Brazilian companies’ digital maturity still leaves AI in the background

The latest edition of the Digital Mindset Panorama, a survey by consultancy Brivia Group that assesses the digital maturity of companies in Brazil, reveals that despite the constant presence of artificial intelligence in board meetings and CEO speeches, the AI and Efficient Management pillar is the weakest of the seven analysed. In all 15 economic sectors evaluated, companies scored lowest in this criterion, indicating that rhetoric still outweighs practical results.

The study interviewed 350 companies between September 2025 and March 2026 and assigned each a Digital Maturity Score (DMS), a scale from 0 to 100 that classifies maturity into four stages: Critical, Awareness, Optimisation and Excellence. For the first time, the use of AI was isolated as an autonomous pillar, separate from traditional digitisation, which made it possible to identify a previously hidden gap.

Christiaan van Hattem, marketing and business strategist at Brivia Group and one of those responsible for the research, explains that the problem lies in the implementation strategy. “Companies release the tool to everyone, without criteria, and then realise that it has become unsustainable from an investment standpoint, without a corresponding improvement in productivity,” he says. The consumption of tokens—a unit that measures the processing of an AI model and is charged by volume by technology companies—increases when AI is applied in a generalised and unfocused manner, directing efforts toward problems that are too broad and increasing costs without proportional results.

For Van Hattem, the most effective path is the opposite: start with smaller, specific problems, avoiding a generalist approach that increases token consumption and reduces profit margins. In addition, he points to the absence of central governance as a second critical challenge. “We see that there is usage, but governance is still not well structured within companies, because it is very centralised in the hands of the individual employee, and not in a strategy defined by the company,” he observes.

Disparities between sectors

The most striking contrast appears when comparing two segments of the healthcare sector. The pharmaceutical industry leads the overall ranking, with 83 points (Optimisation stage), driven by excellence indicators in brand purpose and analytical use of data—characteristics of a sector more focused on internal research and regulatory rigour. The healthcare sector (hospitals, laboratories and operators), on the other hand, ranks last, with only 54 points in total and just 22 points in the AI pillar, classified as “Very Critical.”

The difference, according to the study, is not a matter of willingness, but of the complexity of the journey. While the pharmaceutical industry deals mostly with internal laboratory processes, the healthcare sector interacts directly with the patient, who often has to repeat their history in each new care channel, without integration between hospital systems.

In February, the Federal Council of Medicine published the first national resolution to regulate the use of AI in medical practice, effective from this month. The rule is a regulatory response to the type of dispersed adoption identified by the research, in which each hospital and clinic adopts tools in an isolated manner, without a common standard.

Internal contradictions and more rigorous self-assessment

The study also revealed internal inconsistencies in companies. Van Hattem describes the pattern: “Sometimes a company says it has a strong brand, that it is recognised, that it creates community, and then, in the next question, disagrees that customer acquisition depends less and less on paid media. Or it says it knows the customer deeply, but does not conduct behavioural research with them.” These contradictions were repeated in virtually all 15 sectors analysed, suggesting that companies’ perceptions of themselves do not always withstand a more detailed analysis of their own numbers.

Fabio Rios, CMO of Brivia Group and also leader of the research, notes that some sectors that performed well in the previous edition seemed to stagnate in this one. “Every time a new wave of technology emerges, it leads the company to a strategic reflection, and sometimes what seems like regression is actually an executive team becoming more discerning about its own needs, because it now understands better what true maturity means,” he explains. Thus, a lower score may reflect a more demanding self-assessment, and not necessarily a deterioration.

When the machine should not replace the human

One of the most delicate issues for companies that automate customer service is defining how far AI can replace human contact without harming the experience. For Van Hattem, the answer depends on the emotional context of the interaction, not the available technology. “To track the delivery of a product I bought online, I don’t care at all about full automation, I’ll type in a number and get the answer. But for a consultation about physiotherapy, if the person on the other end doesn’t have my history or the sensitivity to understand what I’m feeling, the demand simply isn’t met,” he argues.

The advertising market has already experienced this limit. Over the past two years, Coca-Cola used generative artificial intelligence to recreate its classic Christmas commercial “Holidays Are Coming” and faced strong backlash. Consumers described the result as “soulless” and complained that the brand had “ruined Christmas,” precisely because it was a moment associated with human warmth and nostalgia—feelings that the public did not want to see delegated to an algorithm.

Three steps to accelerate AI results

Brivia Group summarises in three areas the fastest way for companies to move from a low score: first, eliminate internal “data fiefdoms,” ensuring that information used for decision-making is not restricted to one department; second, cultivate a culture of controlled trial and error, with isolated environments (so-called sandboxes) to experiment without risk to operations; and third, form a central AI governance committee that prioritises specific, measurable problems, rather than generic solutions applied to the entire company.

According to the researchers, this is the difference between companies that will begin to reap the first financial results of AI in the next research cycle and those that will continue to spend more to produce the same as always.