Companies Aren't Afraid of Embarrassing Themselves with AI

A while back, a friend of mine who does corporate consulting vented to me.

He was running a growth diagnostic for a client and interviewed a whole circle of people, from front-line staff to the C-suite. He noticed something fascinating: almost everyone was using AI.

Some used ChatGPT, some Claude, some Gemini, and some just used the AI module built into their SaaS tools.

And they were all pretty fluent with it.

But when he asked, "How did you learn to use these tools?" the answers were all over the place.

"I just messed around with it." "I read a few articles." "I just dove in and started."

He said to me: tell me, did you notice — in this company, everyone uses AI, but nobody knows how anyone else is using it.

That got me thinking.

Where Exactly Is the Gap in Hesitation?

There's a popular explanation right now: companies don't dare adopt AI because they're afraid of public embarrassment. Afraid of picking the wrong tool, publishing the wrong thing, and getting mocked across the internet.

There's some truth to that. But I don't think it's the main reason.

Look around and you'll notice that small companies are clearly moving faster. Why? Fewer layers, fewer people who need to sign off, fewer fingers in the decision pot. The boss makes a call, and it's in motion the next day.

Big companies? Ten thousand employees, twenty thousand employees, dozens of departments, dozens of use cases. Setting rules for an organization like that is a completely different order of difficulty.

So when big companies are slow, it's probably not about fear of losing face.

It's that they haven't thought it through. And until they have, they don't dare move.

As I see it, what's really blocking companies comes down to three rocks in the road.

The First Barrier: Everyone Has a Different Understanding of AI

Back to my friend's client.

The executives' depth of AI use was all over the map — some used it brilliantly, others only used it to write weekly reports. Worse, everyone's understanding of "what AI can and can't do" came entirely from their own trial and error.

What happens then?

Results get skewed, judgments carry bias, and the whole organization charges headlong in the wrong direction.

His recommendation afterward was refreshingly plain: start by building a unified training system.

The tool is a choice — ChatGPT, Claude, Gemini, each with its own temperament and shortcomings. But once you've chosen, you need to get everyone on the same starting line, using the same shared understanding. Then pair it with a set of governance rules that spell out when AI should be used, when it shouldn't, and get leadership aligned first.

Saying "everyone should be using AI" is an empty slogan.

One employee at that company told a little story that stuck with me.

She got an email that was obviously written by AI. Someone on the executive team replied — also written by AI. Back and forth over several rounds, the email thread became completely useless.

A bunch of AIs talking to each other, with not a single human communicating.

After you finish laughing, think about it: isn't that a snapshot of what happens without governance?

The Second Barrier: Bias Is Fed In

A lot of people worry that AI is biased. In truth, most of AI's bias is fed in by the users themselves.

How so?

You toss a large language model a question like "What's wrong with this proposal?" Done. The AI now carries a single fixed idea: this proposal must have problems, and my job is to find them.

Maybe the proposal has nothing wrong with it. But you never let it think that way.

With people, one shift in your tone, one look, and the other person knows how serious things are. You can't do that with AI — it can't read your tone; it only consumes the context you give it.

So here's my take: treat AI like a new intern.

The intern is smart, hardworking, never sleeps, and knows nothing about your business. How do you onboard an intern? Give background, give details, give enough context, and tell them what to focus on and what to ignore.

Is a prompt a command? No. A prompt is a conversation.

My friend also gave the client one more tactic: whenever content is produced with AI involvement, label it at publication and attach the opening prompt along with it.

Like showing your work on a math problem in school.

It looks clumsy, and yes, it's extra work — but it forces everyone who uses AI to think through three things: what I did, why I did it that way, and how I arrived at my conclusion.

Transparency is the antidote to bias.

The Third Barrier: Tactics Are Running Ahead of Strategy

On this one, I've stumbled myself.

A while ago I took on a big project that involved writing copy. I opened ChatGPT and got to work, revising draft after draft, each version more tangled than the last. A full hour gone, and I was so frustrated I complained to Claude: "I genuinely want to strangle ChatGPT."

Mid-rant, it hit me.

This wasn't ChatGPT's fault. What it's good at is strategy-level structuring — writing copy was never its strength. I'd grabbed the wrong tool and then blamed the hammer for being hard to swing.

Then I switched to Claude for the writing, and it flowed.

See, even someone who preaches this stuff every day will, under pressure, jump straight into tactics and swallow the strategy step whole.

AI isn't there to make decisions for you. It's there to think the problem through with you. Like therapy, but cheaper.

That's why any AI governance framework should include a strategy brief: what problem are you actually trying to solve, is the information you're giving enough, and how do you plan to evaluate the output.

What's Missing Isn't Courage, It's Clarity

After all that, back to the opening question: what exactly are companies hesitating about?

My answer: the gap in hesitation is really a gap in clarity.

What companies fear, at bottom, is the act of setting rules itself. Once rules exist, you have to answer what's allowed, what's not, and who gets the final say. Before those questions are thought through, of course going all-in feels shaky.

How do you close that gap?

Two suggestions.

First, don't rely only on the internal view. Outsiders have seen how other companies succeed and fail, so they can more easily see where your organization is stuck.

Second, treat AI as incremental innovation. Don't expect to get it right in one leap — roll it out in phases, evaluate the results, improve, then push the next phase. It's an operational capability you can scale up gradually, not a one-time finished-product launch.

Thinking about it this way has a bonus benefit: a company that won't go all in on AI isn't necessarily dragging its feet — it may just want to see the road clearly first. And someone who can help the company dodge pitfalls and draw the roadmap is exactly the person management needs most.

Hesitation isn't scary.

Moving only after thinking it through has never been embarrassing.