Field Report #002 | The Pioneer Playbook

Two weeks ago, Anthropic and Google walked into ICE Experience 2026 in Las Vegas and sat on a panel about mortgage AI.

Not a tech conference. Not a developer summit. A mortgage conference. The same event where people normally talk about rate locks and compliance updates.

That tells you everything about where this is heading.

ICE Experience is the annual conference hosted by Intercontinental Exchange, the company behind Encompass (the loan origination system most of the industry runs on). It draws thousands of lenders, servicers, and technology vendors.

This year, the headline wasn't a new LOS feature or a servicing update. It was a panel featuring Greg Jacobi from Anthropic (the company behind Claude) and Toby Brown from Google Cloud, sitting alongside ICE's own CTO, talking about agentic AI in mortgage.

Here's the line that stuck with me.

Anthropic's Jacobi said the biggest gains from AI don't come from using it to "do your job." They come from treating it as a thought partner. Arguing with it. Pushing back. Going back and forth until something better comes out the other side.

"You're not just pressing a button," he said. "You're debating with it. You're seeing what comes out and asking for a different perspective."

That's not automation. That's augmentation. And it's exactly how I've been using these tools for the past year.

It wasn't just talk. Multiple companies unveiled AI products at the conference:

Ocrolus launched an automated conditioning engine. Instead of underwriters manually cross-referencing documents against AUS findings, AI generates conditions and matches them to the right paperwork automatically. It went live April 1.

Blend released an autopilot agent for borrower engagement. It reviews documents, checks compliance, and issues follow-ups around the clock. Loan officers see every action in a real-time feed with explanations of what was found and why. The LO still makes every credit decision. The AI handles the paperwork in between.

Optimal Blue introduced a virtual economist. Loan officers can ask questions in plain English about how economic shifts might affect their secondary market pricing. Instead of calling the capital markets desk and hoping someone picks up, they get detailed, data-backed answers instantly.

Teraverde launched Cohi, an agentic AI tool that examines loan profitability by analyzing staff productivity, fallout patterns, and cost structure. It doesn't just report. It predicts where closings are at risk and recommends coaching actions.

Mortgage Automation Tools unveiled Big AI: a conversational mortgage assistant built into the point-of-sale, powered by AWS. Borrowers, loan officers, and executives can all ask questions and get answers in the form of charts, tables, or documents. With a built-in compliance portal that tracks every interaction.

Here's where I break from the conference recap and tell you what I actually think.

Most of these tools are built for large lenders and servicers. They're enterprise products with enterprise pricing. If you're a loan officer at an independent mortgage bank or running a small branch, you won't be logging into Cohi or Blend's autopilot tomorrow morning.

But that's not the point.

The point is directional. The biggest AI companies on the planet are now building specifically for mortgage. Not "financial services broadly." Mortgage. They're sitting on panels at mortgage conferences. They're building products that understand HUD-1 statements and 4506-C forms and automated underwriting findings.

Two years ago, if you wanted AI to help with mortgage work, you had to build it yourself. No templates. No playbooks. Custom workflows and a lot of trial and error.

Now the infrastructure is being built from the top down at the same time practitioners like us are building from the bottom up. I've been mapping this out in a framework I call the Pioneer Framework, which tracks exactly where AI is landing across every department in an independent mortgage bank. Not just origination. Compliance, marketing, ops, secondary, HR, IT.

Those two paths are going to meet. And when they do, the loan officers who already understand how to work with AI will have a massive advantage. Not because they'll adopt the enterprise tools faster. Because they'll already know what to ask for, what to build on, and where the human judgment needs to stay.

There's a harder question underneath all of this. One the conference didn't touch. If AI handles the transactional work that most loan officers spend their day doing, what happens to the loan officers whose entire value was the transactional work? The industry isn't going to lose "loan officers" as a category. It's going to lose a type of loan officer. The value is shifting from process management to advisory judgment. The ones who only knew the process are already feeling it.

I'm writing about that next.

ICE's CTO, Mayur Kapani, said something I keep coming back to:

"Two years from now, we won't be talking about AI. We'll just be using it like a day-to-day tool, and it will be hidden in all the workflows that are part of our platforms. We won't be talking about AI agents. We'll be talking about how I can quickly look up information, get my job done."

That's the endgame. AI disappears into the workflow. And the question for loan officers isn't "should I learn AI?" That window is closing. The question is: when AI is embedded in every tool you touch, will you know how to get the most out of it? Or will you be the person who uses 10% of every platform because you never built the muscle?

After reading through the ICE announcements, I made three changes:

1. I started stress-testing my AI workflows against what's coming. If Blend can autopilot borrower document review, my manual version of that needs to be better or I need to be ready to switch. I'm mapping every AI-assisted task I do against the enterprise tools being built, so I know where my custom workflows will get replaced and where they'll stay ahead.

2. I'm paying closer attention to my LOS. Encompass is going to get AI features baked in. Understanding the platform deeply now means I'll be positioned to use those features on day one instead of spending six months figuring them out.

3. I'm documenting more aggressively. Every workflow. Every prompt. Every process. When the enterprise tools arrive, the practitioners who can articulate exactly what they need will be the ones who shape how those tools work. The ones who can't will get whatever the default settings give them.

Anthropic and Google didn't show up in Vegas to make a sales pitch. They showed up because mortgage is now a priority vertical for the most advanced AI companies in the world.

That should make you pay attention. Not because the robots are coming for your job. Because the infrastructure is being built, right now, that will separate the loan officers who use AI as a competitive advantage from the ones who are still figuring out where the button is.

The tools announced at ICE won't hit most of our desks for 12 to 18 months. That's not a reason to wait. That's a runway.

Use it.

First in. Documenting everything. Building the frameworks for what comes next.

If this landed, share it with a loan officer who's still on the fence about AI. Or reply and tell me what you're seeing on the ground.

- Joshua Rios