Why Trust in AI Dropped 35%: What Business Leaders Must Do Now
Table Of Contents
- The Numbers That Should Alarm Every Business Leader
- What Is Actually Driving the Collapse?
- The Shadow AI Problem: A Trust Crisis With a Dangerous Side Effect
- Why This Is a Leadership Problem, Not a Technology Problem
- A Framework for Rebuilding AI Trust in Your Organisation
- The Business Cost of Getting This Wrong
- Conclusion: Trust Is the New AI Competitive Advantage
Why Trust in AI Dropped 35%: What Business Leaders Must Do Now
Imagine rolling out an AI programme across your organisation — months of planning, vendor negotiations, change management workshops — only to discover that employee trust in those tools has quietly cratered by 35% in a single quarter. That is not a hypothetical. It is exactly what the data now shows, and it is happening inside companies that thought they were doing everything right.
In the summer of 2025, Deloitte's TrustID Workforce Index captured something that most executives had not seen coming: a sharp, sustained drop in employees' willingness to trust and use the AI tools their own companies had provided. The collapse was not uniform. It hit harder in certain roles, certain industries, and certain age groups. And it produced a behaviour that is quietly undermining AI return on investment across the board: employees bypassing approved AI tools and using unsanctioned alternatives instead.
This article unpacks what the data actually says, why the trust decline happened, who it is hitting hardest, and what business leaders can do right now to reverse the trajectory. The good news is that trust is not gone. It is recoverable — but only if leaders stop treating this as a communications problem and start treating it as the organisational change challenge it actually is.
The Numbers That Should Alarm Every Business Leader {#the-numbers}
The headline figure comes directly from Deloitte's TrustID Workforce Index, one of the most granular real-time measures of employee sentiment available. Despite increasing access to GenAI in the workplace, GenAI usage decreased by 15%, and the research shows trust in GenAI decreased by 35% — even as overall organisational trust remained stable. In other words, employees did not stop trusting their employers in general. They stopped trusting the specific AI tools those employers gave them.
Deloitte's TrustID Index tracked a 31% decline in trust in company-provided generative AI between May and July of 2025. Trust in agentic AI systems — those that can act independently rather than just recommend — collapsed by 89% in the same window. That second number is particularly striking. Agentic AI, which many organisations are betting their productivity roadmaps on, is precisely where employees are most resistant.
Confidence in AI agents making key decisions dropped from 48% in 2025 to 34% in 2026 — a 14-point decline in a single year. Meanwhile, the confidence gap between executives and practitioners is striking: 81% of CEOs trust AI-driven delivery, compared to just 50% of QA Directors who feel the same. The problem is not that AI is failing technically. The problem is that the humans being asked to rely on it have lost faith in how it is being deployed.
What Is Actually Driving the Collapse? {#what-is-driving}
Trust does not erode for a single reason, and diagnosing it correctly matters. Several forces are converging simultaneously, and leaders who address only one of them will find the others continue to do damage.
1. AI Is Being Deployed Faster Than Employees Can Process It {#deployed-faster}
While regular AI usage jumped among workers by 13% in 2025, confidence in the technology's use plummeted by 18%. As Mara Stefan, VP of global insights at ManpowerGroup, put it: "AI adoption is accelerating, but confidence is collapsing. Workers are being handed tools without training, context, or support." This is a structural problem. When companies race to deploy, they often skip the human infrastructure — the training, the governance communication, the honest conversations about role impact — that employees need to feel safe.
Only 34% of workers have received training in ethical AI use, yet nearly 80% of those who do apply it regularly. This is not just a training gap; it is a trust opportunity. The data suggests that upskilling is one of the most direct levers available to leaders who want to reverse the trust decline — not because it makes the AI better, but because it makes employees feel less exposed.
2. Fear of Replacement Is Not Irrational — It Is Rational {#fear-of-replacement}
The primary reason for frontline scepticism is fear — specifically, fear of being replaced. This perception of AI as a job killer is widespread, deeply ingrained, and spans multiple industries. A survey by Accenture found that nearly 60% of workers expressed concerns that AI would eliminate their jobs.
What leaders often miss is that this fear does not need to be factually accurate to be psychologically real. With 53% of workers concerned that AI tools will make their role feel less necessary, communicating workflow changes becomes just as strategic as the tech integration itself. When a company announces a new AI tool and simultaneously runs a hiring freeze, employees draw their own conclusions — and those conclusions are not about productivity gains.
3. The C-Suite and the Frontline Are Living in Different Realities {#csuite-frontline}
One of the most consistent findings across research this year is the dramatic gap between how leaders perceive AI adoption and how workers experience it. Trust tracks proximity to the decision. Sixty-one percent of workers say they trust senior leaders when those leaders discuss AI, but that figure ranges from 80% among directors and above to just 47% among individual contributors. Another 38% of individual contributors say they're unsure whether they trust leadership on AI.
Employee resentment of AI is the frustration, distrust, and quiet resistance workers feel when AI is introduced in ways that bypass them. It usually traces back to four rollout failures — no clear plan, no employee input, no honest connection between AI and job security, and no reduction in actual workload — rather than to opposition to AI itself. This is a crucial insight: most employees are not ideologically opposed to AI. They are practically opposed to how it is being handed to them.
Generational differences compound this gap. The confidence gap is most pronounced among older workers, with a 35% decrease in confidence among baby boomers and a 25% drop among Gen X workers. As the technology becomes a daily requirement, the most experienced members of the workforce are feeling most left behind.
4. Agentic AI Has Crossed a Psychological Line {#agentic-ai}
There is a meaningful distinction between AI that assists and AI that acts, and employees feel it viscerally. Although managers and employees alike are getting more comfortable using AI in their workflows, those leading adoption efforts generally feel better than frontline workers about AI. Employees reported that they feel greater personal risk from autonomous systems because of their lack of control over how and when AI is deployed.
Although three-quarters of respondents said they use AI at least once a day, 61% said they'd prefer it if AI had no independent action abilities. The implications for organisations investing in agentic workflows are significant. Deploying autonomous agents into environments where the workforce has not yet built foundational trust in basic AI tools is likely to accelerate the very resistance organisations are trying to avoid.
The Shadow AI Problem: A Trust Crisis With a Dangerous Side Effect {#shadow-ai}
Here is where the trust collapse creates a second, less obvious problem. When employees do not trust the AI tools their company provides, many do not simply stop using AI. They go around it.
Employees are still using AI — just not their company's. For workers with GenAI available at work, 43% report non-compliance with their employer's AI policies through use of unapproved GenAI solutions. This matters for reasons that go well beyond IT governance. When employees use unsanctioned tools, they often input sensitive data, client information, or proprietary workflows into platforms that sit entirely outside the company's security perimeter.
Almost half (44%) of US workers are using AI tools without proper authorisation or in inappropriate ways, and 46% also admit to uploading sensitive company information and intellectual property to public AI platforms. For any organisation operating in a regulated industry — financial services, healthcare, legal — this is not a culture problem. It is a compliance and liability problem. The trust gap does not just slow AI ROI; it actively creates risk.
Why This Is a Leadership Problem, Not a Technology Problem {#leadership-problem}
The instinct in most organisations is to treat a trust decline as a communications challenge — publish a FAQ, hold a town hall, record a video from the CEO. These are not useless, but they are insufficient on their own. The research is clear that employees are calibrating their trust in AI tools through the lens of their trust in leadership.
AI adoption in the workplace is growing rapidly, but employees remain sceptical due to concerns over trust and benevolence — both in AI and in leadership. Sentiment around AI reflects broader confidence in leadership, as employees question how these tools will be used and whether they will serve their interests.
Fear, uncertainty, and a lack of transparency create a climate where employees question not just how AI works but why it's being introduced in the first place. Executives see AI as a path to growth. Employees often see it as a threat. Closing that perception gap is not a technology task. It requires leaders to demonstrate, consistently and credibly, that AI deployment is being done with employees rather than to them.
Higher adoption does not mean blind trust. The more workers understand AI, the more discerning they become about where and how to use it. Daily GenAI users are familiar with the current pitfalls and fallibilities of AI, and are accordingly alert to AI's limits. This is actually encouraging. It means that deeper AI literacy, rather than creating complacency, produces more thoughtful and discerning users — exactly the kind of workforce that makes AI deployment more successful, not less.
A Framework for Rebuilding AI Trust in Your Organisation {#rebuilding-framework}
Rebuilding trust is not a single initiative. It requires sustained action across four dimensions simultaneously.
1. Make the plan legible and honest. Employees need to know that AI supports decision-making rather than completely replacing human judgment, especially for important decisions involving hiring, performance, pay, or promotions. This means communicating specifically, not vaguely. Rather than announcing that AI will "improve efficiency," tell teams which tasks will change, which decisions will remain human, and what the timeline looks like. Explore our AI consulting services to understand how structured AI strategy conversations can help leadership teams get this communication architecture right before they launch.
2. Invest in capability before compliance. Employees need real training, expert access, and hands-on time — one-time tutorials are not enough. Recurring touchpoints, including regular sessions on new tools and use cases, a go-to person for questions, and hands-on practice tied to people's actual jobs, are needed to build capability and confidence. Our Business+AI Workshops are designed precisely for this — equipping teams with practical AI skills in the context of their actual workflows rather than generic introductions to the technology.
3. Involve employees before you deploy, not after. The strongest predictor of whether employees accept AI is not the quality of the tool — it is whether they were told what was happening and asked what they thought. Organisations that consult individual contributors before tool selection, and then publish what changed because of their feedback, build a fundamentally different relationship with AI adoption than those that hand down decisions from the top.
4. Build accountability structures that employees can see. Usage of AI and accountability should be driven from the top down, with boards and management answering for AI incidents similarly to cyber events or regulatory noncompliance. Management that cannot effectively speak to the organisation's use of AI risks fostering a culture of non-accountability. Governance frameworks that are visible and explained — not just filed away in a policy document — are one of the fastest ways to signal that leadership takes AI seriously as a shared responsibility.
The most effective approach is an "AI-leader combination," where leadership fosters AI literacy, facilitates open discussions, and frames AI as a supportive tool for career growth. By aligning technological transparency with empathetic leadership, companies can cultivate both cognitive and emotional trust, ensuring successful AI adoption. For leaders who want a deeper dive into what this looks like in practice, our AI Masterclass programme brings together executives and practitioners to work through these frameworks with real organisational context.
The Business Cost of Getting This Wrong {#business-cost}
Trust in AI is not a soft metric. Its erosion has measurable consequences on the bottom line. When employees do not trust their company's AI tools, they underuse them — producing a negative return on what is often a significant technology investment. When they bypass approved tools for shadow AI alternatives, they introduce security and compliance risk. When they disengage from AI programmes altogether, organisations lose the institutional momentum that is essential for AI to deliver value at scale.
Trust in AI is eroding faster than trust in organisations, but organisations feel the impact of that drop regardless, with adoption also on the decline. The compounding effect is significant. A 35% drop in trust does not produce a 35% drop in AI performance — it produces a much larger operational drag once you account for the downstream effects on usage, compliance, and team morale.
Involving employees in the decision-making process and ensuring that every employee gets a fair share of company value matters directly. When employees say they have a voice in decisions that affect them and that they are fairly compensated for the value they bring to the organisation, they are 20% and 60% more likely, respectively, to adapt to AI quickly. The business case for inclusive AI deployment is not ideological — it is financial.
For organisations that want to benchmark where they stand and connect with peers navigating the same challenges, the Business+AI Forum brings together executives from across Singapore and the region to share what is working and what is not in real-world AI deployment.
Conclusion: Trust Is the New AI Competitive Advantage {#conclusion}
The 35% drop in AI trust is a warning, but it is also an opportunity. Most organisations are experiencing some version of this decline, which means the companies that move decisively to address it will find themselves ahead — not just in employee satisfaction, but in the speed and quality of their AI returns.
The data makes one thing clear: this is not a problem that more powerful models, faster rollouts, or bigger AI budgets will solve. It is a human problem, rooted in how organisations communicate, lead change, and share power with the people doing the work. The companies that treat AI trust as a strategic asset — measuring it, investing in it, and building governance structures around it — will be the ones that turn AI from an expensive experiment into a genuine competitive edge.
The conversation about AI in business is shifting from can we deploy it to can our people actually trust it enough to use it well. That is a more important question, and it deserves a more serious answer.
Ready to turn AI from a source of organisational friction into a genuine business advantage? Join Singapore's most active community of business leaders navigating AI adoption — from strategy to execution. Explore Business+AI Membership and get access to peer networks, expert consulting, hands-on workshops, and the flagship Business+AI Forum.