Artificial intelligence is going through a curious moment. In Brazil, technology is no longer restricted to technical niches and has become part of the routine of millions of people. The search Consumption and use of Artificial Intelligence in Brazilfrom Infobip with Opinion Box, shows that 70% of users use AI to support work, leisure and study activities — and 50% already apply the technology directly in the professional context. At the same time, 69% still only use free versions of the tools, a sign that the market is at an early stage of maturity and monetization.
This advance in use coexists with an equally relevant movement in the corporate world. After an acceleration phase driven by FOMO (fear of missing outor fear of being left behind), the ecosystem enters a more selective stage: projects are no longer driven solely by executive pressure or marketing and start to be charged for a measurable return for the business.
This movement marks a change in maturity. The question stopped being “how to use AI” and became “where does it actually work”. It is in this context that vertical AI begins to consolidate itself as a path that is both promising and challenging for Brazilian companies.
Continues after advertising
If in the beginning the logic was to apply generalist models in any context, practice has shown its limitations. In more sensitive and regulated sectors such as legal, healthcare and finance, accuracy, context and accountability weigh more than speed or “newness”.
Why Most AI Projects Fail
The distance between expectations and execution is still large. The report The GenAI Divide: State of AI in Business 2025from MIT, points out that only 5% of AI pilot programs are able to quickly boost revenue. Recent studies by Goldman Sachs (2024), Gartner (2024) and BCG (2024) go in the same direction: many executives remain dissatisfied with the return on investment of their AI pilots and projects.
This reinforces that the challenge is not in access to technology, but in the way it is applied.
Continues after advertising
Most initiatives are still limited to “packaging” existing models, without real control over the technology or the value chain in which it is inserted. These are solutions that work well in demonstrations, but do not sustain competitive advantage in the long term.
In this new, still nascent stage, it is becoming clear that the AI that brings results is not necessarily the one that trains the biggest models, but the one that controls critical elements: proprietary database, deep integration into the user flow and an application that solves a real problem with superior performance.
The ruler went up very quickly.
Continues after advertising
Infrastructure, data and capital
AI infrastructure connects technology, strategy and geopolitics on a global scale.
Developing AI at scale requires balancing three variables: access to data, computational capacity, and ongoing investment. GPUs remain a critical and scarce resource, and the cost of infrastructure is still one of the main barriers to entry — and survival — for those seeking to develop their own technology.
This scenario reinforces the role of large global platforms. Innovation in AI does not happen in isolation: it depends on an international ecosystem in which cloud, chips, tools and distribution channels operate in an integrated manner. Without this support, innovation will not scale.
Continues after advertising
In practice, AI’s value architecture is already organized into layers: infrastructure at the base, tools in the middle and vertical applications at the top. It is at this top that the most tangible opportunities emerge in the short term — and where a virtuous cycle begins.
Solutions like Forlex illustrate this dynamic by developing vertical AI with a clear purpose: to reduce the overload of the judicial system and return efficiency and, consequently, humanization to the practice of Law. This type of application only becomes viable when connected to a robust and scalable infrastructure, such as that offered by global players, such as AWS (Amazon Web Services), which not only provides computing capacity, but also accelerates the go-to-market and international insertion of these startups.
At the same time, venture capital begins to operate in a more sophisticated way, directing resources to businesses with clear fundamentals: proprietary data, deep integration into the legal flow, superior quality in responses, economic efficiency and access to distribution channels. These factors are decisive in supporting the allocation of resources.
Continues after advertising
When all these vectors operate in a coordinated way, AI stops being an experiment and begins to consolidate itself as a valuable infrastructure — a movement endorsed by contributions from funds such as Vinci Ventures.
The Brazilian difference: adaptation as a strategy
If access to capital and infrastructure is still more restricted, Brazil has a less obvious and potentially more relevant advantage: the capacity to adapt.
In environments with restrictions and scarcity, Brazilian companies were forced to develop more efficient, value-oriented solutions focused on concrete problems. This helps explain why the country has been excelling more in applying AI than in building foundational models.
The Brazilian entrepreneur cares less about creating the next OpenAI and more about solving real frictions in complex sectors.
This dynamic is reinforced by another factor: adoption. AI first became popular in consumption, with high penetration among individuals, before advancing to the corporate environment. This reduces internal resistance and accelerates the incorporation of technology into the day-to-day activities of companies.
From FOMO to criteria for investing in useful AI
This maturity also appears in the way venture capital looks at the sector.
If previously FOMO — the fear of missing out — predominated, a more discerning look is now gaining ground, with greater technical rigor and fundamental analysis. Investors began to better differentiate solutions with structural potential from those that only depend on third-party models.
The discussion stopped being “use AI” and became “where is the moat”. O moator economic moat, is the competitive advantage that allows a company to defend its market and gain share consistently.
Today, five elements tend to define defensible AI:
- Proprietary Data
- Workflow integration
- Superior quality in specific contexts
- Economic efficiency
- Privileged distribution access
Without these pillars, the risk of becoming a commodity is high.
AI leaving the hype and becoming infrastructure — like electricity
Part of the current confusion comes from a mistaken analogy. AI does not behave like the internet or mobile. Its nature is closer to electrification: a transversal technology, which not only creates new markets, but reorganizes all existing ones.
In this scenario, AI tends to stop being a competitive differentiator and become a basic operating condition.
It will not be impossible to act without AI, but it will be economically irrational.
Companies that do not incorporate this layer will lose structural efficiency, competitiveness and speed of response. Technology is no longer optional and becomes “embedded” throughout the production chain.
What lies ahead: consolidation, not collapse
Despite discussions about a possible bubble, the most likely scenario is not collapse, but consolidation.
The market must go through a process of natural selection, with mortality of startups, mergers and acquisitions and a more careful allocation of capital. At the same time, major players continue to invest heavily in infrastructure, in a long-term game that involves a new duality: market and technological sovereignty.
There is, however, a clear sign of maturity: AI revenue begins to leave a closed loop — in which companies buy infrastructure, develop technology and reinvest in the same cycle — and starts to come from applications connected to concrete gains in productivity and efficiency.
Brazilian opportunity: apply better, not compete equally
For Brazil, the challenge is not to lead the race for foundational models, but to occupy a strategic space within this new global architecture.
The opportunity lies at the intersection of practical application, local knowledge and execution capacity. In a scenario where basic technology tends to become commoditized, value migrates to those who solve real problems in depth.
Vertical AI is not just another trend: it is increasingly the natural path to transforming technology into impact and return.
In this game, Brazil may not be the largest producer of infrastructure, but it has all the conditions to be one of the most efficient in transforming AI into highly scalable businesses.
