The observation that defines our view of 2025 is one about deployment rather than development. The AI-native financial infrastructure companies we have been tracking since 2021 are no longer in experimental deployment. They are in production systems at regulated financial institutions, processing real credit decisions, real fraud assessments, and real compliance workflows at scale. The question that defined 2023 — whether AI-native underwriting and AI-native compliance infrastructure could meet the regulatory and operational standards that financial institutions require — has been answered affirmatively by a sufficient number of successful deployments that the early majority of potential enterprise customers is now seriously evaluating rather than cautiously exploring. That shift in customer posture has direct consequences for the funding environment, the competitive landscape, and the infrastructure categories that will see the most commercial activity in 2025.
The credit infrastructure theme is where we see the most immediate commercial acceleration. The combination of fully reliable Open Banking data connectivity, mature ML model governance tooling, and enterprise customers who have now seen AI underwriting perform across a credit cycle rather than only in a benign environment has created the conditions for mainstream adoption. The institutions that moved early — smaller specialist lenders and challenger banks who were willing to take the product risk of deploying AI underwriting before the track record was deep — have built two to three years of performance data that they are sharing with their networks. The mainstream institutions evaluating AI underwriting in 2025 are not doing so speculatively; they are evaluating it against a body of performance data from real deployments. The sales cycle is still long, but the qualification conversation is shorter.
The regulatory development most material to our portfolio in 2025 is the continued evolution of the FCA's Consumer Duty implementation. The initial implementation deadline passed in July 2023, but the FCA has signalled clearly — through supervisory letters, through published outcomes testing guidance, and through enforcement action in adjacent areas — that it expects to see firms demonstrate good customer outcomes through quantitative evidence, not just governance documentation. The firms that have invested in compliance infrastructure to automate outcomes monitoring are finding that the investment is now essential for maintaining their regulatory standing rather than merely advantageous for their supervisory relationship. For the infrastructure providers serving them — including Reflow in our portfolio — this represents a broadening market rather than a narrowing one.
The theme we are watching most carefully in 2025, with both conviction and caution, is the convergence of AI decisioning and embedded lending infrastructure. Several of our portfolio companies are positioned at the intersection of these two areas, and the commercial logic is compelling: an embedded lending platform that combines proprietary credit decisioning with distribution through software platforms has both the data advantage and the distribution reach that makes a truly durable business. The caution is about sequencing. The companies that have tried to build both the decisioning model and the distribution platform simultaneously have consistently found that the complexity is greater than anticipated and that maintaining quality in both dimensions under resource constraint is difficult. Our guidance to founders in this space is to own one layer deeply before attempting to expand into the adjacent layer — and to be honest about which layer is truly proprietary and which they are borrowing from a partner.
Looking at the funding environment, 2025 appears to be a year when the infrastructure fintech category separates clearly into two cohorts. The companies with genuine enterprise traction — measurable net revenue retention, visible expansion revenue, reference customers willing to speak publicly — are finding investor appetite at growth stages that is materially stronger than the previous two years. The companies that have built against the infrastructure thesis but without yet establishing that depth of enterprise relationship are facing a longer path to growth capital than they expected when they raised their seed rounds in 2021 and 2022. This is uncomfortable for some of the teams involved, but it is the correct market outcome: enterprise infrastructure that enterprises genuinely depend on should attract growth capital, and infrastructure that enterprises are still evaluating should spend more time building enterprise depth before seeking to scale. We remain committed to the thesis and more selective than ever about the specific points in the stack where we back new teams.