately, and AI is enabling that across the organization. It’s helping our teams better understand policy, procedures, and the best next step in real time.
That said, there are still real challenges. The biggest one is cost. How do we actually drive down cost per loan? That’s the real question.
Right now, most solutions are fragmented. You’ve got a tool here and a tool there, but they’re not connected. So while they may help with information or a specific task, it’s not always clear how they improve profitability or the overall customer experience.
What we’re really focused on is how to connect everything. How do you orchestrate intelligence across the full ecosystem so it actually drives outcomes? That’s what’s most exciting to me right now.
We’re starting to see some early wins, especially around information and process support, and we’re about 45 days away from going live after UAT. What we’re seeing so far is really promising.
Corban Wells (PRMI):
Thanks. I’d agree with a lot of what Tony said.
Right now, costs in the industry are still going up. About a year ago, we were talking about roughly $11,000 per loan to originate. Now it’s closer to $12,000. So we’re not seeing costs come down yet.
Part of the issue is that we keep adding point solutions. Every new tool or handoff, unless it’s fully integrated across the lifecycle, tends to add cost rather than reduce it. The promise is always efficiency and speed, but in reality, costs often go up after implementation. And that’s what we’re seeing across the industry.
So that’s both the opportunity and where AI is falling short today.
What’s more promising, and what we’re excited about with JazzX, is a more integrated approach. It’s about using AI for what it does well, and pairing it with traditional automation where that makes more sense.
For example, AI is very good at ingesting large amounts of information and providing useful, context-aware answers. That’s where we’ve seen early success. One of our first use cases was building an internal knowledge base using a RAG approach, where the system can access all of our policies and procedures and provide answers 24/7.
Now, whether someone is in the field or in the office, they can quickly get reliable answers without having to track someone down. That’s been a meaningful step forward.
Siddhartha Agarwal (JazzX AI):
That’s great. I think a few things really stand out from what you both said.
First, how do you actually reduce the cost to originate, which as Corban pointed out, is still going up. Second, how do you improve loan officer productivity so they can handle more volume and deliver a better customer experience. And third, how do you be thoughtful about where AI is used.
AI isn’t cheap, and it’s not the right tool for everything. You need deterministic outputs in a lot of cases, especially in mortgage. Being able to clearly show what conditions were met, why they were met, and where in the document that evidence exists. That’s where a combination of AI and structured workflows really matters.
So let’s shift to prioritization. How did you think about where to apply AI, and what makes a use case worth investing in? If you can, it would be great to ground that in a specific example across the lifecycle, whether that’s loan officer, processing, or underwriting.
Corban, let’s start with you.
Corban Wells (PRMI):
Sure. As we’ve looked at automation over the past several years, the industry has mostly focused on automating the major steps every loan requires—things like income, assets, credit, and AUS review.
We’ve gotten pretty good at that. But then the file gets handed to a processor or underwriter, and the question becomes, “What now?” They still have to review everything, check for gaps, and catch edge cases. That rework ends up eating into a lot of the value automation was supposed to create.
So when we think about prioritizing AI, the biggest opportunity is not just automating parts of the workflow, but directing the workflow.
When a file shows up, it shouldn’t just say, “Here’s what’s been done.” It should say, “Here’s what’s done, and here are the five or ten things you still need to handle.” That clarity is what unlocks the value.
Now the processor or underwriter can focus on what actually requires their judgment, complete their work, and confidently pass the file along.
That’s the shift—moving from partial automation to guided execution. And that’s where AI can make a real difference.
Tony Grothouse (Revolution Mortgage):
Yeah, I couldn’t agree more with Corban.
When we talk about directing the workflow, we’re really talking about intelligence. It’s about leveling up everyone in the organization. Not every underwriter thinks the same or interprets guidelines the same way. There’s a lot of variation, and that creates inconsistency.
What AI can do is create a shared intelligence layer. As you encounter edge cases and make decisions, that knowledge gets captured and applied consistently across the organization. Instead of constantly retraining people, you’re elevating the system itself.
That changes how we work. Instead of asking someone to manage everything, you’re narrowing the focus. Out of 80 things on a file, maybe they only need to focus on 8, because the system is handling the rest with confidence.
That also shifts everything earlier in the process. You’re giving borrowers and teams more certainty upfront, instead of discovering issues late, right before closing.
At the end of the day, this is about quality and confidence. We’re packaging loans into securities, so the question is, how do we improve the quality of what we’re delivering? How do we create more consistency and trust in the outcome?
That’s what this intelligence layer enables. It gives you a way to continuously improve, apply decisions consistently, and raise the bar across the board. That’s what excites me most.
Corban Wells (PRMI):
Tony made a really important point. Training people to operate at a high level is hard, especially when guidelines are constantly changing. Fannie and Freddie are updating rules every month, and pushing that knowledge across the organization is painful, expensive, and often inconsistent.
We run trainings, we communicate updates, but gaps still happen. We’re asking people to function like machines, and that’s just not realistic.
This is where AI can really help. It can ingest those updates continuously and apply them in real time, so the right guidelines are being used on the right loans without relying on memory or manual updates.
But that only works if it’s implemented well. It’s easy to say AI can do this, but actually deploying it in a reliable way is much harder. That’s where the difference is, and there are very few teams that can execute on it at a high level.
Siddhartha Agarwal (JazzX AI):
Love it. There are a couple of things that really stood out from what you both said.
Corban, your point about directing the workflow versus just automating it is key. It ties directly to what Tony said about shifting work earlier in the process. The goal is to build trust with underwriters by showing exactly what’sbeen done, which conditions have been met, and why. If they want to go deeper, they can trace it back to the guideline and the specific data in the file.
And then, just as importantly, they’re told what actually needs their attention. Instead of reviewing everything, they can focus on the few items that require judgment. It may take time to build that trust, but once it’s there, the efficiency gains are significant.
Tony, your point on shifting left is also critical. Today, there’s a lot of duplication between what processors and underwriters do. If we can move more of that work earlier, we improve productivity and can start compressing timelines from 45–60 days down to something much faster. That’s the outcome everyone wants.
And on the regulation side, these guidelines are long and constantly changing. AI is uniquely suited to handle that complexity and apply it consistently.
So with that, let me shift to the next question. As you evaluated AI solutions, what criteria did you use to decide who to partner with? You mentioned ROI and speed to value—how did you think about that?
Tony, let’s start with you.
Tony Grothouse (Revolution Mortgage):
Yeah, thank you, I appreciate that.
For me, a big part of this comes from experience. Over the past 10–12 years, we’ve all heard the same promises from technology—lower costs, more efficiency—and a lot of it hasn’t fully delivered. There are plenty of shiny tools out there, but that’s not what I’m looking for anymore.
I’m not looking for vendors. I’m looking for partners.
I need partners who understand the business, understand where we’re trying to go, and can help orchestrate intelligence across the entire platform—not just solve one piece of the puzzle.
Security is also a big factor. The more tools and vendors you have, the more exposure and risk you introduce. With everything coming out from Fannie and Freddie around compliance and data, you have to be thoughtful about who you’re working with and how these systems are connected.
At the end of the day, it comes down to vision. What are you trying to build? For us, it’s about creating a more intelligent, more autonomous operation—where loan officers can focus on relationships and growth, and AI handles the heavy lifting. That improves the borrower experience and strengthens everything downstream, from fulfillment to secondary markets.
So when I evaluate partners, I start there. Do they align with our vision? Can they actually help us get there? And are they solving something meaningful across the business?
That’s really how I’ve approached it over the past nine months. And when you find a partner that thinks the same way, that’s when it gets exciting.
Corban Wells (PRMI):
Yeah, Tony hit on a couple things I strongly agree with.
First is the shift from vendors to partners. Over the past year, we’ve really been rethinking that. With AI, we now have the ability to build more ourselves than we could in the past. It used to be that if something was a common industry problem, we relied on vendors to solve it and spread the cost across everyone. That’s changing.
Now, there are really two cases where we still lean on partners.
One is when they have proprietary data we don’t have access to, like a pricing engine.
The second is when they bring deep capabilities that we simply can’t replicate ourselves. And that’s an important distinction. A lot of systems in mortgage are broad but shallow. They cover a lot of ground, but they don’t go deep into solving the hardest parts of the workflow.
What we’re looking for are partners who go deep—who’ve invested in real AI, machine learning, and automation to solve meaningful problems in the process.
That’s what stood out to us with JazzX. When we evaluated them, we saw a team with real AI expertise and the ability to build deep, workflow-driven solutions. From there, it became a true partnership, combining their capabilities with our operational experience to build something that works in the real world.
At the end of the day, that’s how we think about ROI. It comes from combining deep technical capability with practical, end-to-end workflow impact, not just solving isolated parts of the process.
Siddhartha Agarwal (JazzX AI):
Really appreciate you both putting a fine point on ROI.
At the end of the day, it comes down to a few things. Cost per loan, cycle time, and throughput. How do we reduce cost, eliminate rework and multiple touches, and increase how many loans each person can handle? That’swhat really matters.
On evaluation criteria, a couple things stood out. First, having the intelligence layer sit above the system of record, instead of being locked inside the LOS. That gives you more flexibility and control, rather than being limited by what the LOS can support.
Second is partnership. Is there alignment in vision, and do you have a partner who is willing to build with you and evolve the solution over time?
And third is governance. Are you getting the right answers, and can you prove it? Because ultimately, you have to stand behind that with auditors.
Quick note for the audience, we’ll leave about 10 to 15 minutes at the end for questions. Please keep them coming in the chat and we’ll make sure to address them.
Let’s shift to implementation. Once you’ve selected a solution, there’s often internal resistance when introducing AI.
What kind of resistance did you run into, and what advice would you give on how to work through it? Either of you.
Corban Wells (PRMI):
Yeah, there’s definitely resistance, and it’s understandable.
There’s a good historical example people often point to with ATMs. When they were introduced, the expectation was that bank teller jobs would disappear. Instead, the number of bank branches grew, and teller roles actually increased slightly. The ATM automated one part of the job, which allowed tellers to focus on higher-value activities.
But AI feels different. It has the potential to automate a much larger portion of a role, and people recognize that. So the concern is real.
We’ve seen that internally. People will say, “Why should I help test this if it might replace my job?” That’s a fair question.
So for us, it comes down to values. Every organization has to decide how they handle this. Some will take a purely economic approach. Others will take a more balanced view.
Our perspective is that we have a responsibility to our people. If automation changes roles, we want to help create a path forward. That could mean reskilling, redeploying, or helping people transition into something new. If you make that commitment clear, people are much more willing to engage and help move things forward.
At the same time, this shift is happening whether we like it or not. Faster cycle times and better experiences are going to become the standard. If you’re not part of that, the entire organization is at risk.
So we’re trying to balance both. Move forward aggressively, but do it in a way that reflects our values and supports our team through the transition.
Tony Grothouse (Revolution Mortgage):
ou’re giving me the chance to play the villain here, Corban.
I respect that approach, and I think there’s a lot of value in it. But for me, this comes down to performance and reality.
Over the past several years, we’ve been exposing our teams to AI and encouraging them to use it to level up. Some have embraced it. Others haven’t. And that’s just the truth. Even within technical teams, not everyone leans in.
But this isn’t optional. This technology is moving fast, and it’s going to change how this industry operates across every role—loan officers, processors, underwriters, closers.
If we’re going to invest in this and fundamentally change how the business runs, we have to be willing to make tough decisions.
At the same time, the upside is real. The people who embrace this will be in a much better position. They’ll be more productive, more valuable, and ultimately have the opportunity to earn more than they ever have before.
We’re building a business to win. That means driving results, serving customers better, and becoming the best at what we do. And I believe AI is a major part of how we get there.
This is coming fast. It’s evolving every day. And the companies and individuals who lean in and adapt are the ones who are going to pull ahead.
So my message to the team is simple: level up. Learn it, use it, and make yourself indispensable.
Siddhartha Agarwal (JazzX AI):
Yeah, it’s interesting hearing you both talk about job security.
It really comes down to a cultural shift. You almost need a mindset of “why not AI?” Let AI handle the manual work, because that’s where it’s strongest. But human judgment, accountability, and owning outcomes—that still sits with people. Underwriters and teams are still in the loop.
That said, beyond job security, there’s another big concern, which is trust and compliance.
How do you know the system is giving the right answer? Can you actually audit it? Can you stand behind it with regulators?
So I’d love to get your perspective on that. Where have you seen pushback around trust, and how have you addressed concerns around auditability and compliance?
Corban, let’s start with you, and then Tony, feel free to add on.
Corban Wells (PRMI):
Yeah, let me build on what Tony said.
We’ve all heard this idea that you’re not going to be replaced by AI, you’re going to be replaced by someone who knows how to use it. I think that’s exactly the point Tony was making. The expectation is that people level up, and we’re taking the same approach.
I also think it’s helpful that you’re hearing two perspectives here. There’s more than one way to approach this, and that’s valuable.
On the trust and compliance side, it really comes down to transparency and experience. Over time, as you see the system perform and you get more reps with it, trust builds.
But the key is that it can’t be a black box. You have to be able to see why a decision was made. What inputs were used, what guidelines were applied, and how the system reached its conclusion. That’s what makes it defensible.
That becomes especially important when you’re dealing with investors. If there’s a question about a loan, you need to be able to clearly explain the reasoning behind the decision.
And ultimately, people still have to own that outcome. The AI can support the decision, but the responsibility sits with the team. It has to be something you can stand behind, not just something the system produced.
Siddhartha Agarwal (JazzX AI):
You have to own the consequences, right?
Corban Wells (PRMI): 42:05
That’s right. We have to own the outcome. That responsibility doesn’t go away.
And regulators are reinforcing that. They want accountability from a person, not just an algorithm.
Tony Grothouse (Revolution Mortgage):
Yeah, I agree. That intelligence layer around decision-making is critical.
You also need multiple ways to validate outcomes. Whether that’s comparing outputs across models or having humans review decisions, you can’t rely on a single path. Human involvement is still essential.
Trust is everything early on. If the system makes obvious mistakes out of the gate, you lose people quickly. So it has to be solid before you go live. It needs to be tested, validated, and backed by clear policies so teams know how it’s being used and where their responsibility sits.
From my perspective, it’s about structure. You have human ownership, supported by AI, with a layer in between that validates decisions. For us, that’s our credit and risk team. They ultimately oversee and stand behind the outputs.
You also need to audit continuously. Look at the data, find anomalies, and pressure test decisions. AI actually allows you to do more of that, not less. You can review a much larger percentage of loans and focus attention where something looks off.
At the end of the day, it’s about putting the right controls in place so people can trust the system and still win.
Siddhartha Agarwal (JazzX AI):
That’s great. Thank you.
What I’m hearing is that there will always be a human in the loop, but AI needs to clearly explain its decisions. Why a condition was passed, why a loan was approved, which guidelines applied, and what data was used. As that transparency improves, confidence grows, and QC gets stronger because you have clear documentation to support every decision.
We’ve had a number of great questions come in, so thank you to everyone on the call. There are quite a few, and some are specific to JazzX. Our Head of Product, Jagjit Singh, is responding to those in the chat.
Before we get to the broader questions, I’d love to close this section with one final topic.
Based on your experience, what are three practical pieces of advice you’d give to lenders trying to adopt AI successfully today?
Tony, Corban, quick thoughts.
Corban Wells (PRMI): 46:57
I’ll jump in.
First, pick the right partner. Tony emphasized that well.
Second, apply AI to the right problems. It’s not the right tool for everything, especially highly deterministic tasks.
Third, measure the results. Focus on use cases where you can clearly track impact.
Those would be my three takeaways.
Tony Grothouse (Revolution Mortgage): 47:27
Yeah, similar themes. I agree with Corban.
First, you need a clear vision. What are you trying to solve?
Second, define the problems and make sure they’re measurable. If you can’t measure it, you won’t know if it’s working.
And third, simplify. I’m a big believer in taking a process from start to finish and making it as close to fully automated as possible. That means reducing complexity and edge cases where you can.
The simpler and more structured the process, the easier it is to drive real ROI.
Siddhartha Agarwal (JazzX AI):
That’s great, really helpful advice.
Let’s move to audience questions. Corban, Tony, I’ll just ask these and either of you can jump in.
First question: can you share an example of a complex process where AI has helped in a meaningful way?
Corban Wells (PRMI):
I’ll share one.
A good example is building new referral relationships. For a loan officer, breaking through to a real estate agent is tough. Emails get ignored, and traditional outreach doesn’t always work.
We’ve used AI to improve that. By giving it context about the person we’re reaching out to, it can generate a highly personalized message that’s much more likely to get opened and responded to.
That’s something that’s traditionally been very manual and hit-or-miss, and AI makes it much more scalable and effective.
Siddhartha Agarwal (JazzX AI):
I think what you’re highlighting, Corban, is how important personalization has become.
With AI, we can move toward true one-to-one engagement. Instead of marketing to a broad audience, you can tailor outreach to each individual. We all get generic emails today, even from companies that know a lot about us. AI changes that.
It allows you to create much more relevant, personalized communication at scale.
Great example.
Siddhartha Agarwal (JazzX AI):
Alright, next question.
You both emphasized measuring results. So the question is, how do you monitor AI to ensure it’s producing accurate outcomes, without increasing costs?
In other words, how do you validate that the right conditions are being evaluated and the outputs are correct, without adding more overhead on top of an already rising cost per loan?
Tony Grothouse (Revolution Mortgage):
Yeah, I think it starts with understanding your own capabilities.
What does your technology stack look like? What do you have internally across engineering, data, and operations? If you do not have enough of that in house, then you need the right partner to help you monitor and manage this well.
For us, we do a lot internally. I am fortunate to have a strong CTO and engineering team, and we build multiple agents that check each other, run different tasks, and create checks and balances in the process.
The way I think about it is this. You are building systems that reflect how your best people think. If you have a strong compliance leader, a strong credit leader, or a strong operations leader, you want to bring that knowledge into the system and point it to the best resources so it keeps getting better.
Then on top of that, you need visibility. You need dashboards, anomaly detection, and human review of what stands out. That is how you monitor quality without just throwing more people at the problem.
It is also a mindset shift. You do not need humans manually reviewing everything the way they always have. AI can validate a much larger share of the work, and then people can focus on the exceptions and the things that need judgment.
That is really the key. Use AI to expand your control and coverage, not to create another layer of manual effort.
Corban Wells (PRMI):
Let me add one thing.
When you’re building trust in AI, you have to validate it. You need to know that it performs as well as, or better than, your current process.
That means running both in parallel. Have a human underwrite or process the loan, and have the AI or automation solution do the same thing. Then compare the results.
That phase is not going to save you money. It’s an investment. But it gives you real data and evidence to make a confident decision.
Once you have that, you can move forward knowing the solution actually works.
Tony Grothouse (Revolution Mortgage):
Can I ask a quick follow-up on that, Corban?
In that comparison phase, would you use AI, or even a different model, to evaluate the results and identify differences?
Corban Wells (PRMI):
Yeah, absolutely.
We do use AI to help with that comparison. It can take both outputs and highlight the differences.
But the goal isn’t to convince the AI. It’s to convince people. So the output has to be clear, reliable, and easy for humans to understand and trust.
Tony Grothouse (Revolution Mortgage):
I agree 100%.
One important thing to watch is how models can pick up patterns over time. If you’ve used tools like ChatGPT or Claude, you’ve probably seen how they start to reflect prior interactions.
In this context, you need to be careful about separation. Make sure your systems are properly isolated so models aren’t carrying over bias or previous context that could influence decisions.
You want the evaluation to be objective. It has to be clean, independent, and unbiased so you can trust the output.
Siddhartha Agarwal (JazzX AI):
Yeah, this is great. What you’re describing is model drift.
In the past, systems were deterministic. You didn’t have to monitor outputs in production the same way. Now, with AI, answers can shift over time, so you need ongoing evaluation.
That means having systems in place that continuously monitor outputs, flag when things start to drift, and trigger the right controls. Just like bugs in traditional systems, you need processes to catch and handle these issues.
Final question, and quick answers if you can. What kind of impact do you expect from AI in terms of cost reduction or headcount efficiency? Even a rough percentage is helpful.
Corban Wells (PRMI):
Yeah, we’re targeting around $6,000 or less in total cost per loan over the next couple of years with AI.
That would come from a meaningful reduction in the headcount needed to process a loan.
Tony Grothouse (Revolution Mortgage):
Yeah, similar numbers on my end.
We’re targeting about a 55 to 60% reduction in cost, and roughly a 70 to 80% reduction in fulfillment staff.
Siddhartha Agarwal (JazzX AI):
That’s significant—around 50% from Corban and even higher from Tony.
First, I just want to thank both of you. You’ve taken the time to share real insights and experiences, and that’s incredibly valuable for everyone on this call.
We did get more questions than we could cover, so we’ll follow up with responses for those who registered.
But overall, this was a great conversation. Really practical, real-world perspective from the field—so thank you both again.
