We Are Simultaneously Underusing and Overtrusting AI

September 30, 2026

I keep hearing two concerns from organizations that seem impossible to reconcile. The first is that employees are not using AI enough. The second is that employees are trusting AI outputs too much.

At first glance, those concerns sound contradictory. But after spending the last year watching AI move from experimentation into everyday work, I think both are true.

Many users are still employing AI in relatively basic ways. They're drafting emails, summarizing meetings, rewriting documents, or brainstorming ideas. Those activities are useful, but they're also fairly modest.

Most employees haven't yet built repeatable workflows, reusable prompts, personal knowledge systems, or department-specific processes that fully leverage what these tools can do.

In that sense, we're underusing AI.

At the same time, AI has become remarkably persuasive. A well-written response arrives instantly, presented with confidence and polished language. It's easy to mistake confidence for accuracy.

The risk is in humans becoming less critical because the answers arrive so quickly.

I've seen people accept summaries without checking the source material. I've seen AI-generated analyses forwarded without validation. I've seen organizations focus heavily on training people to prompt effectively while spending relatively little time teaching them how to evaluate what comes back.

In that sense, we're overtrusting AI.

The interesting thing is that most conversations about AI focus on prompting. Prompt engineering has become a skill people actively discuss and develop.

I'm not convinced that's where the long-term value lies.

Prompting becomes easier every month. The tools keep improving. Interfaces get simpler. Templates become reusable.

Judgment remains difficult.

The professionals who create the most value won't necessarily be those who generate the most AI content. They'll be the ones who know when an AI response deserves trust, when it deserves skepticism, and when it deserves further investigation.

The emerging divide isn't between AI users and non-users. It's between thoughtful users and passive users.

Passive users ask a question, receive an answer, and move on. Thoughtful users challenge the answer, verify assumptions, improve their process, and build systems that compound value over time.

Organizations shouldn't aim for more AI usage. They should aim for better AI usage.

The future doesn't belong to people who reject AI. It also doesn't belong to people who blindly trust it.

It belongs to people who can do both: use AI aggressively and think independently.

That may end up being the most important professional skill of the next decade.