Banks struggle to turn AI spending into real revenue gains

Dyna.Ai has published new research that argues only a small share of banks are translating artificial intelligence investment into measurable revenue, with early results concentrated in a limited set of use cases.

The executive insights report, produced with GXS Partners and Smartkarma, focuses on banks in Southeast Asia, Latin America and the Middle East. It sets out a view that progress depends less on the number of AI pilots and more on scaling deployments into day-to-day operations with clear accountability for outcomes.

Dyna.Ai points to a sharp rise in expected spending. It says global BFSI AI spend is projected to increase more than tenfold from USD $35 billion in 2023 to USD $368 billion by 2032. It characterises the next phase of adoption as a shift from experimentation to production deployment tied to measurable commercial results.

The report cites a wide gap between perceived progress and measurable returns. It says 77% of financial services executives report positive ROI within the first year. It also says enterprise-wide impact remains limited in many organisations, with unclear ownership for operational outcomes.

"Most banks believe they are progressing with AI, yet research shows only 10% of the organizations using agentic AI are seeing significant, measurable ROI," said Tomas Skoumal, Chairman and Co-founder of Dyna.Ai.

Revenue levers

The research highlights a narrow group of revenue-related use cases that banks have started to scale. It describes AI-driven personalisation as one of the clearest examples. It says banks that operationalise personalisation can achieve up to a 6% revenue uplift.

It also points to wealth management tools used by relationship managers and advisers. It describes "RM co-pilots" as one area showing measurable effects. It says these tools correlate with a 20% year-on-year boost in adviser sales in wealth management, according to the research summary.

The report also flags "next-best-action" approaches for cross-sell as a current focus. It frames these as decisioning systems that recommend products or actions across digital and mobile channels.

Southeast Asia

The research places significant emphasis on Southeast Asia, which it describes as a region where banks are moving from pilots into scaled deployments. It links that shift to mobile-first customer behaviour and regulatory conditions that encourage digital financial services.

As an example, it cites DBS Singapore. Dyna.Ai says DBS Singapore generated USD $565 million in AI-driven revenue from 350 use cases in 2024. It says the bank is targeting USD $745 million by 2025.

The report also connects AI-driven revenue strategies with credit expansion for smaller firms. It references a USD $300 billion ASEAN MSME financing gap. It frames this as a pool of unmet demand where banks are using digital channels and data-driven decisioning to increase lending volumes while managing risk.

Middle East

For the Middle East, the report points to government-led AI strategies and the growth of fintech activity. It cites PwC estimates that AI could add USD $320 billion to the Middle East economy by 2030. It positions financial services as a central sector in that projection.

It identifies wealth management and cross-border payments as early areas of impact. It describes deployments that aim to scale relationship management and strengthen compliance processes. It also links AI adoption with faster and more reliable regional transactions in payments corridors.

Latin America

In Latin America, Dyna.Ai's research describes fraud and risk as key constraints. It also notes large numbers of adults remain outside formal financial services. It places the number at more than 200 million.

The report describes AI-based credit decisioning and fraud prevention as tools banks are applying to expand lending while maintaining risk controls. It cites BBVA Mexico as an example of AI-enabled decisioning used in support of broader financial inclusion, while maintaining risk discipline.

Operating model

Beyond use cases, the report argues that organisational conditions determine whether AI spending produces revenue. It lists data fragmentation, governance uncertainty and adoption friction as common barriers.

It also argues that leading banks set explicit revenue outcomes and build accountability into delivery from the start. It describes workflow embedding as a differentiator, with AI tools placed into frontline processes rather than run as side programmes.

Dyna.Ai also highlights a contracting approach it calls Results-as-a-Service. It describes this as a model where providers are paid for outcomes rather than software tools.

"The issue isn't experiments, it's accountability," said Skoumal. "Results-as-a-Service means tying AI deployments to measurable business outcomes, not tool adoption. That shift changes how enterprises think about execution when moving from pilots to production."

The company said the report's conclusions reflect a broader shift in banking AI strategies across emerging markets, with more institutions looking for measurable revenue contribution as deployments move into production environments.