What is CashAI?
Dave was an early adopter of AI, launching its first machine learning underwriting models in 2019. CashAI is a proprietary AI-driven underwriting engine which ingests cash flow data from members’ connected bank accounts alongside proprietary credit performance data to generate a member-specific risk score that determines if and how much credit to extend through bank partners. Dave updates CashAI on an approximately annual basis. The latest version, CashAI v5.5, launched in September 2025 and nearly doubled its feature set compared to prior models.
How does CashAI provide a competitive advantage over traditional underwriting?
Dave underwrites members for credit who connect an external bank account to our platform, providing access to up to 24 months of historical transaction data and near real-time visibility going forward. Dave has originated (either ourselves or through our bank partners) nearly 200 million ExtraCash transactions since inception, creating what we believe to be one of the largest proprietary datasets in the industry that enhances predictive performance. This depth of insight gives us a detailed understanding of where, how, and when our members earn, spend, and manage their cash flow. Every origination adds a new data point connecting a member's cash flow pattern to a credit outcome, which improves CashAI's predictive accuracy and enables us to offer higher limits to more members.
Why is CashAI a durable competitive moat?
CashAI's competitive advantage stems from the nearly 200 million ExtraCash transactions originated since inception, each allowing us to link a member's cash flow profile to a credit outcome–a dataset that takes years of origination history to build. We believe that this creates a flywheel: more data improves model accuracy, which enables higher limits for more members, which generates more transactions and more data.
Beyond the dataset itself, Dave has invested significant time and capital in building the regulatory and operational infrastructure necessary to operate at scale — including bank partnerships, payments infrastructure, compliance capabilities, capital markets relationships, and a large network of customized vendor integrations. Additionally, over 460 million servicing interactions with members have further refined our ability to separate credit risk, providing a behavioral signal layer that cannot be replicated through transaction data alone. This combination of proprietary data, purpose-built infrastructure, and deep servicing intelligence is what we believe makes CashAI’s competitive moat durable and difficult to replicate.
How does Dave think about the risk and opportunity of broader AI disruption?
We believe broader AI advancement is a tailwind, not a threat, for Dave. From a defensibility perspective, our moat is not merely algorithmic. We've invested significant time and capital in building the necessary regulatory and operational infrastructure and relationships across bank partnerships, payments infrastructure, compliance, capital markets and a large network of customized vendor integrations to operate at scale. More importantly, as referenced above, we have established a massive proprietary data set on credit performance and servicing interactions to refine our models, which is difficult to replicate without significant user scale and capital investment to absorb losses. Second, in a scenario in which AI creates dislocation in the economy, leading to lower incomes or higher unemployment and government-assisted income, while origination per user could potentially decrease slightly, we believe this will be more than offset by the large increase in Americans looking for and for whom we can underwrite for short-term liquidity. Overall, we believe our business will continue to benefit from AI innovation. AI technology allows us to make CashAI more powerful, bring to market more valuable products for our members with an efficient team, and support speed and scalability across all aspects of our operations.
What advantages come with ExtraCash’s short duration?
ExtraCash’s short duration (approximately 12 days on average) allows us to operate without a capital-intensive balance sheet or taking significant credit exposure at any one time. Our net ExtraCash receivables balance is approximately one-tenth the size of quarterly originations, underscoring the product’s high-turnover, capital-light nature.
Short duration also confers a structural learning advantage that compounds over time. First, because every ExtraCash cycle takes less than two weeks on average, CashAI continuously ingests new credit outcome data and allows us to refine its underwriting in near real-time. Second, short duration accelerates product and fee structure iteration. Testing requires weeks or months of cohort observation to establish signal; that cycle can take years at traditional lenders originating longer-duration products — a timeline exposed to shifting macroeconomic conditions, undermining the efficacy of the testing process. When we rolled out our new fee structure in early 2025, for example, we were able to assess member retention, reactivation, repayment behavior, and portfolio economics in a matter of weeks for new members and less than three months for existing members. This speed of learning across underwriting, pricing, and product development compounds our advantage over time.
How do we believe Dave’s credit performance will hold up in a challenging macroeconomic environment?
In stressed macro environments, CashAI is designed to allow us to adjust approval rates and approval amounts in near real-time, while maintaining losses within our targeted bands. This speed and precision give us high confidence in our ability to help control credit outcomes and deliver consistent business results across economic cycles. Lastly, ExtraCash focuses on essential use cases such as rent, groceries, and gas, which we believe supports strong repayment behavior and disciplined portfolio performance across economic cycles.
We also believe a stressed macroeconomic environment would expand our addressable market. Economic stress typically increases the number of Americans living paycheck to paycheck, growing the population that benefits most from access to short-term liquidity. At the same time, traditional lenders that rely on FICO-based underwriting models have historically pulled back from lower income demographics during downturns. FICO scores are inherently backward-looking — a measure of how a borrower has behaved over years, not how they are positioned currently — and offer little forward visibility into a consumer's ability to repay when their financial circumstances are changing rapidly. Unable to accurately price risk in a fast-moving environment, these traditional lenders often retreat from the market. The result is a credit supply gap precisely when demand is rising. We believe Dave is built for exactly this environment: our cash flow-based underwriting reflects a member's then-current financial reality, not their repayment history from years past. Moreover, our short duration allows us to adjust pricing to reflect risk in weeks rather than quarters or years. Both forces — rising demand and retreating supply — would drive more consumers toward Dave.
How does Dave measure and evaluate the economics of ExtraCash?
We believe the correct way to evaluate our economics is ExtraCash Net Monetization Rate — which is gross spread less 121-day losses — and net revenue per transaction. We are focused on these metrics to optimize our business, and we will continue to highlight these numbers in our earnings reports.
A good example of this is in early 2025, when we replaced our optional fee model with a simplified 5% fee structure, including a $5 minimum and $15 maximum. The result was greater credit revenue retention as customers retain on our platform, resulting in better portfolio spreads. The larger and more predictable monetization spreads also gave us an opportunity to increase approval limits for new and existing customers, which helps with both conversion and monetization. While these higher limits led to a one-time increase in loss rates, the impact was far outweighed by the gains we achieved in the incremental gross spread. The net result was better monetization per transaction, higher member lifetime value, and more competitive offers for members — a better outcome for both Dave and the consumers we serve.
1 ExtraCash Net Monetization Rate defined and calculated as ExtraCash revenue (i.e. processing and service fees, net) less 121 day losses divided by total ExtraCash originations over a given period.
What is the provision for credit losses?
The provision for credit losses is a non-cash accounting estimate under U.S. GAAP that ensures our allowance for credit losses reflects expected losses on outstanding ExtraCash receivables at period-end. Given the short duration and unique characteristics of our product, we believe a more complete view of credit performance is provided through the static-pool metrics discussed below.
What are the primary drivers of changes in Dave’s provision for credit losses?
Changes in our provision for credit losses are driven primarily by portfolio growth, composition, timing, and underlying credit outcomes. For example, ExtraCash receivables typically peak intra-week on Tuesdays, resulting in a higher allowance, and therefore a higher provision, when a reporting period ends on that day, all else equal. These dynamics can create short-term variability in the provision that may not reflect changes in actual credit performance.
When are ExtraCash receivables written off?
ExtraCash receivables are written off after 120 days from bank origination, or sooner if determined to be uncollectible.
What loss metric should be assessed to evaluate Dave’s credit performance?
For a short-duration, single-repayment product like ExtraCash, we believe static-pool metrics—including the 28-Day past due and 121-Day charge-off rates—provide a more accurate and consistent view of credit health. The 28-Day past due rate is a leading indicator of credit performance while the 121-day charge-off rate is the realized outcome it is designed to predict. These measures eliminate distortions from growth and timing effects and offer a clearer assessment of portfolio quality and credit risk management effectiveness across customer cohorts over time.
