AI’s Biggest Barrier Isn’t Capability. It’s Trust.

As AI takes on a larger role in core business workflows and customer engagement, the challenge is no longer adoption; it’s trust. Even as organizations scale AI to increase efficiency, customers don’t automatically extend confidence in it.

According to KPMG’s 2025 global study on trust, attitudes and use of artificial intelligence, only 41% of Americans are willing to trust AI, while 70% expect it to benefit them. The issue isn’t that people don’t see the value in AI, it’s that organizations are asking them to trust technology they don’t fully understand or see consistent results from. That’s because these tools often sit on top of segmented data, disconnected workflows and inconsistent communications. When those systems aren’t working well together, AI doesn’t build trust; it can actually erode it.

AI Scales What’s Already There

AI doesn’t automatically improve the systems behind it. In some cases, it can expose existing weaknesses within an organization’s internal frameworks. Consider a customer who receives one answer by email, another by SMS, and something different when they speak with a representative. Introducing AI into this process doesn’t solve the underlying problem, it simply accelerates it. These errors frustrate customers and compromise their confidence and trust in the AI processes that promised to make their lives easier.

Since AI is only as reliable as the data and operational systems that feed into it, organizations can’t simply tack AI onto existing operations and expect better results. Fragmented workflows force models to operate on insufficient, outdated or conflicting information. Connecting those systems provides AI with a consistent flow of data to work with, helping it deliver more accurate information to internal teams and consumers.

Customers may never see the processes behind an AI application, but they do see what it produces. If an AI-generated response is wrong or makes their lives more difficult, they lose confidence in the organization and the technology.

Responsible AI Needs Clear Boundaries

Organizations need to define where AI delivers the most value. A recent study found that consumers are concerned about data security and lack of human oversight, with only 40% saying they trust AI to handle their personal data and 47% expressing concern about the latter. Those concerns need to influence how organizations design and deploy AI, not just how they talk about it. 

For consumers to be confident that AI will only be used for their best interest, organizations need to create clear AI governance frameworks that specify exactly what information it can and can’t access, what actions it can take, and when a human will intervene. That doesn’t mean putting a human in the loop every time AI is used, but instead, the focus should be on identifying where human judgement matters and building those points into AI workflows.

When AI operates within defined boundaries, uses customer information appropriately and has clear rules around its actions, organizations have a better foundation for earning trust. 

Transparency Can’t Be an Afterthought

As organizations expand their use of AI, they need to be clear with customers on where and how the technology is used in their experiences. Deloitte’s 2025 Connect Consumer Survey found that consumers often find that technology is advancing without sufficient transparency around it. They aren’t necessarily asking organizations to explain how a model works; rather, they want to know what it’s doing with their information, how it affects their experience, and who is accountable if something goes wrong.

In practice, this means proactively communicating how AI is being used, so people have enough clarity to understand what it does and doesn’t do. These questions are much easier to address when organizations build it into the design of AI systems, because consumers are more likely to rely on AI when they understand how it works.

Trust Will Determine AI’s ROI

Although customers may never see what’s happening inside an AI system, they experience the results. They notice when an AI-generated response is wrong, when they have to repeat themselves, or when they receive different answers depending on where they ask. Those experiences damage trust, but the real problem is what’s underneath: segmented data, disconnected processes, weak governance, or unclear communications.

That is why organizations need to look beyond the AI models themselves when evaluating their success. ROI depends on whether the people who rely on it can trust it, and if it fosters customer engagement. An advanced system that people can’t trust won’t deliver the results the business expected. AI will continue to become more capable, but those capabilities won’t determine how far organizations can take it — the people who depend on it will.