UpTrajectory Review

Rocket Mortgage's technology strategy deserves attention not because it is novel in the abstract, but because it inverts a failure pattern that has burned billions in small-business and enterprise software alike. The piece argues that most technology projects collapse when engineers build first and ask user questions later—whether that means blockchain platforms searching for problems to solve, or AI tools that miss actual workflow needs. Rocket instead embedded developers with mortgage brokers, letting the people who actually navigate loan applications, compliance checks, and client conversations dictate what the tools should do. This is less 'user feedback' as an afterthought and more structural co-creation, with brokers effectively serving as product managers during build cycles.

For small-business operators, this is not a distant case study in mortgage lending. It is a diagnostic for every software purchase and technology partnership you have likely regretted. How many of us have adopted a CRM, inventory system, or customer-facing app that required our staff to contort around its assumptions? The 95% AI failure rate cited in the piece lands hardest on businesses without IT departments to absorb the friction—when a tool fails, the owner or a key employee becomes the workaround. Rocket's approach suggests a harder but more reliable procurement standard: refuse to buy or build until the vendor can demonstrate that actual frontline workers, not just executives, shaped the feature set. That shifts power in negotiations and protects against shelfware.

What is genuinely new here is the framing of broker-as-architect rather than broker-as-beta-tester. Most 'user-centered design' in financial services means focus groups after the framework is set. Rocket's reported method—developers sitting at brokers' desks, observing the full chaotic context of a workday—resembles ethnographic research more than conventional software development. Where we are skeptical: the piece does not address whether this process scales, or whether it becomes performative once a company grows past its insurgent phase. Large organizations often codify 'voice of customer' programs that produce theater, not change. We would want to see whether Rocket's recent tools, post-IPO and under public market pressure, still carry the same broker-imprint or have reverted to roadmap-by-committee.

The downstream effects matter for the technology vendor ecosystem, not just for Rocket's competitors. If this model gains credibility, it could reshape how SaaS companies sell to small businesses. Vendors might need to prove embedded development partnerships rather than promise future customization. That could raise upfront costs but lower total cost of ownership. Conversely, it could disadvantage smaller software shops that lack Rocket's resources for intensive field research. For mortgage brokers specifically, the power dynamic shifts: they become talent that technology companies compete to access, not just a channel to be disintermediated. That is a meaningful reversal in an industry that has spent a decade predicting broker obsolescence.

What to watch: whether Rocket publishes specific metrics tying this development approach to outcomes—loan velocity, broker retention, error rates—that would let other businesses benchmark. What to do now: audit your next technology decision by tracing who defined the requirements. If the answer is a vendor's product team with no embedded time in your operation, negotiate for it, or build in a termination clause that triggers if user adoption stalls. The real lesson is that 'built with input from users' has become marketing noise; the standard should be 'built with users in the room when architecture was debated.' That distinction separates tools that survive implementation from those that join the 95% failure heap.

“Rocket's broker-designed tools show the fix: users define the build.” — Forbes Business

Takeaway: Demand proof that frontline workers shaped any technology before you buy, not just feedback after build.