Organizational Culture and AI: Your Company Is Already Using It (Even If Its Plan Says It Isn’t)

Your company’s AI plan says one thing, but your employees do another. In the European Union, one in five companies officially uses artificial intelligence. But when you ask people one-on-one, three out of four people who work with data already use it every day—almost always with tools that no one has approved for them.

That gap between what the organization endorses and what its people actually do is the sign that matters. The question “Is my company ready for AI?” isn’t about the future—it’s about what’s happening this week, often outside of management’s radar. And it isn’t answered with a speech about curiosity and trust, but by observing a few very specific behaviors.

What does it really mean for a culture to be ready for AI?

It’s far less glamorous than it’s made out to be. Most articles on culture and AI urge you to adopt virtues (curiosity, humility, trust) as if they were a software update. It sounds good, but it can’t be measured: no one knows when their company has enough “curiosity.”

There’s a more honest way to look at it. Every organization has two cultures at the same time: the stated culture (what the values, AI policy, and committee presentations say) and the culture in practice (what people actually do when they open their laptops). With AI, those two cultures have drifted further apart than ever before.

The numbers speak for themselves. According to Eurostat, in 2025 only 20.0% of European companies with ten or more employees were officially using AI (13.5% in 2024), and Spain is among those adopting it most slowly. However, a global survey by Microsoft and LinkedIn of 31,000 professionals (2024, including Spain) found that 75% of those who work with data were already using it, and of those, 78% were using tools not provided by their company. These are different metrics—formal adoption versus individual use—but the trend is clear: the actual culture of use is far ahead of the stated one.

AI isn’t waiting for your culture to be ready. It’s already here; the question is whether you can see it.

Whether a culture is prepared does not depend on its values, but on four observable factors:

1. You could say that AI has made mistakes—even regarding a boss’s work—without any consequences.
2. Drug use is visible and shared, rather than hidden.
3. A distinction is made between what AI proposes (and someone verifies) and what AI decides on its own.
4. It’s the person who knows the process who decides, not the person who signed the license.

Can anyone say, “This result from the AI is wrong,” without it costing them anything?

It depends on whether pointing out a mistake is a plus or a minus in your company. Generative AI gets things right a lot and makes mistakes with complete confidence: it writes false information with the same certainty as true information. That’s why the first sign of a prepared culture isn’t enthusiasm, but something more uncomfortable: that anyone can stop a deliverable and say, “What came out of the tool is wrong,” even if their boss used it, even if management approved the tool.

When that’s not possible, the predictable happens. The text generated by AI makes its way to the client without anyone thoroughly reviewing it, because questioning it seems like questioning the person who sent it. The Microsoft survey gets to the root of the problem: more than half of those who use AI avoid admitting that they use it for their important tasks, for fear of appearing replaceable. And those who hide how they work cannot review that work.

The cost: An AI error that reaches the customer because no one dared to review it ends up being much more expensive than the time saved in generating it.

Correction: Separate criticism of the result from criticism of the person. Make reviewing what the AI produces a normal part of the process, not a sign of distrust toward anyone.

Do your friends share their prompts, or do they keep them to themselves?

Take a look at what happens with prompts that work. In a well-prepared culture, someone who has figured out how to summarize a contract in thirty seconds shares it at the next meeting; in one that isn’t, they keep it to themselves as a personal advantage. That seemingly minor detail determines whether AI benefits one person or the entire company.

We’ve already seen the underlying issue: most people who use AI do so with tools the company hasn’t provided them, and many hide this fact. It’s not a discipline problem—it’s a design problem. When usage happens behind the scenes, the organization can’t build on what already works, identify where it’s risky to enter sensitive data, or learn from its own successes. You have adoption, but it’s invisible—and what’s invisible can’t be governed.

Start with: a safe, non-judgmental space where people can share the prompts that work for them—as well as the ones that didn’t. Not a twenty-page policy; a channel where sharing pays off and hiding doesn’t.

Do you distinguish between what AI suggests and what AI decides?

Almost never, and that’s exactly the blind spot. Not all AI tasks are the same: in some, the machine makes a proposal and a person checks it before the result is released (a draft, a summary, an initial analysis); in others, the AI makes the decision and its output is used as is (an automated response to a customer, a score, a ranking). Let’s call them verifiable tasks and tasks of faith.

A prepared culture isn’t one that uses the most AI; it’s one that knows which of the two categories each task falls into. The problem arises when a task is assigned to AI without anyone having decided to do so: the AI has been responding, scoring, or filtering on its own for months, and the company only finds out the day it fails. Drawing that line is an act of culture, not technology: it forces us to ask, task by task, how much it costs for the AI to make a mistake—and who is held accountable if it does.

Deliverable: a list of your AI use cases organized into two columns: “suggests and someone verifies” and “decides on its own.” The second column should only include cases where a mistake is inexpensive to correct. If there’s anything costly in that column, either move it to the first column or assign someone to monitor it.

Who decides what to do with AI: the person who understands the process, or the person who purchased the license?

More often than not, it’s management that decides who signs off on the bill: management approves the license, and with it comes the idea of where AI “must” be used. But the person who truly understands a process—where it gets stuck, which step is unnecessary, and which exception throws everything off—is usually the one who works on it every day, not the one who bought it.

When decisions are made far removed from the actual work, the process is automated as-is—flaws and all—and AI merely speeds up something that was already flawed. A well-prepared organization brings the people who actually do the work to the table before making any purchases, and adheres to an unassuming rule: if a process is broken, fix it first, then automate it. Automating what’s broken only multiplies the problems.

Question for your committee: Of the last three AI projects you decided to undertake, how many of them had the person who does that work on a daily basis in the room? If the answer is “none,” you know why some of them didn’t quite work out.

Quick Quiz: Is Your Usage Culture Ahead or Behind?

Answer yes or no based on your actual team, not the one you’d like to have:

1. If you ask who uses AI for work, almost everyone says they use “something,” but no one shows you a specific prompt they use.
2. There is an approved AI policy, but you wouldn’t know which tool the person sitting next to you uses every day.
3. The last time something went wrong with AI, the conversation focused on finding someone to blame, not on learning from it.
4. You held an AI training session, and no one has used what was taught since; the knowledge stayed in the room.
5. There is no task for which you can say with certainty whether the AI is merely making suggestions or making decisions on its own.
6. The last thing you automated using AI was decided by someone who doesn’t do that job every day.

Zero or one “yes”: your usage culture is healthy; the challenge is to maintain it as you grow.

Two or three: Your stated culture is ahead of your actual practice. You talk about AI, but its actual use is scattered and obscure. Start by making it visible.

Four or more: You have AI operating behind the scenes that you don’t control. You don’t lack technology; what you lack is the ability for people to use it, make mistakes, and talk about it without facing consequences.

What if your company is already better prepared than its plan acknowledges?

It probably is. AI is already here: your people are using it, in their own way, often better than management realizes. Being prepared isn’t about installing a new culture, but rather bringing the one that already exists to light: making its use visible, verifying what matters, letting people make mistakes and talk about them, and starting in areas where mistakes are inexpensive to fix.

That can’t be fixed with a talk about curiosity or a twenty-page policy. It’s built by using AI on real work, on a small scale, where the results can be verified, and by observing what behaviors emerge. A prepared culture isn’t a prerequisite that must be established before getting started; it’s what remains when the tool is used properly.

So the question isn’t whether your company is ready for AI. It’s more specific: of the four signs mentioned in this article, how many would you see today if you took a look at how your team is working this week?

If you want to move beyond talk and put these habits into practice, at SmartWay we apply our expertise to real-world cases through training and development, and when the underlying organizational structure is the problem, we provide organizational consulting. Reach out to us via the contact form, and we’ll take a look at your specific situation. 🚀