As artificial intelligence takes over more of the work of writing software, the competitive advantage for banks may increasingly shift from software development itself to proving that AI-generated applications are correct, secure and behave as intended, and they have been thoroughly tested and validated.
That is the argument emerging from a new analysis by Sudhir Pai, Executive Vice President and Deputy Group CTO at Capgemini, who believes software engineering is undergoing a “foundational re-architecting” driven by autonomous AI agents.
The implications could be profound. If AI increasingly generates applications from high-level intent, software quality becomes less dependent on how code is written and more dependent on how rigorously systems are verified before and after deployment.
Pai argued that “the discipline of software engineering is undergoing a foundational re-architecting, driven by the rise of autonomous AI agents.”
“Such a transformation is not incremental; it redefines the engineer’s very identity by elevating the role from coding, a cognitive task being automated, to orchestration, a strategic, metacognitive task,” he wrote.
“The discipline of software engineering is undergoing a foundational re-architecting, driven by the rise of autonomous AI agents.”
– Sudhir Pai
For financial institutions investing heavily in generative AI, agentic AI and AI-assisted software development, that could represent a significant shift in where engineering effort is focused.
Rather than spending most of their time writing software, development teams may increasingly concentrate on defining intent, orchestrating AI-driven development pipelines and validating the behaviour of the resulting applications.
That places QA functions squarely in the centre of software delivery. Rather than asking whether developers wrote good code, banks will increasingly need to demonstrate that AI-generated systems behave correctly, remain secure, comply with regulatory requirements and continue operating reliably as models, prompts and business rules evolve.
Pai believes engineers themselves will have to adopt a fundamentally different role.
“The new era of artificial Intelligence is not simply asking engineers to learn a new tool; it is asking them to unlearn their core identity,” he stated.
Arrival of compilers
Pai compares the shift with the arrival of compilers, which abstracted away assembly language and allowed engineers to focus on higher-level programming.
“Today, AI agents are the ‘new compiler.’ Engineers will state their intent, and the AI will then consider a pool of AI agents to create and build into a fully tested, secure and documented application by orchestrating a swarm of other agents.”
For QA professionals, that single sentence may be one of the most significant observations in the paper. If AI systems are expected to produce “fully tested, secure and documented” applications, somebody still needs to verify that those claims are true.
That moves software assurance away from reviewing individual code commits and towards validating entire AI-generated systems, including behaviour, security, resilience, integrations, documentation and governance.
The approach also aligns closely with a broader regulatory trend already emerging across financial services.
The European Central Bank has repeatedly stressed that banks need stronger evidence of resilience as AI transforms both software delivery and cyber risk.
Under DORA, supervisors increasingly expect institutions to demonstrate not simply that controls exist, but that systems can detect, respond to and recover from realistic scenarios.
Likewise, growing interest in behavioural testing for generative and agentic AI reflects a wider move away from one-off validation towards continuous assurance throughout an application’s lifecycle.
In that environment, QA teams are no longer validating only software functionality. They are increasingly validating AI behaviour.
Pai argued that this requires an entirely different set of capabilities. “This new reality requires a ‘mastery of something else.’ It demands a shift to systems thinking, a mastery of ‘toolchains,’ and an ability to be the ‘connector of dots.'”
He added: “The AI is not a ‘co-pilot’ sitting next to you; it is a cognitive peer.”
“An engineer’s job is no longer to be the developer; it is to make the best possible use of that developer.”
– Sudhir Pai
For software testing teams, that evolution changes the nature of quality engineering itself. As AI becomes responsible for generating increasing amounts of software, testing shifts towards verifying whether AI systems consistently produce reliable, explainable and compliant outcomes under realistic operating conditions.
The paper also highlights a growing governance challenge. “The ‘Return to Build’ brings immense opportunity but also a familiar danger,” Pai wrote.
“Without a standardised enterprise-wide blueprint, the ease of AI development will lead to a new generation of ‘spaghetti.'”
He added: “This will result in thousands of siloed, unmanageable applications and agents that are unable to be coordinated and tracked efficiently.”
That warning is particularly relevant for large financial institutions, many of which are already experimenting with hundreds of AI use cases across different business units.
Without common testing frameworks, governance standards and validation processes, the rapid expansion of AI-generated software could create significant operational and technology risk.
Pai concluded that “the re-architecting of software engineering is an immediate, foundational shift.”
However, the larger story may be that software engineering itself is becoming a verification discipline. As AI increasingly generates applications, competitive advantage shifts from writing code to proving that AI-generated software behaves safely, securely and predictably. In that new model, quality engineering moves from supporting software delivery to becoming one of its primary control functions.
16 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
- AI adoption strains JPMorgan testing
- Can banks ‘outsource’ AI accountability?
- HDFC Bank raises testing stakes
- Is observability banking QA’s next discipline?
- Barclays on AI testing, telemetry and kill switches
WATCH NOW
QA FINANCIAL PODCASTS
CLICK HERE TO LISTEN TO OUR EXCLUSIVE CONVERSATIONS
