As organizations raced to deploy AI, the technology outran oversight capabilities. Many are now grappling with basic governance issues that track all the way to the top, where organizational leaders have become prime suspects in the hunt for shadow AI, according to a two-part TrustedTech report.
The Microsoft cloud solutions provider analyzed responses from 2,001 UK and U.S. employees surveyed by Censuswide in March. More than three-quarters of decision-makers said they felt confident using AI effectively in their roles. Only 43% of employees below the decision-maker level said the same, per the report’s second installment published Tuesday.
Meanwhile, 36% of respondents said they were primarily self-taught in AI, while only 23% credited employer training. Among U.S. respondents, 46% said their organization lacked adequate training on safe and secure AI use. Across the full sample, 41% said they lacked clear workplace guidance on AI, and 48% said employers should bear primary responsibility for upskilling.
The findings, TrustedTech said, point to a readiness problem, not a worker problem.
“Organizations have prioritized deploying AI tools over preparing the people who use them,” TrustedTech founder Julian Hamood said. “The workers most at risk of displacement are receiving the least investment and that is not a coincidence. The confidence gap our data reveals is not a talent problem. It is a leadership failure, and it has a solution.”
The research helps explain an eyebrow-raising data point that surfaced in the first part of the report, published in May. Executives are disproportionately responsible for the spread of shadow AI. Nearly two-thirds of decision-makers acknowledged using unapproved AI tools, more than double the proportion among lower-level employees.
“The first announcement from our research revealed that Shadow AI isn’t primarily being driven by junior employees experimenting with ChatGPT, for instance,” Andy Nolan, VP of technology at TrustedTech, told Channel Dive. “It’s being driven by senior leadership at the highest levels: the people making strategic decisions while often operating outside established governance.”
The latest installment of the report sheds further light on the problem. Companies have moved aggressively on tools, but not on the training, oversight and cultural groundwork needed to make sanctioned AI adoption workable across the business, Nolan said
Governance first
The standard partner playbook — deploy a platform, publish a policy and train end users — is insufficient for AI.
“Our research shows executives are often the earliest adopters of AI, but also among the most frequent users of unapproved AI tools,” Nolan said. “That’s not because they’re intentionally ignoring security. It’s because they’re trying to move faster than the governance process allows.”
In many organizations, he added, leadership sets the cultural norm for AI adoption before formal policies exist. For partners, that means implementing top-down oversight.
“Executive teams need to understand both the opportunities and the risks because their behavior sets the tone for the rest of the organization,” Nolan said. “If executives aren't part of the governance conversation from day one, they're often the first users to unintentionally work around it.”
Organizations need to build governance before or as they begin deployment.
“If Microsoft Copilot is rolled out before organizations establish data access reviews, acceptable use policies, and governance guardrails, shadow AI behaviors often continue, at all levels, even after approved AI tools become available,” Nolan said.
Employees should be trained based on role rather than title.
“Frontline and mid-level employees — the people whose roles are often most susceptible to AI-driven disruption — are receiving the least organizational support,” Nolan said. “Different departments, different workflows, and different risk profiles require different guidance and practical use cases.”
Beyond licensing
Workers aren’t rejecting AI. TrustedTech found they are using it because it works. Seven in 10 respondents said AI had a positive impact on performance, while more than half reported saving at least three hours per week using AI tools.
Restrictions on AI use become problematic in that context. More than one-quarter of respondents acknowledged they would continue using AI even if their workplace banned it. More than one-third of decision-makers said the same.
Channel partners with AI adoption expertise face a significant opportunity that goes well beyond deployment and licensing. Customers want governance frameworks, role-based adoption plans, security controls, acceptable-use policies, executive advisory services, workforce training and ongoing measurement.
“Purchasing AI licenses isn’t the same as transforming how an organization works,” Nolan said.
Furthermore, partners must measure whether employees are actually using approved tools confidently, securely and in ways that produce business outcomes.
“Customers aren't simply asking, ‘Can you deploy this technology?’” Nolan said. “They're asking, ‘How do we make AI successful across our entire organization?’ That's a much deeper, longer-term, and more meaningful engagement pathway.”
The strongest partners won’t just sell the most AI seats; they'll also enable organization-wide productivity gains while reducing the risks of shadow AI.
Employees want to be more productive and executives are trying to move faster, even as organizations seek to balance innovation with governance. Those priorities don’t inherently conflict, but they do call for alignment from the beginning of an AI launch.
“Organizations don't need more restrictions around AI,” Nolan said. “They need trusted pathways that make secure AI adoption easier than unapproved alternatives.”
