When Executives Say “AI-Generated Data Can't Be Trusted,” Maybe the Real Problem Is That They Have Never Used It

One of my friends works at a company that has recently been running a series of trials with AI tools.

It seems that the people doing the actual work have already encountered a major obstacle.

According to my friend, a fairly senior person made the following statement:

“We cannot trust data produced by AI. We should maintain the traditional system in which people create the work and other people double-check it.”

I understand the concern.

AI hallucinates. It can produce errors that sound entirely plausible.

I understand why that makes a company cautious about using AI in its operations.

But this executive was not proposing a cautious way to use AI.

The proposal was to preserve the traditional process in which one person creates the work and another person checks it.

In other words, AI would be removed from the workflow entirely.

Then What Is the Trial For?

The idea that a company must either use AI for everything or rely entirely on people is too extreme.

Suppose AI reduces a task that once took ten hours to two, and a person then reviews the output in another two hours.

The entire job would take four hours, saving six hours overall.

The answer is neither to trust AI without limits nor to remove it from the process entirely.

The organization needs to redesign the work according to risk: what AI should handle, what people should decide, and how much human review is actually necessary.

If the old workflow remains completely unchanged, the AI trial itself is a waste of time.

Over drinks, my friend sounded half resigned and said, “If that is what the company has decided, there is nothing we can do. It will simply spend more on labor.”

The people on the front line have no choice but to follow the rules they are given.

Is AI What They Really Don’t Trust?

What follows is only my speculation.

People often say, “AI lies, so it cannot be trusted.”

But is that really the only reason?

I suspect a stronger motive may be the fear of introducing something into the organization that senior leaders do not understand themselves.

By the time people become department heads or executives, they rarely spend their days writing complex Excel formulas, building macros, or processing large datasets themselves.

Many of them probably do not use AI intensively every day either.

In other words, they have experienced neither its usefulness nor its limitations firsthand.

You cannot evaluate something you do not understand.

You cannot take responsibility for something you cannot evaluate.

The safest response is therefore:

“Let’s keep doing things the way we always have.”

If the company adopts AI and a serious problem occurs, the executive who approved its adoption may be held responsible.

If that executive delays adoption, however, they are unlikely to be blamed for it.

From the perspective of the individual executive, opposing AI can be a perfectly rational decision.

Humans Have Always Made Mistakes

There is still something that bothers me.

People have always made mistakes.

Errors in Excel calculations.

Copy-and-paste mistakes.

Emails sent to the wrong recipient.

Numbers omitted during transcription.

These have always been routine workplace problems.

Yet we did not respond by saying:

“Computers cannot be trusted, so we should stop using them.”

We simply said:

“Let’s check the work.”

With AI, however, the conclusion easily becomes:

“That proves AI cannot be trusted.”

I find the difference curious.

The Problem Is Not AI. It Is the Workflow.

People should review AI output. That is correct.

But how much should they review?

Which tasks should be delegated to AI?

Which decisions should remain with people?

Designing that division of responsibility is what matters most.

If a company introduces AI but preserves the old process in which people create everything from the beginning and verify everything through to the end, productivity will not improve.

The problem is not AI.

It is the way the work is organized.

Where “AI Cannot Be Trusted” Leads

Companies need to be cautious.

It makes sense to start small and require human review for critical work.

I understand the need to introduce the technology in stages.

But if an organization continues to say, “We do not understand it, so we will keep doing things the old way,” it will never understand AI.

I am somewhat curious to see what will happen to companies that choose not to use it.

Then again, history has already given us the answer.

Companies that refused to use computers.

Companies that refused to use the internet.

Companies that refused to use the cloud.

Their subsequent fortunes give us a reasonable idea of what comes next.

I do not believe AI will be the sole exception to that history.