The Problem With "Trusted AI"

Do you trust people?

Most of us would instinctively say yes. But if that were entirely true, organizations wouldn't have background checks, approval limits, audits, or dual signatures on large expenditures. No one interprets those controls as evidence of distrust. They exist because trust has never been unconditional. A controller may prepare the financial statements, but someone else reviews them before they're released. A purchasing manager may be trusted to buy office supplies, but not to authorize a $20 million acquisition.

Trust has always been contextual. We trust people to perform specific responsibilities under specific circumstances, with safeguards appropriate to the level of risk.

Which brings me to AI.

A recent WalkMe study found that only 9% of employees trust AI to make complex, business-critical decisions. Among executives, the figure is 61%. That's a 52-point gap between the people using AI and the leaders championing it.

So what's driving that gap?

Honestly, we don't know. And that's the interesting part.

The survey asked about "trust" without defining it, which means every respondent answered their own version of the question. An executive might have heard "Do you believe AI can make your organization more productive?" An employee might have heard "Would you stake your reputation on this output?" Those are entirely different questions, and a single word, trust, absorbed them both.

Which means the 52-point gap may not measure a difference in trust at all. It may measure how many different questions hide inside that one word.

And it's not just surveys. Organizations everywhere talk about "trusted AI." Vendors promise it. Regulators encourage it. Boards ask about it. Yet I rarely hear anyone define what they actually mean. We keep asking "Can we trust AI?" as though it has a yes-or-no answer. I don't believe it does, because AI isn't one thing. We may trust AI to summarize a meeting but not to negotiate a contract. We may rely on it to flag unusual transactions while refusing to let it approve payments. Those aren't contradictions. They're examples of sound management, the same contextual judgment we've always applied to people.

So if trust is too vague a word to measure, or even to discuss productively, what should we be asking instead?

I'd suggest this: Do I have sufficient confidence to rely on AI for this particular decision?

Confidence isn't based on hope, impressive demonstrations, or marketing claims.

Confidence is earned through evidence. Can the output be validated? Can exceptions be identified before they become business problems? Can someone explain how the recommendation was developed? Is there meaningful human oversight?

Those are not questions about artificial intelligence. They're questions about decision governance. They're the same questions organizations have asked for decades about financial reporting, cybersecurity, and every other area where important decisions are made. We don't have confidence in financial reporting because accountants are trustworthy. We have confidence because a system of internal controls is designed to produce reliable information and identify problems before they become material.

AI deserves the same treatment.

That perspective should also change how we think about governance itself. Too often, governance is presented as a way to make AI trustworthy. I see it differently. Governance doesn't create trust. Governance creates the conditions that justify confidence. It establishes where AI is being used, what decisions it influences, what evidence is required, who remains accountable, and what happens when something goes wrong. In other words, it helps leaders determine when reliance on AI is appropriate and when additional oversight is needed.

It may also be the only way to close that 52-point gap. Not by persuading employees to trust more or executives to trust less, but by replacing an undefined question with answerable ones.

So the next time someone asks whether AI can be trusted, pause before answering.

Then ask a better question: What exactly are we trusting, and what evidence gives us confidence that we should?

About Root Idea

Root Idea helps CFOs protect the business from AI decision risk. Root Idea works directly alongside finance teams to map AI influence, establishes decision governance controls that hold up to board scrutiny, and delivers training and change management to make governance stick. 

If your organization is scaling AI and governance hasn't kept pace, that's exactly the conversation we're built for. Learn more at rootidea.ai.