Article illustration 1

Main article image: J Studios/DigitalVision via Getty Images

A startling contradiction lies at the heart of corporate AI adoption: businesses are racing to implement artificial intelligence while simultaneously distrusting the very tools they deploy. According to a comprehensive global study by SAS and IDC surveying over 2,300 technology leaders, 65% of organizations already use AI with another 32% planning implementations within a year—yet 78% admit to lacking complete trust in these systems. This trust deficit directly impacts ROI, with MIT research suggesting 95% of enterprise AI use cases fail to deliver measurable value.

"This misalignment leaves much of AI's potential untapped," warns Chris Marshall, IDC's VP of Data, Analytics, and AI Research. "ROI remains lower where there's a lack of trustworthiness." The study identifies three critical roadblocks eroding confidence:

The Triple Threat to AI Trust

  1. Infrastructure Inadequacies: Weak cloud foundations and data pipelines prevent reliable AI performance
  2. Governance Gaps: Only 40% of organizations implement explainability frameworks despite transparency being crucial for trust
  3. Skills Drought: Workforces lack specialized AI competencies needed for proper implementation and oversight

The Human Bias Blind Spot

Perhaps the most surprising revelation involves a psychological quirk: respondents reported higher trust in generative AI (like ChatGPT or Gemini) than traditional machine learning models—despite GenAI's notorious hallucinations and black-box nature. Researchers attribute this to an innate human tendency:

"The more 'human' an AI feels, the more we trust it, regardless of its actual reliability. We're biologically wired to respond to human-like interaction, even when it's algorithmically simulated."

This explains why employees sometimes develop misplaced confidence in conversational AI outputs, with 43% admitting to sharing sensitive data with AI systems according to supplemental studies.

The Path to Trustworthy AI

Addressing the trust crisis requires concrete actions:
- Prioritize explainability: Implement governance frameworks that document data sources and decision pathways
- Upskill strategically: Develop AI literacy programs instead of replacement-focused layoffs
- Modernize infrastructure: Build robust data pipelines before deploying complex models

As Marshall concludes: "Trust isn't an AI feature—it's an organizational discipline. The companies closing the ROI gap treat trustworthy AI as engineering practice, not magic."

Source: SAS-IDC Global AI Trust Study (2025); Original reporting by Webb Wright/ZDNET