Nearly two-thirds of financial services organisations are knowingly deploying untested, unchecked software into production as artificial intelligence dramatically accelerates software development.
The findings highlight a growing headache for banks attempting to balance delivery speed with software quality, operational resilience and regulatory compliance.
The data suggests many financial institutions are no longer struggling simply because software development has become more complex. Instead, AI is generating code faster than many organisations can realistically test or validate.
As a result, many banks are forced into difficult decisions over what gets tested before release and raising fresh questions about governance as autonomous AI tools become embedded across the software development lifecycle.
Those are among the conclusions of the 2026 Quality Transformation Report, published by software testing firm Tricentis and shared exclusively with QA Financial today.
The Tricentis team said they surveyed around 2,500 CEOs, CIOs, CTOs, engineering leaders, QA professionals and software developers across six countries and multiple industries, including financial services.
For banks, the results arrive at a significant moment. Regulators across Europe, the UK and Asia are placing increasing emphasis on AI governance, operational resilience and technology risk, expecting firms not only to deploy AI safely but also to demonstrate that critical software continues to be tested, monitored and governed as development cycles become increasingly automated.
Untested software
The report found that 64% of financial services organisations admit releasing untested software into production, making banking one of the sectors under the greatest pressure. More significantly, most financial organisations are no longer describing these releases as accidental.
While the 64% figure represents only a one percentage point increase on last year’s survey, Tricentis said there is little indication the problem will ease over the coming year.
Across all sectors, one-third of organisations said leadership is prioritising speed over quality, while 30% reported AI is generating more code than development teams could realistically test before release.
Together, the findings suggest the underlying pressures behind untested software are continuing to grow rather than recede.
Last year’s survey largely attributed untested releases to quality slips and development mistakes, the researchers said. This year, respondents acknowledged they are consciously making the trade-off.
Leadership pressure to prioritise delivery speed and the sheer volume of AI-generated software are increasingly driving decisions to release code before testing has been completed. That represents a notable shift in how software quality risk is emerging inside large organisations.
As AI enables developers to produce significantly more software in less time, testing capacity has not expanded at the same pace. Rather than slowing delivery, many organisations appear to be accepting higher levels of software risk.
Far-reaching consequences
For banks, where software increasingly underpins customer channels, payments, trading systems, regulatory reporting and financial crime controls, the consequences extend well beyond software defects.
More than a quarter of financial services respondents said poor software quality would be most likely to result in security breaches or compliance failures. The same proportion highlighted rising technical debt and maintenance costs, while one quarter identified loss of customer or partner trust as the greatest consequence.
The financial impact is equally significant.
Nearly two-thirds estimate poor software quality costs their organisation more than £372,000 annually, while almost one in five believe annual losses exceed £744,000.
Although every organisation measures these costs differently, the findings reinforce a broader trend emerging across financial services: software quality is increasingly becoming a business and governance issue rather than simply an engineering concern.
Regulatory direction
That shift closely mirrors the direction regulators have been taking over the past two years.
Under the EU’s Digital Operational Resilience Act (DORA), financial institutions are expected to demonstrate continuous resilience across critical ICT services, including robust testing and governance.
Similar themes appear throughout the Bank of England’s operational resilience framework, the Monetary Authority of Singapore’s technology risk guidance and the growing body of supervisory work around AI governance from regulators including the FCA.
Increasingly, regulators are asking not simply whether firms are adopting AI, but whether they can prove they remain in control of the software those AI systems help create.
“Financial services organisations face unique pressure to balance innovation, speed and control.”
– Andrew Power
The Tricentis research suggested that governance may now be becoming the real bottleneck. AI adoption itself is accelerating rapidly.
Seven in ten financial services organisations have already implemented AI across some or all software delivery workflows, while 80% say they trust AI agents to make software release decisions.
Yet confidence is not matched by operational readiness as only 40% of financial services respondents said they are very prepared to operationalise, govern and scale AI agents across software delivery.
Implementing AI agents within software development and testing is now cited as the sector’s leading IT priority over the next twelve months, closely followed by establishing governance and trust in those AI systems.
That distinction is important.
The accompanying analysis published alongside the report argues that organisations are moving beyond AI experimentation into production-scale deployment, where governance becomes considerably more difficult than simply implementing the technology.
During pilot projects, AI can be closely monitored and treated as an experiment. Once AI agents begin making release recommendations, generating tests or influencing production decisions across enterprise software delivery, organisations require visibility, auditability, approval processes and clearly defined human intervention points.
According to the report, many firms are only now discovering that those governance structures have not kept pace with adoption. The report also highlighted a widening gap between executive confidence and operational reality.
Across all sectors, 81% of CEOs express high confidence in AI-driven software delivery, compared with just 56% of QA and DevOps professionals. While 44% of C-level executives believe their organisations are very prepared to operationalise and govern AI agents, only 23% of QA and DevOps practitioners share that assessment.
Disconnect
The accompanying blog argues that this does not necessarily indicate disagreement over AI’s potential. Rather, executives typically see strategic progress through faster delivery and increased automation, while engineering and QA teams experience the operational realities of governance gaps, fragmented tooling, accountability questions and the practical challenge of validating AI-generated software at scale.
That disconnect may become particularly significant inside banks. Only 38% of financial services organisations say software developers and senior executives are aligned on what constitutes release readiness and high-quality software.
For organisations operating under increasingly demanding operational resilience frameworks, differing interpretations of release readiness could create governance risks extending far beyond software engineering teams.
Andrew Power, Head of UKI at Tricentis, believes the challenge has fundamentally shifted.
“Financial services organisations face unique pressure to balance innovation, speed and control,” Power said.
“AI is helping teams deliver software faster, but it is also increasing the volume and complexity of software being produced,” he added.
Power continued: “The challenge is no longer simply deploying AI; it is governing AI at the speed it is being deployed. Just as cybersecurity became a board-level priority over the past decade, AI governance is becoming a strategic concern for business leaders.”
He stressed that “organisations need AI-powered testing and quality engineering capabilities that can keep pace with AI-driven software development while maintaining trust, meeting regulatory expectations and strengthening operational resilience as adoption accelerates.”
Other industries
Across the wider survey, financial services was not alone. Overall, 60% of organisations admitted deploying untested code into production, with retailers and energy companies reporting similar pressures.
One-third cited leadership pressure to prioritise speed over quality, while 30% said AI was generating more code than teams could realistically test.
The report argued that organisations are not losing confidence in AI itself. Rather, they are recognising that production-scale AI requires governance models, audit trails and oversight mechanisms that many organisations have yet to build.
That observation is likely to resonate with financial institutions. Over the past year, regulatory discussion around AI has increasingly shifted away from experimentation towards accountability.
Whether under DORA, operational resilience frameworks or emerging AI governance initiatives, supervisors are placing growing emphasis on evidence, assurance and demonstrable control over increasingly autonomous systems.
That changes the role of software testing. Testing is no longer simply about finding defects before software reaches production. It is increasingly becoming part of the evidence firms use to demonstrate operational resilience, effective AI governance and responsible software delivery.
The organisations best positioned to benefit from AI are therefore unlikely to be those generating software the fastest. Instead, they may be the ones capable of proving that speed has not come at the expense of software quality, governance or control.
As AI continues to compress software delivery cycles across banking, the competitive advantage may ultimately lie not in releasing more code, but in demonstrating that every release can still be trusted.
THIS SEPTEMBER IN LONDON
REGISTER TODAY – SIMPLY CLICK HERE
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
- Goldman puts AI coding to the test
- How to test AI models that banks do not control
- OpenAI, Filigran and SunTec: the latest vendor and product news
- Sygnum: Testing AI is ‘a measurement problem’
- Banks’ ‘code for all’ push raises testing risks
QA FINANCIAL PODCASTS
CLICK HERE TO LISTEN TO OUR EXCLUSIVE CONVERSATIONS
