Everyone bought AI. Few turned it into culture.

The math almost every company did in 2025 was simple. Buy the license, hand it out to the team, tell people to use it. The next quarter someone asks what changed, and the answer comes back as anecdotes. One person saved an afternoon. Another wrote an email faster. Nobody can point to a line in the results that is different.

The numbers match the scene. In McKinsey's 2026 survey, almost nine in ten companies use AI day to day. Only 37% say AI has already contributed to EBIT. And only 6% attribute 5% or more of their results to AI. In Brazil the picture is of a door that has barely opened. Cetic.br measured 17% of companies using some kind of AI in 2025, up from 13% the year before.

The distance between the nine in ten and the 6% lies in what the company did after buying.

What comes with the license

AI culture is the team knowing why it uses AI, where it does not, and who decides when the machine gets it wrong. The subscription does not bring that.

The subscription brings access. And access alone has a known effect. Each person finds their own way, nobody tells anyone else, and the knowledge dies on the computer of whoever found it. In a survey by Writer and Workplace Intelligence of professionals in the United States, 31% of employees admitted to sabotaging their company's AI strategy.

When people do not understand the reason, they do the task and hand back the minimum. With AI that gets worse, because the minimum looks productive. The report comes in on time and the deck looks good. Nobody checks the standard.

The two axes

There is a horizontal axis, which is the download of knowledge. Prompts, tools, automation, what the machine does. It installs fast, and it is what almost every training sells.

And there is a vertical axis, which is judgment. Knowing what is worth doing, what is not, and what the quality bar is when the answer comes back polished and wrong. This axis develops over time, with people looking at real work.

One axis without the other makes a strip. Both together make an area.

A team with plenty of tools and little judgment produces more of the same, faster. A team with plenty of judgment and few tools debates well and delivers slowly. Both cost money. Neither shows up on the chart as an AI problem. It shows up as an expensive team.

The mistake of buying at the wrong level

The Esquadro method has three levels. The first is the person learning to work with AI in their own flow. In the second, the team shares what works, agrees on what to automate and reviews mistakes together. The third changes how the whole company works, with policy, dashboards, onboarding and an audit trail.

Most companies are at the first level and buy as if they were at the third. They book the talk on autonomous agents before the team has agreed on how to review a text the machine wrote.

McKinsey itself shows the path. Among the 6% that capture results, almost three in four redesigned their workflows because of AI. Among the other companies, one in four. Process redesign is the third level. Nobody gets there by skipping the first two.

What to do on Monday

Three moves, in order.

First, pick one area and a sponsor with a name. Culture comes in through a place where someone answers for the result. Then it spreads through the company.

Next, make visible what already happens. Someone on the team already uses AI well. Ask that person to show how, in front of the others, with real work on the screen. That is worth more than a course.

Last, agree on the standard before scaling use. What the machine can decide alone, what goes through review, what never goes out without a person. Write it on one page and put it where the team works.

All three fit in the budget the company already has. What they ask for is someone who stops to design how AI works in there.

What stays

The technology keeps moving. According to METR, the length of task an AI agent can complete on its own has doubled every seven months since 2019. In the 2024 and 2025 models, it doubled even faster. The distance between those who designed and those who only bought grows on its own, every six months.

The work that closes that distance is human, and it is short. It is deciding where AI comes in, agreeing on the standard and teaching the team to build on top of it.