Tricentis has published research showing that 60% of organisations are deploying untested code into production.
The survey highlights mounting pressure on software teams as AI speeds up development.
The findings are based on a survey of 2,501 respondents in the US, UK, Ireland, Germany, Japan and Singapore, including CEOs, CIOs, CTOs, engineering leaders, QA and DevOps professionals, and developers at organisations with more than 150 employees.
The share of organisations releasing untested code has changed little from the previous year, when 63% reported doing so. The latest results suggest the reason has shifted, with organisations now making deliberate trade-offs rather than suffering accidental lapses.
Leadership pressure to prioritise speed over quality was cited by 32% of respondents as a reason for shipping untested code. Another 30% said the volume of AI-generated code was too large for teams to test fully.
The pattern spans sectors. More than half of organisations in every industry surveyed said they had deployed untested code, including 64% in financial services, 63% in retail, and 58% in energy and utilities.
Boardroom gap
The research also points to a divide between senior executives and technical teams in confidence in AI systems. More than four in five CEOs, or 81%, said they had high confidence in AI-driven systems and tools, compared with 56% of QA and DevOps professionals.
A similar gap appeared in preparedness to govern and scale AI agents through the software development lifecycle. Some 44% of C-level executives said their business was very prepared to do so, compared with 23% of QA and DevOps professionals.
Adoption is already well advanced. Nearly half of organisations, or 48%, said they had fully implemented AI internally. Of that group, more than half said their AI tools and processes changed regularly.
One-third of teams identified tool complexity and sprawl as a major barrier to achieving continuous software quality at scale. Skills gaps were cited by 33%, while 28% said code volume was increasing faster than they could manage and 26% pointed to a lack of clear quality and trust metrics.
The report suggests that confidence in agentic AI is high on paper, but day-to-day operations remain difficult. It found that 83% of organisations trust agentic AI to make release decisions, while 82% believe they are prepared to operationalise and govern AI agents at scale.
At the same time, respondents reported persistent obstacles including untested code, security concerns, skills shortages and data quality problems. Security concerns were cited by 27% of respondents, while 24% pointed to skills gaps and another 24% to data quality issues.
Financial cost
The commercial impact of weak software quality also featured strongly in the research. One in five organisations said poor software quality costs them between USD $1 million and USD $5 million each year.
Nearly half, or 45%, estimated annual losses of between USD $500,000 and USD $1 million. Security and compliance failures were named as the biggest source of those losses by 30% of respondents, followed by technical debt and rework at 28%.
These figures place software testing and governance more firmly in the realm of business risk rather than a narrow engineering concern. The survey linked quality failures not only to internal inefficiencies, but also to broader questions of trust in software as AI-generated output becomes more common.
Kevin Thompson, Chief Executive Officer at Tricentis, said the combination of faster development cycles and weaker controls was becoming harder for companies to manage. "Accelerating business transformation initiatives is one of the top priorities for today's C-suite, and AI has the potential to help software development teams move faster than ever before. However, with increased speed comes increased risk. When software quality processes fail to keep pace with development speed, organizations often respond by taking shortcuts that materially degrade or reduce confidence. Our research highlights the growing pressure teams are facing to balance speed, quality and control as software development accelerates. As risks like financial performance and customer trust become more visible and measurable, software quality can no longer be treated as just an engineering concern. It must become a boardroom imperative," said Thompson.
He added that many organisations were still using methods built for an earlier stage of software development. "Many organizations are still relying on quality processes that weren't designed for software development in the AI era. As development accelerates, leaders need clearer visibility into software quality risk and stronger alignment between engineering, QA and the broader business. The organizations that succeed will be the ones that can scale speed and control together," said Thompson.