AI in servicing: Responsible governance in an era of innovation

Artificial intelligence is no longer “coming soon” – it’s actively being introduced into all aspects of the mortgage industry. This represents a powerful opportunity for servicers to reshape operations, streamline workflows and transform how borrowers access information about their loans. When AI is implemented thoughtfully, servicers can reduce costs, resolve borrower inquiries faster, automate back-office tasks and deliver the 24/7 digital experiences today’s technology-centric homeowners expect.

Despite growing enthusiasm among business leaders and the general population, AI adoption in financial services remains in its early stages. Research from the Digital Banking Report found that nearly 90% of banks are either just starting AI initiatives or haven’t begun. The hesitancy is rooted in three concerns: regulatory and compliance constraints, data privacy and security and limited in-house expertise.

Those concerns are only intensifying. Both Freddie Mac and Fannie Mae have released AI governance guidelines requiring sellers and servicers to establish internal AI policies. Fannie Mae’s guidelines, effective August 2026, and Freddie Mac’s guidelines, effective since March 2026, mandate a designated internal overseer and the ability to disclose at any time what AI tools are in use, how they’re being used and what safeguards are in place. Servicers are also on the hook for noncompliant AI use by their vendors.

State legislatures are also stepping in. With federal AI legislation stalled, states have become the primary drivers of AI regulation, and their focus has largely centered on consumer transparency, algorithmic accountability and automated decision making. Colorado illustrates how quickly this landscape is evolving. The state was the first to attempt comprehensive AI consumer protection, originally targeting algorithmic discrimination in high-stakes decisions. But before the law could take effect, Gov. Jared Polis signed SB 189 on May 14, 2026, repealing and replacing it with a fundamentally different framework that shifts focus to automated decision-making technology (ADMT) and consumer disclosure requirements, effective Jan. 1, 2027.

The message from regulators is clear: AI adoption is expected, but it must be governed, documented and defensible.

ICE’s approach to responsible AI is grounded in three core commitments: every solution built must be governable, auditable and explainable. This framework shapes how AI development is being approached across the company and it starts well before anything goes to market.

Here’s how AI is being developed to meet those commitments:

Servicers who wait too long to adopt AI risk falling behind on efficiency, borrower experience and competitiveness. Those who deploy AI irresponsibly risk regulatory exposure, reputational damage and borrower harm. The sweet spot is the “fast follower” position: partnering with proven, responsible technology, learning from early implementations and scaling with confidence.

ICE is investing across servicing, including applying AI capabilities to MSP®, ICE’s loan servicing system. As announced earlier this year, ICE is incorporating AI into borrower-facing spaces like conversational chat and voice agents, as well as call prediction for call center employees. And all of it is built the same way: with compliance and security at the center, not the margins.

To move forward and embrace innovation, servicers shouldn’t be asking whether to adopt AI. Instead, they should question how to adopt it in a way that earns and keeps the trust of borrowers, employees and regulators. That’s the only kind of AI ICE is building.