When we wrote about the “AI speed trap” in the software development process for TechRadar Pro a year ago, the central concern was whether faster delivery would come at the expense of quality. A year on, what we’re seeing are actually deeper issues around trust, governance and organizational alignment.
Head of UK & Ireland at Tricentis.
As AI-generated code becomes more routine and ecosystems become more complex, the challenge is no longer simply keeping pace with delivery. It is deciding who owns quality, how much risk is acceptable, and what confidence looks like in an AI-enabled environment.
AI has become firmly embedded across the software development lifecycle, accelerating delivery at an unprecedented rate. But the real story is no longer just speed. Organizations are now being forced to make more explicit decisions about where quality ends and acceptable risk begins, and to reconcile trade-offs around testing, trust and accountability.
Latest Videos FromTechRadar
AI is accelerating software delivery, but quality strategies are falling behind
Nobody would dispute that AI has transformed software quality over the last year. AI-assisted coding, automated testing and increasingly autonomous workflows are no longer experimental; they are becoming standard practice.
Nearly seven in ten organizations have now implemented AI tools across at least some of their software delivery workflows, and almost half have fully embedded it into their development environments.
The impact on productivity is undeniable. Development teams can generate code faster than ever, automate repetitive tasks and accelerate release cycles that once took weeks into days, or even hours. But the ability to create software has advanced far more quickly than organizations’ ability to validate it.
As a result, 60% of teams are knowingly releasing untested code. Not accidentally, as our 2025 research showed, but because they are under pressure to ship faster, and because the sheer volume of code is outpacing what many teams can realistically test. In other words, these are conscious business decisions. Organizations aren’t even trying to test everything; they’re deciding what they can afford not to test.
As software delivery accelerates, quality has become less about eliminating every defect and more about determining which risks are acceptable. But that requires stronger governance, clearer accountability and a shared understanding of what constitutes an acceptable release.
The definition of quality is changing
In an AI-driven world, organizations will never be able to test every line of AI-generated code. While AI has removed many of the bottlenecks associated with writing code, it hasn’t created more time to validate it. Traditional quality models were designed for an era where development speed largely dictated testing speed. That assumption no longer holds.
The answer, however, isn’t to lower the bar for software quality. It’s to rethink what quality actually means. Rather than attempting to preserve old approaches by asking teams to test everything, organizations need to rethink what software quality actually means.
Increasingly, success will depend on understanding where quality matters most. Which applications are business-critical? Which customer journeys carry the greatest commercial or regulatory risk? Which changes genuinely require exhaustive testing, and which can be validated through risk-based approaches?
In other words, software quality is becoming intelligence-led rather than activity-led. The focus is shifting from test volume to confidence in release decisions, based on how well organizations can identify and manage the risks that matter most.
Why governance, visibility and continuous validation are becoming the new foundations of software quality
As organizations move towards risk-based quality models, a critical gap is emerging: how those risks are actually governed in real time.
In traditional software delivery, governance was often applied at fixed stages. In AI-accelerated environments, that model breaks down. Code is generated continuously, pipelines move faster than human review cycles, and release frequency has increased beyond what stage-gated governance was designed to support.
This requires three structural shifts.
First, visibility becomes foundational. Without real-time insight into what AI is generating, what is being tested, and where coverage gaps are emerging, organizations are effectively making risk decisions without a complete view of the system.
Visibility is the precondition for trust, allowing leaders to understand not just whether software works, but whether it is being adequately validated as it evolves.
Second, governance must move closer to code creation. Instead of acting as a final checkpoint before release, governance increasingly needs to be embedded earlier in the lifecycle, where risk is first introduced. This means bringing quality signals, compliance requirements and risk thresholds into the development process itself, rather than evaluating them retrospectively.
Third, continuous validation replaces discrete testing phases. As AI increases both the speed and volume of code production, organizations are shifting from episodic testing cycles to always-on validation. Rather than asking whether software has been tested, the more relevant question becomes whether it is being continuously validated against the organisation’s risk appetite as it changes.
In AI-driven environments, confidence in software quality depends less on point-in-time approvals and more on the ability to observe, measure and validate quality as it is being created.
Organizations are making deliberate decisions about risk
Software quality has always involved balancing competing priorities, but AI is making those trade-offs more explicit, with “good enough” increasingly defined at the intersection of engineering constraints and business pressure.
The challenge is that organizations don’t always agree on what that looks like. Nearly half report only partial alignment between executives and software teams on what high-quality software actually looks like, suggesting that release readiness is increasingly open to interpretation rather than governed by consistent standards.
This also explains a key finding in our data: that while more than nine in ten C-level executives express confidence in their testing strategies, almost one-third of quality assurance (QA) and DevOps leaders remain uncertain that those strategies adequately address the most critical software risks.
That gap reflects fundamentally different perspectives. Executives naturally focus on strategic outcomes and business velocity, while engineering teams experience first-hand the operational realities of validating increasingly complex software systems.
Closing that gap will become increasingly important as AI continues to accelerate software delivery. Trust cannot exist if different parts of the organization operate under different definitions of quality or acceptable risk.
Software quality has become a business decision
For years, software quality was viewed as a technical responsibility – a discipline concerned with finding defects before customers did. That is no longer sufficient.
When software underpins customer experiences, regulatory compliance and critical business operations, decisions about quality inevitably become business decisions. Shipping software with known gaps can no longer be merely a technical compromise; it represents an organizational risk assessment.
The consequences extend well beyond development teams. Poor software quality can increase technical debt, introduce security vulnerabilities, create compliance challenges and erode customer trust. Many organizations are already attributing hundreds of thousands (if not millions) of pounds to the downstream impact of software failures and rework.
At the same time, governance has struggled to mature at the same pace as AI adoption. While many trust agentic AI to make release decisions, significantly fewer believe they are ready to do so at scale. That disparity between adoption and oversight is likely to become one of the defining challenges of the next phase of AI-enabled software delivery.
Ultimately, software quality is no longer something that can be delegated solely to engineering teams. It requires leadership alignment, shared accountability and governance that keeps pace with increasingly autonomous development.
Conclusion
AI has accelerated software delivery to the point where existing approaches to testing and governance are struggling to keep up. The organizations most likely to succeed are those that make deliberate, well-governed decisions about risk and build trust into the software delivery lifecycle rather than treating it as a final checkpoint.
Software quality can no longer be just an engineering discipline. In the AI era, it has become a leadership responsibility.
We’ve featured the best laptop for programming.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit
