Financial institutions are rapidly accelerating AI deployment across core banking operations, fraud detection, payments, document processing and customer service, with growing adoption of agentic AI now raising fresh questions around software testing, validation, operational resilience and production reliability.
For QA and software testing teams inside banks and financial institutions, the shift marks a major transition from controlled AI pilots toward large-scale production environments where testing, governance and resilience requirements become significantly more complex.
New research from NVIDIA’s latest State of AI in Financial Services survey, based on responses from more than 800 financial services professionals, found that 65% of respondents said their company is actively using AI, up from 45% in last year’s report, while 61% are using or assessing generative AI, up 52% year over year.
The findings suggest banks and financial firms are moving beyond experimentation and proof-of-concept deployments toward enterprise-wide production rollouts that increasingly touch sensitive customer workflows, payments infrastructure, compliance systems and decision-making processes.
One of the report’s most notable findings is the growing rise of agentic AI inside financial institutions. According to the survey, 42% of respondents are now using or assessing agentic AI, while 21% said they have already deployed AI agents.
“Open source models are fundamentally changing the competitive dynamics in financial AI.”
– Helen Yu
The report defines AI agents as advanced AI systems designed to autonomously reason, plan and execute complex tasks based on high-level goals, a development likely to create major new challenges for software testing, quality engineering and operational governance teams across the banking sector.
The report also highlights that institutions are increasingly deploying AI systems into high-risk operational environments where performance reliability, accuracy drift and production stability become critical concerns.
“Performance reliability. Accuracy drift. Agents that look impressive in a demo and wobble in production. Fair concerns when the workflow in front of you decides whether a family gets a home,” one industry executive said alongside the report findings.
Autonomous AI systems
For banking QA and quality engineering teams, the shift toward autonomous AI systems introduces entirely new testing and assurance challenges beyond traditional software validation approaches.
Unlike deterministic banking applications, agentic AI systems can autonomously reason, plan and execute tasks, increasing pressure on firms to strengthen continuous validation, monitoring, adversarial testing and runtime governance frameworks.
The rise of AI agents is also likely to increase focus on model explainability, auditability and evidence generation, particularly as regulators continue scrutinising AI governance, operational resilience and third-party dependencies across financial services.
“Agentic AI systems can now make decisions in under 200-milliseconds that traditional rule-based systems simply can’t match.”
– Helen Yu
Industry observers increasingly warn that many financial institutions are deploying AI faster than their testing and assurance frameworks are evolving, creating risks around hallucinations, model drift, workflow instability and inconsistent outputs across customer-facing systems.
The report found that document processing and management emerged as one of the largest drivers of AI return on investment across the sector, while customer experience and engagement, algorithmic trading and risk management also ranked highly.
“Creating operational efficiencies is the largest improvement AI has made in financial services,” according to 52% of respondents, while 48% cited employee productivity gains.
The survey found that 89% of respondents said AI is helping increase annual revenue and decrease annual costs, with many institutions now reporting measurable commercial benefits from production AI deployments.
For many organisations, the gains are becoming substantial. According to the report, 64% of respondents said AI had helped increase annual revenue by more than 5%, including 29% who reported revenue increases above 10%.
Similarly, 61% said AI had helped decrease annual costs by more than 5%, with 25% reporting cost reductions above 10%.
“The most tangible ROI I’m seeing is in payment operations, specifically authorisation optimisation and intelligent routing,” commented Dwayne Gefferie, a payments strategist at the Gefferie Group.
“Agentic AI systems can now dynamically adjust retry logic based on real-time issuer signals and make routing decisions under 200-millisecond routing that traditional rule-based systems simply can’t match.”
“What makes this compelling is that every basis point improvement in authorization rates translates directly to revenue, there’s no ambiguity in measurement,” Gefferie added.
At the same time, the survey suggests financial institutions are increasingly embracing open-source AI models and infrastructure as they seek to build differentiated AI systems using proprietary banking data.
According to the report, 84% of respondents said open-source models and software are important to their AI strategy.
“Open source models are fundamentally changing the competitive dynamics in financial AI,” said Helen Yu, the current CEO of Tigon Advisory and a technology and AI strategy executive who advises banks and financial services firms on AI transformation, digital strategy and emerging technologies.
“The real value capture happens when institutions fine-tune these models on their proprietary transaction data, customer interaction histories and market intelligence, creating AI capabilities that competitors cannot replicate.”
The growing use of open-source models is likely to intensify focus on software supply chain assurance, third-party model governance and AI validation frameworks inside regulated financial environments.
Banks are increasingly being forced to balance flexibility and innovation with governance, security and testing obligations as open-source foundation models become embedded within enterprise banking environments.
“Open source models can help banks close the gap with early movers, unlock cost efficiencies and safeguard against vendor lock-in, but they’re not without their limitations, proprietary approaches can unlock superior performance for domain-specific tasks,” said Alexandra Mousavizadeh, a financial services intelligence and AI analyst focused on how banks adopt and operationalise AI technologies.
She is the co-founder and co-CEO of Evident Insights, the company behind the widely cited Evident AI Index, which tracks and ranks AI maturity across major global banks.
“Leading banks need to demonstrate proficiency in both approaches, applying the right kind of model to the right problem, in the right context.”
The report also suggests banks are increasingly shifting investment toward production optimisation and AI infrastructure resilience.
About 41% of respondents said investment would go toward “optimizing AI workflows and production,” while 30% said spending would focus on expanding AI infrastructure, including cloud and on-premises deployments.
For QA, DevOps and platform engineering teams, that expansion creates additional pressure around performance testing, failover validation, observability and resilience engineering for GPU-intensive AI workloads operating in production banking systems.
The growing complexity of AI infrastructure is also expected to increase demand for synthetic data testing, continuous monitoring, model benchmarking and automated governance tooling capable of validating AI systems at scale.
Meanwhile, nearly 100% of respondents said AI budgets would either increase or remain stable over the next year, reinforcing expectations that financial firms will continue accelerating AI deployment despite growing governance and operational concerns.
As financial institutions continue moving AI deeper into critical banking infrastructure, the findings also reinforce growing regulatory concerns around AI governance, explainability, resilience and testing standards across the sector.
“The institutions winning in AI are treating their proprietary data as a strategic asset for building differentiated AI products,” Yu concluded.
Why not become a QA Financial subscriber?
It’s entirely FREE
* Receive our weekly newsletter every Wednesday * Get priority invitations to our Forum events *
REGULATION & COMPLIANCE
Looking for more news on regulations and compliance requirements driving developments in software quality engineering at financial firms? Visit our dedicated Regulation & Compliance page here.
READ MORE
WATCH NOW
QA FINANCIAL PODCASTS
