For a long time, assessing software quality meant counting bugs and reviewing test results. That approach is outdated. According to the World Quality Report 2025-26 by Capgemini and Sogeti (based on a survey of more than 2,000 executives in 22 countries), 89% of organizations now test or use AI in their quality processes, but only 15% have truly scaled it organization-wide, while 52% remain stuck in the pilot phase. The number of companies that aren’t using AI at all even rose from 4% to 11%—a sign that failed pilot projects are causing some organizations to backtrack.

Where it does work, AI delivers an average productivity gain of 19%. The biggest barriers are not technical, but structural: data sensitivity (67%), integration complexity (64%), doubts about the reliability of AI output (60%), and a lack of internal AI expertise (50%).

It is also striking to see where AI is being used: no longer just to analyze defects after the fact, but increasingly earlier in the process, during the definition of requirements and test design. The use of synthetic test data for quality analysis rose from 14% to 25% in just one year.

For anyone structuring a quality audit, ISO/IEC 25010:2023 remains the benchmark: the second edition of this international standard describes nine quality attributes—ranging from functional suitability and performance to security—as criteria throughout the entire software lifecycle.

At M2Q, our Quick Scans and Audits are based on exactly this kind of up-to-date framework, not on outdated checklists.