Firms with trusted AI practices report stronger returns, study finds

Organisations with stronger artificial intelligence governance and data practices are significantly more likely to report higher returns from their AI investments, according to a new study by SAS and IDC. The second annual Data and AI Impact Report, titled ‘The New Economics of Trust‘, found that organisations applying what it describes as trustworthy AI practices were 15 times more likely to report strong or high returns on investment from their AI projects. The study found that organisations with the strongest governance, data quality and auditability practices reported at least twice the return on investment from AI deployments compared with other organisations, although SAS said this group represented a comparatively small part of the market. Fewer than one in 20 organisations identified as laggards in trustworthy AI reported similarly strong returns, according to the report. “When AI works, it’s incredibly impactful,” SAS chief technology officer Bryan Harris said. “However, it is well documented that state-of-the-art agents can have error rates that exceed 25 per cent on complex tasks, which is unacceptable in high-stakes decision-making,” he added. “In order to achieve accuracy and repeatability, organisations must embed domain expertise into agentic workflows, while keeping people at the centre of governance and oversight,” he continued. “Organisations that do this successfully will close the trust gap and gain a competitive advantage in the market with AI,” Harris stated. “As AI becomes more autonomous, organisations face a new challenge: maintaining confidence in systems people don’t fully understand,” IDC vice president Chris Marshall explained. “Our findings show that stronger oversight, explainability, accountability and data foundations are becoming prerequisites for scaling AI successfully,” he added. The report found that employees frequently override AI-generated recommendations, with a lack of explanation for an AI decision emerging as the main reason for doing so. Researchers said employees were increasingly reluctant to rely on systems when they could not establish whether their output was correct or understand how an AI system had reached a final decision. The report said the problem could become more significant as AI systems gained greater autonomy, making the ability to explain decisions increasingly important to successful adoption. It also found that a lack of confidence in AI decisions could lead employees to make manual corrections, potentially reducing productivity and profitability while increasing the time required to complete tasks. According to the study, 97.2 per cent of users override AI-generated recommendations in at least some cases. The most common reason for overriding an AI recommendation, regardless of whether employees considered the output correct, was that the system could not explain the reasoning behind its decision. The report also found that trust in AI declined as systems became more autonomous, falling from 76 per cent for generative AI to 66 per cent for agentic AI. SAS and IDC said the findings pointed to a widening gap in returns between organisations that prioritise trustworthy AI practices and those that do not. The report suggested that organisations obtaining the greatest value from AI were not necessarily using different technologies, but were instead managing and governing AI differently. Organisations investing in trustworthy AI measures were 15 times more likely to report strong or high returns on investment, with 62 per cent doing so compared with 4 per cent among organisations that did not have such practices. Organisations with the strongest trustworthy AI practices also reported 1.85 times greater gains across 13 business outcomes, including revenue growth, cost savings and customer experience. The study found that 85 per cent of AI leaders with strong trustworthy AI practices were increasing their investment in the area by more than 10 per cent during 2026, potentially widening the performance gap identified in the report. The research also highlighted weak or outdated data foundations as a major obstacle to AI adoption. SAS and IDC said many organisations were deploying AI on underdeveloped or outdated data and data infrastructure, limiting their ability to provide the transparency and explainability required to govern AI effectively. Only 17.5 per cent of enterprises were found to have a fully optimised data infrastructure considered mature enough to meet the demands of agentic AI. The report said organisations with an optimised data foundation were four times more likely to expect strong returns from AI projects. They were also six times more likely to require data quality and explainability controls intended to build trust in AI systems. The findings were based on a global survey of 2,699 decision-makers with knowledge of or influence over their organisations’ data and AI initiatives. The survey covered 28 countries and four focus industries, namely banking, insurance, life sciences and the public sector. The report included industry-specific use cases and findings examining how organisations in those sectors were approaching AI. In banking, 85 per cent of AI leaders had established AI governance frameworks, compared with 29 per cent of organisations classified as laggards in the study. The report said banking leaders were increasingly treating robust AI governance as a competitive and operational issue rather than solely a compliance requirement. In the public sector, 41 per cent of leaders said they were increasing investment in trustworthy AI by more than 20 per cent in the year ahead, a rate SAS said was among the highest across the industries surveyed. In life sciences, 23 per cent of organisations had scaled AI across their entire companies, the highest proportion recorded among the four industries examined. The report defined trustworthy AI as artificial intelligence designed to be reliable, fair, secure and compliant with regulatory requirements, while also being capable of showing how it reached a decision. It said users and decision-makers throughout an organisation should be able to hold an AI system against a predetermined chain of accountability when its output was incorrect or missing. AI systems should also be governed according to clear rules and their compliance with those rules should be demonstrated, according to the report. SAS and IDC assessed organisations against five dimensions of trustworthy AI and assigned them scores out of 100. Organisations with an average score of at least 80 were classified as trustworthy AI leaders for the purposes of the study. The assessment covered data quality and governance, model governance and oversight, explainability and fairness, responsible AI policy, and audit and accountability.