AI adoption in companies: why 74% still can't scale projects

The phase of doubt is a thing of the past. By 2026, executive committees and boards of directors will no longer debate whether Artificial Intelligence will redesign the work ecosystem. The challenge has now shifted to a strictly operational level: how to transform strategic conviction into real execution capability?

IA June 18, 2026 6 min read By: Skyone

The phase of doubt is a thing of the past. By 2026, executive committees and boards of directors will no longer debate whether Artificial Intelligence will redesign the work ecosystem. The challenge has now shifted to a strictly operational level: how to transform strategic conviction into real execution capability?

To answer this question, MIT Technology Review Brazil, in partnership with Skyone, developed an in-depth study entitled "AI at Work: 20 Insights on Hybrid Teams". The results provide a compelling diagnosis of the current corporate landscape: we are experiencing a clear mismatch between the ambition of the discourse and the reality of the internal infrastructure.

Although 99% of technology and business leaders believe that AI agents will be central to competitiveness in the next three years, the truth is that the vast majority of organizations are trapped in the permanent pilot—projects that impress in isolated demonstrations but fail to scale and generate continuous structural value.

Below, we analyze the main bottlenecks identified by the MIT study with Skyone and how companies can overcome this barrier.

A portrait of the mismatch: high ambition, low ability

The data collected in the research shows that access to technology or lack of familiarity with the subject are not the real current problems. The obstacle lies in coordination and the operational base.

The AI adoption gap in 2026

99% They believe that AI will be central to business in the next 3 years.
74% They are still in the early or intermediate stages of adoption.
59% They do not consider themselves prepared to operate hybrid teams in the short term.
57% They do not have a dedicated budget for Artificial Intelligence.

This scenario of stagnation is not unique to the Brazilian market. The global report The GenAI Divide: State of AI in Business, published by MIT NANDA, analyzed more than 300 public initiatives and identified that 95% of organizations are still not capturing measurable returns from generative AI.

The reason? The extreme difficulty of connecting the algorithm to real workflows, corporate goals, and the systems that support day-to-day operations.

The 3 major bottlenecks that fuel "permanent pilot projects"

For Artificial Intelligence to move beyond being a mere laboratory experiment and become central to operations, leaders need to overcome three invisible barriers:

1. The illusion of the "off-the-shelf solution" and fragmented systems

Many companies buy generic tools from external vendors believing that AI will autonomously understand the operation. However, as Luiz Pecci, IT and Digital Director at Mundo do Cabeleireiro, points out:

"This is a journey of construction, in which the organization needs to teach the AI its business rules, its decision criteria, and its operational context.".

When the algorithm attempts to run on scattered data, isolated spreadsheets, and fragmented legacy systems, the intelligence fails due to a lack of correct inputs.

2. Organizational silos between business and IT

AI doesn't belong to just one department; it requires cross-functionality. However, Skyone research reveals that 40% of companies cite integration between areas as the main challenge to including AI in their processes. In 46% of organizations, business and IT still operate in isolated silos, without a clearly defined collaborative dynamic.

3. Technical weaknesses and local (on-premises) infrastructure

Ambition has grown much faster than data architecture. Only 41% of companies have the cloud as a consolidated foundation for data and AI. The other 59% still operate with partial cloud or mostly on-premises infrastructure, a condition classified by the study as inadequate to support and scale robust Artificial Intelligence initiatives.

The flaw in metrics and the limited focus on efficiency

Another relevant insight brought by the partnership between MIT TR Brazil and Skyone concerns how the market measures the success of implementations.

Outdated metrics: only 14% of organizations use Return on Investment (ROI) as the primary indicator to evaluate their AI projects. This demonstrates that many corporations have not even clearly defined what tangible results they expect to achieve.
Short-sighted view of productivity: approximately 46% of companies invest in hybrid teams (humans + AI) solely to achieve marginal productivity gains and reduced time spent on repetitive tasks. Only 18% consider AI with the goal of designing new products or revenue streams.

Productivity is only the first stage of transformation. When technology is used solely to accelerate an old process, it merely masks underlying technical problems and postpones the necessary reorganization of processes, responsibilities, and leadership.

Skyone: the bridge between strategic intent and real scale

To bridge the gap between isolated experimentation and structural gain, organizations need to modernize their foundations before focusing solely on algorithms. This is precisely where Skyone positions itself as the ideal strategic partner to orchestrate this transition.

Through its integrated solutions, Skyone removes the technical friction that permanent pilots experience:

Skyone Autosky: enables the migration of legacy systems and complex ERPs to AI-ready cloud environments with "Zero Code Change". This ensures the performance, on-demand scalability, and digital security (ISO 27001 certification and Zero Trust architecture) essential for keeping corporate data available and integrated.
Skyone Studio: a unique integrated platform that combines iPaaS (capable of connecting over 400 systems and eliminating silos), Lakehouse for advanced data management, and tools for developing intelligent virtual agents.

With this technical foundation, your company gains the ability to teach its own business rules to advanced programming language models (LLMs), allowing humans and systems to operate side-by-side with maximum governance and efficiency.

Final considerations: the optimized human

As Felipe Wasserman, Marketing and Growth Director at Skyone, aptly summarizes, technological advancement does not diminish the role of people, but rather raises the bar for human reasoning:

"It's not the human being being replaced, it's the human being being augmented by a technological layer that speeds up the operation, but it cannot, on its own, solve what requires sensitivity, understanding, and context.".

In the era of hybrid teams, the market winner will not be the one who simply repeats that Artificial Intelligence is inevitable. It will be the one capable of building the necessary architecture to connect it to real work, converting experimentation into real organizational results.

Access the Special Edition and check out the numbers, analyses, and trends that reveal how companies are evolving their AI, data, and innovation strategy to generate a competitive advantage.