Software quality costs companies millions - AI exacerbates the problem

The current Quality Transformation Report 2026 by Tricentis, a leading provider in the field of agentic quality engineering, paints a sobering picture: although companies have significantly increased their development speed through the increased use of artificial intelligence, many are struggling to maintain confidence in the quality of their software. The global study is based on a survey of more than 2,500 managers and experts from Germany, Ireland, Japan, Singapore, the United Kingdom and the USA.

Untested code is deliberately put into production

A key finding of the study: despite significant progress in the use of AI tools, 6 out of 10 companies worldwide state that they are using untested code in production. In Germany, the figure is even higher - at almost 7 out of 10 companies. There has been a significant shift compared to the previous year: While in 2025 most companies still attributed this to unintentional quality deficiencies, today they openly admit to knowingly using untested code. The main reasons cited by respondents were pressure from management to prioritize speed over quality (32% globally, 35% in Germany) and the sheer volume of AI-generated code that teams are simply no longer able to fully test (30% globally, 32% in Germany).

This pattern runs through all sectors: In each of the sectors surveyed, more than half of the companies stated that they put untested code into production. The pressure to act is particularly pronounced among financial service providers (64%), in retail (63%) and in the energy and utilities industry (58%).

AI introduction overtakes quality assurance

Almost half of the companies surveyed (48% globally, 49% in Germany) have fully implemented AI internally. However, of these, more than 50 percent report that their AI tools and processes are constantly changing. Teams cite tool complexity and sprawl (33%), skills gaps (33%), a volume of code that is growing faster than can be managed (28%) and a lack of standardized quality and trust metrics (26%) as key obstacles to scalable and continuous software quality.

Management and operational teams see the situation differently

The discrepancy between the perception at management level and the operational reality in the teams is particularly striking: more than three quarters of C-level executives (81%) state that they have great confidence in AI-supported systems and tools - compared to only 56% of QA and DevOps specialists. The discrepancy is similarly large when it comes to organizational maturity: 44% of executives consider their company to be well positioned to productively deploy and scale AI agents in the software development lifecycle. Among QA and DevOps professionals, only one in four (24%) share this assessment.

It is true that 83 percent of companies worldwide (78 percent in Germany) trust AI agents for release decisions, and 82 percent (76 percent in Germany) state that they are ready to introduce AI agents on a large scale. At the same time, many are still struggling with untested code, tool sprawl, security concerns, skills gaps and data issues.

Financial risks increase significantly

The economic consequences of poor software quality are considerable: one in five companies worldwide (20 percent) loses more than one million US dollars annually due to poor software quality - in Germany it is almost one in three (30 percent). The primary cost drivers are security and compliance breaches (30 percent) as well as technical debt and rework (28 percent). In Germany, higher technical debt and maintenance costs (30 percent) and increased customer fluctuation (26 percent) were cited in particular. Almost half of those surveyed (45%, 53% in Germany) estimate annual losses of between 500,000 and one million US dollars.

Kevin Thompson, CEO of Tricentis, sums up the situation: «Accelerating business transformation initiatives is one of the top priorities at C-level. AI has the potential to move software development teams forward faster than ever before. However, with greater speed comes greater risk. When quality processes don't keep up with the pace of development, organizations often take shortcuts that measurably undermine confidence in software quality.»

Thompson adds: «Many organizations still rely on quality processes that were not designed for software development in the AI age. As development cycles accelerate, leaders need more visibility into software quality risks and closer alignment between engineering, QA and the broader business environment. Successful organizations will be those that can scale speed and control together.»

The Quality Transformation Report 2026 thus marks a significant shift compared to the previous year: the key question is no longer whether companies can use AI - but whether they can maintain trust, control and quality assurance in their releases on a large scale. 

Source: www.tricentis.com