60% of firms deploy untested AI code, risking $1M losses

Tricentis’ 2026 Quality Transformation Report found that 60% of global organizations deploy AI-generated code to production without testing it first. Thirty-two percent of respondents reported they ship untested code under pressure from leadership. The share is slightly lower than the 63% reported in 2025; the report attributed the change to a move from accidental lapses to deliberate choices to prioritize speed.

Respondents identified several barriers to maintaining quality. Around 30% said the volume of AI-generated code is overwhelming testing pipelines. Tool complexity and sprawl and skills gaps were each reported by roughly one-third of organizations. Other issues included 28% who said code volume is growing faster than their testing capacity and 26% who reported a lack of clear quality and trust metrics. Security problems were listed by 27% of respondents and data quality issues by 24%.

The trend spans industries. Financial services reported the highest rate of untested deployments at 64%, retailers followed at 63% and energy and utilities at 58%. Forty-eight percent of organizations said they have fully implemented AI internally, and more than half of those said their AI tools and processes change frequently.

The survey also showed a divide between executives and engineering teams. Just over four in five CEOs expressed high confidence in AI-driven systems, compared with 56% of QA and DevOps staff. Forty-four percent of C-level executives said their business was very prepared to operationalize, govern and scale AI agents across the software development lifecycle, while 23% of QA and DevOps professionals agreed. Eighty-three percent of organizations reported trusting agentic AI to make release decisions and 82% said they are prepared to govern AI agents at scale.

About one in five organizations reported annual losses exceeding $1 million tied to software problems. Security and compliance failures accounted for about 30% of those losses and technical debt and rework for roughly 28%.

Kevin Thompson, chief executive of Tricentis, warned that faster development can increase risk when quality processes lag. He called for clearer visibility into software quality risk and stronger alignment between engineering, QA and the broader business.