A Peak of $4 Billion and a Fall That Says It All
Blend entered the public markets in 2021 as one of the most talked-about names in the U.S. mortgage fintech space. Its automation platform enabled banks and lenders to process loan applications, streamline onboarding friction, and promised to modernize one of the slowest industries in the financial system. The market believed in its potential: the company surged to a valuation nearing $4 billion at its stock market peak.
What followed requires little interpretation. The Federal Reserve aggressively raised interest rates between 2022 and 2023, causing the volume of mortgage originations in the U.S. to collapse. Blend, whose revenue model was directly tied to its banking clients' transaction activity, felt the impact immediately and proportionately. When the market that fuels your revenue engine contracts by half, no software—no matter how efficient—can absorb that blow without marking the income statement.
Now, as reported by Fortune, CEO Nima Ghamsari is redirecting the company towards AI-driven automation as a recovery vector. The thesis is appealing on paper: if AI can reduce the cost of processing each transaction and widen the platform’s reach beyond mortgage credit, Blend could rebuild its revenue base on a structure less vulnerable to interest rate cycles. The problem is that for this thesis to hold, it requires a type of organizational management that rarely accompanies companies in survival mode.
The Unnamed Mistake
Blend built its business model on an implicit assumption: that the mortgage market would remain static long enough to justify near-total dependency on that segment. This isn’t a critique of the founding team’s intelligence; it's a description of how hypergrowth logic functions in an environment of cheap capital. When money is abundant and investors reward growth at any cost, concentrating resources in the segment with the highest traction makes perfect financial sense in the short term.
The structural issue is that this concentration leaves the company without buffers. A well-designed business portfolio does not rely on a single revenue engine to weather adverse cycles. At its peak valuation, Blend did not seem to be incubating parallel business lines with enough budgetary autonomy to become a second engine when the primary one failed. Mortgage automation was everything: the product, the customer, the pitch to investors, and the metrics by which every internal initiative was measured.
This homogeneity in the portfolio is precisely the type of fragility that doesn’t show up in IPO roadshows, but economic cycles ultimately expose. When it appears, the most common corporate response—what we see here—is to point to an emerging technology as the solution. In this context, AI risks becoming a narrative argument before it evolves into a verifiable operational motor.
Automating Isn’t the Same as Exploring
This is where analysis becomes technically complex. Ghamsari is betting on AI to reduce operational costs and enhance margins in the existing business. This is, by definition, an efficiency initiative within the current portfolio, not an exploration of new revenue models. The distinction is important because it has direct consequences for how that bet should be managed, financed, and measured.
If AI is deployed as an efficiency layer atop the existing mortgage business, base risk does not vanish: it merely becomes cheaper to operate. An adverse interest rate cycle will continue to compress transaction volume, and cost reductions from automation merely improve the breakeven point; they do not diversify the income source. For the AI bet to be genuinely transformative in portfolio terms, Blend would need to leverage it to open adjacent categories—other credit products, different types of banking clients, other geographical markets—with a genuine exploration logic, not just an optimization of the present.
The classic organizational danger in this type of pivot is measuring AI initiatives with the same profit metrics expected of mature mortgage businesses. If leadership funds an intelligent automation project but requires it to justify its budget with short-term revenue metrics, what they get is not an internal startup exploring new markets, but a glorified IT initiative with a precarious budget. This type of management does not produce portfolio expansion; it generates reports.
What Blend needs—and what isn’t clear if it is implementing—is a budgetary and operational separation between what keeps today’s business alive and what funds tomorrow’s bets. This involves protecting cash flow from the core operation, defining learning metrics for exploratory projects, and giving them enough autonomy to validate market hypotheses without being evaluated every quarter as though they were already mature businesses.
Organizational Design Determines Whether AI is a Lever or Just a Narrative
The story of companies betting on emerging technologies after a business model crisis has a recognizable pattern. Those that succeed clearly separate the survival budget of the current business from the budget for exploring new lines. Those that fail try to use the new technology to make what exists more efficient, without building the organizational capacity to discover something different.
Blend finds itself at a moment where both pressures coexist: it needs to improve margins in the short term to maintain public market confidence, while simultaneously needing to build a revenue base less exposed to mortgage cycles. These two needs have opposing management logics. The first demands control, efficiency, and predictability. The second requires tolerance for experimentation, long validation cycles, and metrics that aren't tied to quarterly balance sheets.
From the available information, Ghamsari's bet seems to lean towards the first vector: using AI to make what already exists more profitable. This is rationally defensible given the context, but it doesn’t resolve the structural problem that left Blend so exposed when the mortgage market contracted. A portfolio that relies on a single sector to generate revenue, even if that sector operates with greater technological efficiency, remains a single-sector portfolio.
The viability of Blend's reset hinges on whether leadership manages to build, in parallel to current operational efficiency, a capacity for exploration with protected budgets and its own metrics. Without that separation, AI will be a margin improvement, not a portfolio expansion, and the next adverse cycle will find the company in the same structural position that led it here.
