Scaling enterprise AI: from pilots to real-world impact

One topic I often discuss with clients and technology leaders across industries is how to unlock value through artificial intelligence applications. And you don't have to be particularly creative to see the benefits they can deliver. But creating a meaningful competitive advantage for each specific business—and being able to scale it through technology—is a much greater challenge.

In fact, the evidence shows a gap between experimenting with AI and generating real, tangible value from it.

A McKinsey study finds that 88% of organizations already use AI in at least one business function, yet only 39% report measurable impact, and only a small fraction are capturing significant benefits at enterprise scale.

So, what happens in that space between pilots, proofs of concept, and isolated projects?

Organizations commonly measure their AI progress through metrics such as the number of models developed, algorithmic accuracy, or the volume of data and processes involved. All of these are valid metrics, but they are no longer enough to capture the full picture.

The phenomenon of “a thousand flowers blooming”—with innovations, systems, processes, and new ideas emerging from the explosion of Artificial Intelligence—has also resulted in disconnected experiments and a lack of coordination across the experience and deployment of these applications.

Moving forward requires shifting the focus toward business metrics. Gartner recommends new metrics to complement traditional technical measures with indicators tied to revenue growth, cost reduction, customer experience, operational efficiency, and talent retention. In other words, AI success should not be measured by what technology can do, but by what it actually enables the business to achieve.

• Sales conversion rate: measures how much AI contributes to turning opportunities into actual sales.
• Collections efficiency: evaluates how AI helps accelerate and improve payment recovery.
• Labor cost optimization: reflects savings or better use of human resources enabled by automation.
• Time to value: measures how quickly an AI initiative delivers business results.
• eNPS (Employee Net Promoter Score): indicates AI's impact on employee satisfaction and experience.

Technology alone cannot scale if every part of the organization is not aligned around a shared view of information. The transition to real-world impact requires AI to align with each company's production, operational, and decision-making processes.

According to the McKinsey study mentioned above, high-performing companies are three times more likely to fundamentally redesign their workflows to integrate AI. These AI high performers achieve more than 5% of EBIT (earnings before interest and taxes) attributable to the use of the technology.

The difference lies not only in adopting AI, but in creating an ecosystem where these tools can evolve, become more powerful, and be deployed with human involvement.

In this context, having modern, flexible, and secure infrastructure should be “non-negotiable”. It is a basic requirement. This foundation enables organizations to train more reliable models, integrate processes, and drive innovation at scale while building effective and secure AI models.

A concrete example is FS (Fueling Sustainability). By implementing SAP Technology Platform (SAP BTP), the company enabled a new way of managing ethanol sales, addressing more complex processes with greater efficiency and less manual intervention through technology. Success also requires senior leaders to demonstrate active commitment and serve as role models in the adoption of new platforms.

Data quality is just as important as data availability. When information flows in an integrated way across teams and systems, AI can operate with greater accuracy, context, and ability to generate impact.

The raw material for success is clean data that reduces bias in models and also enables the generation of synthetic data to increase the volume androbustness of the models those data will support. In addition, vector databases allow systems to capture context and relationships across data from different sources.

Addressing changes in talent and culture is not a minor consideration. To achieve the synergistic effect of technology and people, organizations cannot assume that even positive changes will happen automatically. They need to be actively supported and intentionally managed. Experiences such as Comfama's implementation of SAP Analytics Cloud highlight the importance of actively managing organizational culture to ensure effective adoption of new tools.

A Deloitte study indicates that, globally, 84% of international companies have not redesigned jobs or operational workflows to integrate AI. This is where fusion skills come into play: the collaborative potential created when human capabilities and technology work together, including:

• The ability to strategically interact with AI systems
• Critical judgment to interpret and validate results
• Continuous learning through a two-way relationship between people
   and technology

Developing these capabilities should be a priority for building an agile culture.

In a mature data model, reinventing the value chain represents the most advanced stage, where AI tools can become central business enablers, fostering collaboration across teams, partners, competitors, and suppliers. Reinvention requires moving beyond organizational boundaries, creating shared data and knowledge environments that support growth.

While many companies are focusing on using AI to reduce costs, those truly looking to unlock its value are setting growth and innovation goals based on the capabilities the technology provides. What does this vision look like in practice? It means being able to solve challenges at unprecedented speeds while expanding the boundaries of what is possible for the business. It is a quantum leap, much like the one taken by São Salvador Alimentos, which placed a development ecosystem at the heart of its operations, embedding it throughout the company's structure.

1. Define ownership for every AI-enabled model and process.
    Without clear ownership, there can be no scalability.
2. Establish business metrics from the outset.
    While metrics that measure specific AI capabilities are important, every initiative should answer one fundamental question: “What business objective are we trying to address?”
3. Start by applying AI to repetitive decisions and high-volume processes.
    This is where the fastest and most measurable returns are often found. It is also important to start with a manageable, agile task that can generate cultural enthusiasm for AI across the organization. The value of a quick win should not be underestimated.
4. Ensure data quality, availability, and governance.
    Only then can organizations generate fast, responsive results that can support critical decision-making. No AI strategy can overcome the limitations of inconsistent information.
5. Think about scaling capabilities rather than isolated projects.
    The goal should be to build platforms, methodologies, and models that understand each context, adapt to its needs, and bring innovation to every challenge.

The coming years are likely to mark a turning point for the industrialization of AI, as the gap between explored potential and real-world value begins to close through the strategic convergence of technology, people, and governance.