AI value in fintech is unlikely to accrue to the firms making the loudest claims about models. Over the next 12 to 18 months, the AI winners are more likely to be the companies that control assets competitors cannot easily replicate: proprietary transaction data, fraud and risk infrastructure, and distribution that keeps them close to the customer.

That matters because AI in financial services will not be judged on branding alone. Investors will want proof that it is improving economics in ways that are measurable and durable. In practice, that means looking at whether AI reduces fraud losses, lowers cost to serve, sharpens underwriting and cuts the volume of decisions that still need manual review.

Tomi Popoola is the founder and CEO of Slash Finances, a fintech company using AI to promote financial inclusion. She is also a former AWS Solutions Architect with expertise in AI, cloud infrastructure and financial technology.

In this interview with the Champions Speakers Agency, Popoola sets out which layers of the stack are best placed to capture value. She also explains why fraud and risk infrastructure may monetise fastest, and which metrics investors should watch to separate real operating progress from inflated claims.

1. Which layers of the AI-fintech stack will capture the most strategic leverage over the next 12 to 18 months?

Tomi Popoola: "Strategic leverage will concentrate where AI meets irreplaceable structural advantages, and three layers stand out.

The first is proprietary data. The durable moat in financial AI isn’t the model. It’s the underlying transaction and behavioural data that trains and continuously refines it. Banks, payment networks, and large fintech platforms already hold datasets that compound in value as AI becomes more deeply embedded in operations. AI can commoditise models, but it cannot commoditise the data feeding them.

The second is fraud and risk infrastructure. This layer will monetise fastest because the ROI is immediate and measurable. Even marginal improvements in fraud detection translate directly into loss reduction, while better risk scoring expands approvals without inflating defaults. These tools also slot into existing workflows with relatively low friction, which accelerates adoption.

The third, and most durable, is distribution. The greatest value in financial services has always accrued to whoever owns the customer relationship. Payments apps, neobanks, and embedded finance platforms can deploy AI across lending, insurance, payments, and financial advice, compounding its impact across the entire customer lifecycle."

2. What metrics should investors track to assess whether AI is genuinely moving the needle?

Tomi Popoola: "The most revealing metrics are those where machine intelligence should produce outcomes that humans or rules-based systems simply cannot match at scale.

Fraud loss rate is often the cleanest signal of model effectiveness. Look for sustained reductions without a corresponding spike in false positives that reject legitimate transactions.

Cost-to-serve per account captures the operational impact of automation across support, compliance, and underwriting. If AI is working, this number should fall without a degradation in service quality.

Credit approval rates alongside default metrics are particularly instructive together. Rising approvals paired with stable or declining charge-offs suggest better risk discrimination, not just looser standards.

CAC to LTV ratios can improve meaningfully if AI enables more precise targeting, better product-market fit at the customer level, and higher cross-sell conversion.

Manual review rates are perhaps the most underused metric. The proportion of transactions still requiring human intervention in fraud, compliance, or underwriting is a direct measure of automation effectiveness. Consistent declines here translate almost mechanically into margin expansion."

This exclusive interview with Tomi Popoola was conducted by Tabish Ali of the Motivational Speakers Agency.

Benzinga Disclaimer: This article is from an unpaid external contributor. It does not represent Benzinga’s reporting and has not been edited for content or accuracy.