Start With the Problem: How Mortgage Brokers Are Reshaping Enterprise AI Adoption

Businessman using digital tablet with virtual property documents and checklist icons. Concept of real estate technology, online mortgage, smart contract, and house investment.

getty

On September 1, three mortgage brokers took the stage at Detroit’s historic Fillmore Detroit and demonstrated working technology prototypes built to their own specifications. Observing from the mezzanine—seated in one of the theater’s original seats, within a building that first opened in 1925 as the State Theatre, weathered Detroit’s long era of decline, and was ultimately restored as the Fillmore—was a reminder that reinvention often emerges from places with deep history.

The theater was designed by C. Howard Crane during Detroit’s grand movie-palace period, when attending the cinema was itself a communal experience. The architecture carried the warmth of the early twentieth century, yet the atmosphere was unmistakably of the present day. Now, nearly a century later, the same hall hosted a different kind of gathering: mortgage brokers demonstrating to a technology company exactly what they believed their industry required next.

The event was “The Big Pitch,” a competition launched by Rocket Pro in June around an unconventional model for enterprise AI innovation. Rather than presenting brokers with pre-built solutions, the company invited them to articulate the problems they faced. Then came the unexpected turn: after months of framing the contest around a single $100,000 grand prize selected by broker vote, Rocket announced it would develop all three finalist tools, awarding $100,000 to Seth Hasan of West Capital Lending for Quick Counter, $60,000 to Andrew Haff of Barren Hill Mortgage for Right Track, and $40,000 to George Chevalier of Clearview Lending Solutions for Signal.

The 95% Problem

This announcement carries significance well beyond mortgages. It arrives at the heart of a defining tension in enterprise AI: despite enormous investment, technological breakthroughs have not consistently yielded measurable business outcomes.

A widely referenced report from MIT’s NANDA initiative, The GenAI Divide, found that approximately 95% of the enterprise AI projects it examined produced no discernible impact on profit and loss. The study analyzed 300 AI deployments, interviewed senior executives, and surveyed frontline employees. Its conclusion pointed less to the quality of underlying models than to how companies choose the problems they aim to solve and how they embed AI into real operational workflows.

Lead author Aditya Challapally told Fortune that the small number of successful initiatives succeed because they “pick one pain point, execute well, and partner smartly” with the people who actually use the tools. This pattern long predates generative AI.

A Strategy That Predates Artificial Intelligence

In 2018, Maersk and IBM launched TradeLens, a blockchain platform designed to digitize global shipping and trade. The solution was engineered first and then presented to an industry expected to adopt it. Maersk discontinued the project in 2022, acknowledging that while the platform functioned, it had not achieved the broad industry collaboration necessary for commercial sustainability.

Walmart Canada took a fundamentally different approach, beginning with a concrete operational challenge: tracking food products through a fragmented supply chain. In an early pilot with IBM, the time needed to trace an item from store back to farm dropped from seven days to 2.2 seconds. Walmart subsequently mandated that leafy-green suppliers adopt blockchain-based farm-to-table traceability, directly tying the technology to faster food-safety investigations and more effective recalls.

The parallel is far from theoretical. FDA investigations this year have repeatedly depended on traceback methods to identify probable contamination sources, even as the agency works with industry stakeholders to accelerate and refine food traceability capabilities. The technology category is the same. The era is the same. The starting point differs entirely. And the critical distinction was not what the technology could accomplish, but what someone needed it to accomplish. Fast-forward to September 1st: that is the premise behind “The Big Pitch.”

Problems First, Prototypes Second

Rocket Pro received more than 350 submissions, open to any licensed wholesale professional—not solely Rocket partners—and asked participants to submit problems rather than completed solutions. According to Rocket Pro Chief Revenue Officer Austin Niemiec, AI quickly emerged as the predominant theme across submissions. The three finalists then collaborated with Rocket’s engineering team to develop prototypes of their concepts. Voting was conducted through a combination of online and live ballots, with one vote per licensed professional.

Each tool targets a specific friction its creator encounters in daily practice:

Chevalier’s Signal addresses a pipeline bottleneck: existing systems can flag outstanding loan conditions but fail to guide a loan officer on the next step. His concept employs AI to surface issues proactively, recommend targeted solutions, and draft the communications needed with clients and agents—effectively serving as a digital specialist working alongside less experienced loan officers.

Haff’s Right Track confronts the research demands of complex loan files, where brokers can spend hours poring over intricate product guidelines. His tool is designed to chart a viable path forward for the borrower rather than merely cataloging additional obstacles.

Hasan’s Quick Counter, the winning entry, targets speed in competitive scenarios. A broker can import a competing loan estimate and generate a counteroffer within minutes—while the client is still on the phone.

Augmentation, Not Replacement

Across all three submissions, the underlying principle is the same: augmentation—how AI-enhanced tools strengthen the professional already responsible for the client relationship.

This philosophy stands in contrast to years of predictions that AI, much like blockchain before it, would fundamentally disrupt and disintermediate trust-based professions. It also reflects a more pragmatic enterprise application: placing AI directly into the hands of practitioners who possess deep knowledge of the workflow, the customer, and the problem at hand.

Rocket encapsulated this philosophy in the partner commitments unveiled at the event: “human expertise, with the speed of AI.”

Where Web3 Converges With Established Industries

The same reasoning applies well beyond AI.

For years, Web3 has been discussed primarily as a technology category—blockchain, digital assets, smart contracts, and decentralized networks. The conversation typically began with what the technology could do and worked backward toward potential use cases.

Established industries offer a fundamentally different starting point. Their customers buy homes, purchase food, transfer money, and make decisions within systems that have been built and refined over decades. These industries also carry layers of accumulated friction that technology companies may not always perceive from the outside.

Consider property records. Georgia’s National Agency of Public Registry now offers online registration of immovable-property rights through a smart-contract service. The significance extends well beyond maintaining a government record digitally. The service is purpose-built to transition a transaction historically dependent on paper records and intermediaries into a remote, digitally executed process.

Or consider diamonds. De Beers’ Tracr platform uses blockchain to trace natural diamonds from their source through the entire value chain. The company has extended that infrastructure into a consumer-facing proposition through its ORIGIN program, providing provenance details designed to give buyers greater confidence about a diamond’s origin and its journey through the supply chain.

The technology is not the product. Trust is.

Financial services are following the same trajectory. J.P. Morgan’s Kinexys platform leverages blockchain infrastructure for payments and digital assets; the bank reports that the platform has processed over $1.5 trillion in notional value and more than $2 billion in average daily transaction volume. The SEC now formally recognizes tokenized securities as instruments whose ownership records are maintained in whole or in part on crypto networks.

The point is not that blockchain replaces banking. It is that banking can harness blockchain to transform how money and assets move.

This is a consequential distinction.

A faster mortgage process is useful. A mortgage process that provides borrowers with greater visibility into the status of their deal, identifies problems earlier, and equips loan officers with better tools to resolve them is more effective. A blockchain-based food ledger may improve supply-chain efficiency. The ability to trace a potentially contaminated product in seconds rather than days can alter public-health outcomes. A digital property record may reduce administrative burden. A remote transaction can expand who is able to participate in a market and how effortlessly. This is why established industries matter to the next phase of Web3.

The mortgage broker knows where a loan breaks down. The title professional knows where property records break down. The retailer knows where its supply chain breaks down. The banker knows where capital becomes trapped in infrastructure built for another era. These professionals do not need to begin with AI, blockchain, or digital assets. They can begin with the problem.

The Next Problem Worth Solving

And the problems identified by these practitioners may prove increasingly significant. Consider deed fraud, which sits at the intersection of real estate, financial services, government records, and emerging technology. Forged documents can enter public records because the recording process is administrative rather than investigative. Generative AI adds a further dimension, enabling more sophisticated forgeries even as governments explore AI-driven detection methods.

Nearly a decade ago, the Cook County Recorder of Deeds demonstrated that property transfers could be recorded on a blockchain. The technology proved viable. Broader legal and institutional adoption did not follow. Perhaps the broader lesson from “The Big Pitch” is this: the most consequential technology adoption in established industries may not begin with a technology roadmap. It may begin with the person closest to the customer identifying what is not working.

AI, blockchain, tokenization, and digital assets become meaningful when they change what the customer or consumer can accomplish—not merely how efficiently a company performs the same familiar tasks. For established industries, the opportunity lies not in bolting tomorrow’s technologies onto yesterday’s processes, but in using them to deliver better outcomes for the people those industries ultimately serve. Start with the problem. Place the practitioner nearest to it in the room.

Then determine what technology makes possible.

Source link