The second edition of the Data and AI Impact Report: The New Economics of Trust finds that companies following the strictest practices in governance, data quality and explainability consistently outperform peers. According to the report, a relatively small group of organizations that invest in trustworthy AI achieve at least twice the returns on their AI initiatives compared with others.
Key findings
- Organizations applying trustworthy AI practices are fifteen times more likely to report “strong or high” returns on AI projects: 62% of these organizations reported such outcomes compared with 4% among others.
- Companies following the most reliable practices achieved 1.85 times greater gains across 13 business metrics, including revenue growth, cost savings and customer experience.
Trust, overrides and explainability
The report warns that AI systems that cannot explain themselves pose a significant business risk. Researchers observed that employees are increasingly reluctant to rely on systems whose correctness cannot be verified or explained. That reluctance often leads staff to override or manually correct AI decisions, which consumes time, reduces productivity and increases costs, further eroding trust.
Specific trust and override statistics from the report:
- 97.2% of users override AI recommendations at least occasionally.
- Respondents most commonly reported overriding systems because the AI could not justify its decision.
- Trust in generative AI stands at 76%, while trust in more autonomous AI agents is lower at 66% — indicating growing concern as systems become more independent.
Bryan Harris, Chief Technology Officer at SAS, notes that while well-functioning AI can have enormous impact, top AI agents can still exceed a 25% error rate on complex tasks — an unacceptable level in high-stakes environments. He argues that domain expertise must be embedded into agent workflows and that humans should retain oversight.
Chris Marshall, Vice President at IDC, adds that as AI becomes more autonomous, organizations face the challenge of trusting systems they do not fully understand. The report underscores that stronger oversight, explainability, accountability and robust data foundations are increasingly prerequisites for scaling AI successfully.
What makes AI "trustworthy"?
The report defines trustworthy AI as systems designed to operate reliably, ethically and securely, to comply with regulations, and to clearly show how decisions were reached. Organizations must establish clear accountability chains so users and decision-makers at every level can hold AI systems responsible for faulty or missing outputs. All AI systems should be subject to governance principles and demonstrable rules.
Data foundations and infrastructure gaps
Most organizations today build AI on weak or outdated data and data infrastructure. Without solid data foundations there is no transparency or explainability, making it difficult to manage AI effectively or extract real value.
- Only 17.5% of companies have fully optimized data infrastructure that meets the needs of AI agents, a shortfall that constrains performance and returns.
Why this matters
The report issues both an economic and operational warning: companies that invest in trustworthiness see markedly better outcomes, while those that do not are far less likely to realize high returns. The conclusion is clear — to deliver sustainable value from AI, organizations must invest not only in models but also in governance, explainability and reliable data foundations.
