AI is becoming ever-more important for the efficient functioning of business operations. But there is a frontier that cannot be surmounted with a model alone: building trust.
While many enterprises are investing heavily in AI capabilities, encouraging employees to use these technologies does not automatically translate into confidence in their outputs or decisions.
For Dr. Cecilia Dones, Founder and Chief Data Officer at 3 Standard Deviations and Adjunct Faculty at Columbia University, understanding this distinction is essential for leaders hoping to realize lasting value from AI investments.
In a recent interview with Corinium Global Intelligence, Dr. Dones argues that organizations must look beyond adoption metrics to consider the relationship between technological reliability, employee confidence, and the organizational culture surrounding AI.
Understanding the Difference Between Trust and Trustworthiness
When organizations discuss building trust in AI, they often treat trust as a single concept. However, Dr. Dones highlights three distinct dimensions: trustworthiness, trust, and trusting behavior.
Trustworthiness refers to whether an AI system performs reliably and produces acceptable outcomes. Trust describes an individual's belief or confidence in that system. Trusting behavior, meanwhile, concerns whether someone actually relies on the technology.
These dimensions are not necessarily aligned.
An employee might use AI regularly because their employer requires it, even while questioning the quality of its outputs. Conversely, someone may recognize that a system is reliable without choosing to incorporate it into their daily work.
This creates challenges for organizations attempting to measure the success of AI transformation programs.
Dr. Dones explains that when employees feel pressured to use technologies they do not fully trust, the disconnect between their attitudes and behaviors can create cognitive dissonance.
For data and analytics leaders, the implication is significant. High usage rates may demonstrate compliance with an organizational mandate rather than genuine confidence in AI.
Building trust therefore requires more than ensuring employees have access to technology. Organizations must demonstrate that systems are dependable while giving people reasons to believe in their value.
Psychological Safety Is Essential to AI Adoption
Beyond technical reliability, Dr. Dones believes leadership has an important role in shaping how employees experience AI.
Organizations must demonstrate that systems are competent, consistent, and genuinely useful. However, employees also need reassurance about how these technologies will affect their roles.
Dr. Dones emphasizes the importance of creating an environment where employees feel comfortable exploring AI without fearing negative consequences.
This psychological safety allows employees to discover how AI can complement their expertise rather than simply viewing it as a potential replacement.
By contrast, organizations that continually pressure employees to use AI risk encouraging fear-driven behavior.
Employees may comply with adoption requirements without exploring how the technology could genuinely improve their work.
For leaders, creating a culture of experimentation is therefore as important as investing in the underlying technology.
Human Oversight Must Come With Genuine Authority
Trust becomes especially important when AI systems begin influencing critical business decisions.
However, Dr. Dones argues that simply placing a human within an AI-driven process may be insufficient for true oversight.
If employees are expected to monitor automated decisions without having the authority to intervene, their involvement may provide little meaningful protection.
"Human in loop to me means if something is going wrong with the machine in the agentic context, human in loop means the human can stop the machine."
For data leaders, this raises important questions about accountability, decision-making authority, and the governance controls surrounding AI systems.
Effective oversight requires clear responsibilities and the ability to act when technology produces unexpected or harmful outcomes.
Without those safeguards, human involvement risks becoming a procedural exercise rather than a genuine mechanism for improving trustworthiness.
_
Dr. Dones contributed to our recent research report, published in partnership with FICO, read the full report: The Confidence Gap in AI Decisions.