Brazilian companies adopt generative AI quickly, but only 16% measure the return

The gap no board should ignore

A survey conducted by BeCulture, with participation from Confi and the AI Brasil community, published by Exame on November 13, 2026, measured the gap between talk and practice on AI at companies that together posted R$25.2 billion in net revenue in Brazil in 2025. The result: 73% of organizations have already incorporated generative AI into at least one process and 64% say that usage is part of employees' daily routine. Yet only 16% measure this progress with indicators or formal targets.

For whoever approves the budget, this gap matters more than the adoption rate. Spending on AI without a return indicator is a decision of faith, not of management: the CFO has no way to justify the next budget increase, and the board has no way to compare this investment with any other capex line. The survey "AI: From Use to Value" polled professionals from more than 15 industries, and the pattern repeats across all of them: the technology arrived before the ruler to measure it.

Security: high concern, almost no policy

For 88% of respondents, the protection of data used in AI tools is a source of high concern. The problem shows up in the next line of the data: even within this group that says it is concerned, 64% do not confirm having a complete formal policy for data use and protection.

In practice, this means a company running a language model on customer data, contracts, or proprietary code without a document defining what can and cannot go into the prompt. This is direct exposure under LGPD, Brazil's data protection law, and in an M&A due diligence process, it is exactly the kind of gap that makes a buyer discount the price or stall the negotiation.

The most cited obstacles confirm the diagnosis:

• Lack of readiness in processes and organizational culture: 48%
• Data integration problems: 40%
• Lack of security: 31%
• Lack of tools: 13%

The right question for the CTO is not which AI tool the company uses. It is which formal document governs what goes into the prompt, and who signs off on it.

The knowledge that walks out the door with the employee

45% of respondents say that when an employee leaves the company, most or almost all of the knowledge goes with them. Another 46% say preservation is only partial. Just 9% confirm having a structured plan or tool for knowledge retention.

64% say it is hard to find the history of similar decisions already made internally. Only 16% record day-to-day knowledge in a centralized tool: 57% depend on the department or the individual, 21% keep informal records, and 6% record nothing at all.

In short: a company that depends on a person, not a process, to retain knowledge carries key-person risk on its balance sheet, even if nobody calls it that. It is the kind of question every due diligence process asks: what happens if the head leaves tomorrow? According to this survey, nearly half of companies would answer "we'd lose almost everything," and that should weigh on the multiple, not just on the HR conversation.

Structured HR correlates with results, not an organizational-climate metric

85% of participants believe AI will transform their companies' culture. But only 22% confirm having a structured initiative to prepare people, processes, and leadership: 57% are building this now and 21% have none at all.

The data point that separates those who convert adoption into value is this: at companies with a structured People department, the average rate of implementation of AI-focused practices reaches 78%. It drops to 42% at organizations where HR is nascent.

That number drops to 42% at organizations where HR is classified as nascent. This points to a clear relationship between HR and the success of the technology, reinforcing the idea that People and Technology departments need to work even more closely together in this new era of organizations.

Renato Teixeira, director at BeCulture

Rodrigo Righetti, head of AI Brasil, sums up what is missing beyond access to the tool:

This involves establishing guidelines for use, preparing teams, and defining responsibilities, so that the technology can be applied consistently and securely.

Rodrigo Righetti, head of AI Brasil

And Pedro Chiamulera, CEO of Confi, frames the problem that comes before the tool:

We need to understand how artificial intelligence is contributing to the business, what results it delivers, and what is missing for its potential to be realized. With this report, we want to broaden this discussion and show that generating value depends not only on adopting the technology, but also on embedding it into operations.

Pedro Chiamulera, CEO of Confi

What the board should ask before approving more AI budget

Four questions separate spending from investment, according to what this survey reveals:

1. What indicator measures AI's return today, and who owns that number?
2. Is there a formal data policy for AI use, signed by someone with the authority to do so?
3. What happens to critical knowledge if the key person leaves tomorrow?
4. Is the People department on the committee that decides on AI, or only Technology?

None of these questions requires knowing how a language model works internally. They only require the same governance yardstick that any other investment front already needs to have. Before approving more AI budget, the CEO should demand from the CTO and HR the same indicator they would demand from any other line of the business: without it, 73% adoption is just untracked spending, not strategy.