Roberta Pek in Rick’s Room

In this episode of Rick’s Room, Rick Grant speaks with Roberta Pek, an experienced mortgage banking, compliance, operations, and process-improvement professional whose career has included major financial institutions. Her background in Lean Six Sigma and regulated financial services has led her naturally into one of the industry’s most pressing questions: How can mortgage lenders take advantage of artificial intelligence without creating unacceptable operational, compliance, and consumer risks?

Too often, governance becomes something organizations add after an AI system has already been deployed. Instead, lenders need to understand where AI operates within their organization, what processes and customers it touches, who owns it, how decisions are being made, and what risks those decisions create.

Peek has developed practical tools using a Six Sigma framework that lenders can apply to their own AI operations. The objective is to help organizations identify where AI models reside, understand how they affect borrowers, quantify their risk severity, determine appropriate monitoring, and establish appropriate human oversight.

The answer isn’t to slow down AI adoption. It’s to make governance part of AI adoption from the beginning.

Transcript Highlights

Rick Grant: Hi, and welcome back to Rick’s Room. This is the podcast where I pretend I’m at the end of a busy day at a business conference. I’m in the lobby, and I finally get to meet somebody I’ve been wanting to talk to but haven’t had the opportunity to catch up with. I didn’t want to wait for the next business conference, so we’ll do it here on the podcast.

My guest today is definitely someone I would wait around to catch up with. Roberta Pek has been around our industry for a long time. She’s worked at some of the biggest and most successful companies in our space, and she’s an expert in compliance. More recently, the new AI tools have caught her attention, so I wanted to get her on the show to find out what she’s doing now.

Roberta, thank you for joining me today.

Roberta Pek: Thank you for having me.

Rick: Before we get into the nuts and bolts of AI, which I’m very interested in, I want to ask about your background. Anybody who looks you up on LinkedIn is going to see all these credentials, including a Master Black Belt in Six Sigma. You could have worked in virtually any industry, and yet you’ve spent a long time with us in financial services and, more specifically, mortgage. What drew you to this industry, and why did you stay?

Roberta: I enjoy financial services. I enjoy improving processes, reducing waste in processes, and making things more efficient.

Mortgage banking is probably among the most regulated areas of financial services. There are hundreds of touchpoints from end to end, and every one of those touchpoints has a chance to go awry.

I took my operations experience and my risk and compliance experience in mortgage banking and coupled them with my Lean Six Sigma Master Black Belt training. I was able to make a difference not only for the companies I worked for but also for the borrower.

Let’s face it: getting a mortgage, or any type of loan, is a huge financial decision for a borrower. We want to make sure the borrower gets the best experience possible—what we like to call frictionless—but we also need to do it compliantly and with a good workflow.

Lean Six Sigma, all the tools of mortgage banking, and now AI all come together. I really enjoy it.

Rick: It’s almost as though all of your experience has prepared you for this moment.

I haven’t talked to anybody in our industry who isn’t excited about applying AI. A lot of people are writing about it. Most of those stories are about capabilities: What can it do? What can it replace? How can it save money? What are organizations missing?

Roberta: From what I’ve seen, the biggest thing organizations are missing is governance.

They have the conversation about capability. Can you underwrite something faster? Can it become more efficient? Can you flag documents?

But they aren’t talking enough about governance.

Governance seems to become an add-on or bolt-on after the AI tool has already been deployed.

If you’re using AI in your shop, you need to understand what it touches, what it does, who owns it, and how decisions are being made.

I’m seeing three major things.

First, people sometimes look at AI as something that will answer all questions. But if you have an AI model making a credit decision, that credit decision comes under the same kind of regulatory scrutiny as if a person were making that decision.

If AI is making a credit decision or another decision that affects the borrower, you need to understand how AI is affecting that decision and what oversight exists.

The second thing I’m seeing is the idea of QC as a bolt-on.

QC has been around for as long as mortgage banking and financial services, but doing it with AI is a totally different animal.

You can’t simply use traditional QC, where you’re taking a small sample of loans and reviewing them after the fact.

AI is making decisions at scale and in real time. It could be making decisions based on patterns that you’re not going to find by reviewing a small random sample.

Shops need to adjust their QC procedures and policies to the reality of AI operating at scale.

The third thing is the idea of the human in the loop.

Many shops feel the AI can make the decision because it’s a model and it’s going to be right.

But you really need to understand where the risk levels are.

Does a human need to review every decision AI makes, or does a particular system require a lighter oversight touch?

It depends on the risk rankings you assign to your AI models. That’s the foundation of everything.

You need to know where your AI models are, what they touch, and then risk-rank them.

Those are three things I think need to be addressed as we move forward with AI in financial services, especially mortgage.

Rick: One of the things I noticed looking at your LinkedIn profile is that you’ve actually created some models lenders can use to start evaluating this.

Roberta: At previous organizations, I created models and prompts to support this work.

For an upcoming Mortgage Bankers Association presentation, I’ve also created four tools using the Six Sigma framework that lenders can take back to their shops and apply.

Rick: What can you share about what you’ll be teaching?

Roberta: The webinar is about AI governance, risk, and quality in mortgage banking operations.

What people are going to get are very practical and tactical ideas about how to implement this in their shops.

There has already been a lot of industry content about AI and AI frameworks. What I’m providing complements that.

Where are your AI models living?

How do they touch the customer or borrower?

What is the risk severity?

I provide ideas for quantifying that risk so participants can take those concepts back to their shops and look at them through their particular lens.

I also have worksheets to help organizations understand how to monitor these systems.

As I mentioned, QC needs to change when we’re looking at AI models.

For example, can you monitor things like time at task to see whether there are indicators of a problem?

Suppose you have an AI model helping underwriting flag exceptions, and you notice that none of those exceptions are being overridden.

That could mean the underwriter is simply accepting what the AI model says rather than doing an independent human-in-the-loop review.

Those are the kinds of practical and tactical ideas that organizations can deploy.

Mortgage banking and financial services are data-rich environments. There’s probably a lot of data, and a lot of it is noise.

You have to tease the signal out of the noise and use it to make informed decisions.

Another issue is determining whether the human in the loop needs to look at every decision or whether a lighter touch is appropriate.

That gets back to the risk associated with the AI model in your mortgage process.

Is it simply flagging documents? Or is it actually making a credit decision that’s going to affect the borrower or affect pricing?

The framework helps organizations think through those questions.

Rick: That’s a lot of value to pack into an hour.

There are probably also people watching who aren’t members of the Mortgage Bankers Association and will want to learn more from you. How can they do that?

Roberta: They can contact me through my LinkedIn profile. I’d be more than happy to have a conversation with anyone.

I also have the four tools—which are basically worksheets, let’s call them what they are—with information about how to do this. I’m happy to have that conversation.

Rick: They’re a little more than worksheets, Roberta. You’re helping people understand how to think about something that’s changing so rapidly that it’s frightening to many organizations. But the big lenders—and anybody who wants to be successful—know they’re going to need AI. They just need to do it right.

Roberta: I agree. And the takeaway I would give them is this: Do it with one small model. One small piece. Get used to doing it with one model and take it from end to end. Choose one piece of mortgage origination, underwriting, closing, or whatever part of the process you want to use, and take it through the entire process. That will give you confidence to do something more complex.

When people go through their shops, inventory where AI touches the mortgage manufacturing process, and risk-rank those applications, I recommend starting with something lower-risk. Learn what the process looks like. Then you’ll have the confidence to tackle the next level of severity.

Rick: There’s so much more I want to ask you, but we’re bumping up against time. I’ll ask you to promise to come back on the show so we can explore more of this fast-changing and vitally important part of our business.

I’ve been visiting with Roberta Pek, an expert and a voice of reason in the new world of AI.

Roberta, thank you for being with me today.

Roberta: Thank you for having me. I’d love to come back.