Enterprise enthusiasm for AI is easy to measure. Business value is harder. Organizations have poured billions into generative AI, but many still struggle to turn pilots into measurable financial returns. A 2025 MIT NANDA report found that despite a running total of up to $40 billion in enterprise GenAI investment, 95% of organizations in its research saw no measurable P&L impact from their initiatives.
AI adoption is moving into a more consequential phase. The goal is no longer to simply give employees a chatbot that summarizes documents or drafts emails. Organizations increasingly want AI systems that can recommend actions, automate workflows and eventually act with greater autonomy.
Yet many employees still have unresolved concerns about AI’s accuracy, security, transparency and impact on their jobs. As organizations ask these tools to do more, those concerns become harder to ignore.
That raises an important question: If employees do not trust AI today, what happens when they are asked to hand it more control tomorrow?
For technology providers and channel partners, trust isn't a communications problem to solve after deployment. It has to become part of the deployment itself.
Trust is different when AI touches critical work
AI use is already becoming routine for many knowledge workers. But the trust threshold rises when the technology moves into IT operations.
General knowledge workers are showing increasing trust in — and even dependence on — AI, according to Andrew Hewitt, Vice President of Strategic Technology at TeamViewer. IT administrators are different. “They often work within highly defined processes built over years or even decades, and the consequences of a bad action can be much higher,” he says.
TeamViewer research confirms that caution. Only 1 in 20 IT respondents surveyed felt comfortable with AI acting without a human in the loop. Instead, IT professionals are generally more comfortable using AI to provide analysis or recommendations while keeping a person in control of the final action.
An AI assistant that produces a weak meeting summary creates inconvenience. An AI system that pushes a bad update across thousands of endpoints can grind the business to a halt.
The trust problem is not simply whether AI is “good” or “bad.” It is whether people understand what it does, know its limits and can retain meaningful control when the stakes rise.
Employees need to know what AI means for them
Accuracy and security are only part of the equation. Employees also want to understand what greater AI adoption means for their jobs.
That concern cannot simply be dismissed as resistance to change. AI will alter roles, especially as more routine work becomes automated. But Hewitt argues that for technical IT positions, the near-term shift is more likely to be a change in responsibilities than wholesale replacement.
An administrator who once performed every technical task manually may increasingly orchestrate AI across multiple tools and review recommendations. That creates a different type of work requiring new skills.
Leadership has to make that transition visible. Employees need training and a credible explanation of where human expertise still fits. Career paths can then be defined with more clarity. Telling workers to embrace AI while leaving them to speculate about whether it will eliminate their role is unlikely to build confidence.
The same principle applies to productivity expectations. Leaders should be clear about what AI is supposed to improve, how employees will be evaluated and where human judgment remains essential.
Without that transparency, adoption risks being seen as a top-down mandate rather than a capability to leverage. Leadership must make that transition viable by creating the conditions for employees to trust the AI tools they're handed.
More autonomy requires more visible control
Hewitt compares the current moment to the early days of bring-your-own-device policies. Organizations once worried about employees accessing company information from personal devices. Over time, technology, policy and governance matured around that behavior.
AI presents a larger version of that challenge. The issue is not only which device can access sensitive data, but where the organization’s accumulated knowledge goes and what an AI system may be able to do with it. That makes transparency and guardrails central to adoption and maintaining employee trust in the technology.
Employees and administrators need clear answers to three questions:
-
What is the AI system doing?
- Why did it make that recommendation?
- What happens when it is wrong?
For higher-risk actions, they also need the ability to approve, stop, or roll back what the system does. Those controls are not barriers to AI adoption. They are what make greater autonomy possible. The MIT NANDA report on agentic AI similarly points to a tension between control and capability:
- Constrain systems too tightly, and organizations limit their value.
- Give them too much freedom and unpredictability rises.
Organizations need processes that let AI systems scale while maintaining control and learning along the way.
Trust should be designed in, not repaired later
For MSPs and other technology partners that advise customers on AI tools, this creates a practical shift in how AI should be positioned and deployed. The old model is to lead with capabilities around automation, speed and reducing manual work. Those outcomes still matter, but customers also need to understand the boundaries.
That means starting with specific workflows, defining what AI can and cannot do, establishing human approval points and enabling rollbacks before expanding autonomy. It also means being honest about where the technology is still learning.
The ROI challenge reinforces the point. Most organizations in the MIT NANDA research had yet to convert GenAI investment into measurable P&L impact, even as experimentation remained widespread. More tools alone will not close that gap.
Trust is one of the operating conditions required to turn AI capability into actual use.
The organizations that get this right will not be the ones that persuade employees to stop worrying about AI. They will be the ones that give employees reasons to trust it, including:
- Clear limits
- Visible oversight
- Useful training
- Meaningful control
As AI becomes more autonomous, that foundation will matter even more. Employees are already using AI. The question is whether they will trust it enough to let it act autonomously.
Learn more about how TeamViewer ONE can help your organization detect issues sooner, automate faster and fix recurring problems to build employee trust.
