Open-Source Innovation: The Best AI Model

For some years now, we have seen the conversation surrounding artificial intelligence fluctuate between hypotheses about how best to feed, train, and implement AI models. In some cases, the operating belief has been that building larger models could translate into greater outcomes.

This approach poses a dilemma: how to build a model and how to apply it to the areas that can have the most transformative impact for a business? 

In artificial intelligence for business – AI4Biz – this is an incredibly relevant question, as companies are waging the best strategies to incorporate AI into their operation with the clear objective of generating value by solving logistical, operational, technological, and commercial paradigms in a more efficient way.

This dilemma is well represented in the following example: In the most recent CEO Study focused on artificial intelligence published by the Institute for Business Value, we found that 59% of CEOs in Mexico believe that the companies that onboard the most advanced generative artificial intelligence are the ones that will have a competitive advantage in the near future; however, the same percentage of surveyed CEOs reported an unwillingness to sacrifice a degree of operational efficiency in order to be more innovative and potentially have better results in the longer term.

In short: companies want to innovate but they are unwilling to disrupt their operations and sacrifice their short-term results.

What is the most advanced generative artificial intelligence model?

The answer might not be the same for everyone. Recent developments in the field, including the launch of DeepSeek, suggest that the premise of building an all-encompassing model is not necessarily the most efficient nor the right answer for every business.

At IBM, our approach to AI is rooted in the principle that the best engineering optimizes for two critical factors: performance and cost. We understand that businesses exist on a spectrum, each with unique requirements and capabilities. Therefore, the most advanced AI solution is that which can successfully resolve the specific needs of each user.

Aligned with this, that is the focus with which we have developed our Granite model: leveraging fit-for-purpose architectures to achieve up to 30-fold reductions in AI inference costs. This makes training more affordable and accessible. 

This hypothesis suggests that open-source innovation, paired with the right data and implemented toward specific use cases that have tangible impact for companies, is the new direction for artificial intelligence. In that sense, our focus is on embedding intelligence into data strategies, transforming proprietary information into tangible, measurable returns through purpose-built enterprise tools.

The evidence supporting open-source AI is compelling. According to the “ROI for AI Study” conducted by Morning Consulting in December 2024, 69% of Mexican respondents plan to escalate their AI investments in 2025. Moreover, 45% of these businesses intend to capitalize on open-source ecosystems to optimize their AI implementations. Similar trends are observed in Brazil, where half of the surveyed companies are eyeing open-source AI to optimize their investments, and 78% are set to increase their AI expenditure in 2025.

After considering the shift in the AI conversation worldwide, as well as the challenges that companies consider when deciding how to invest in artificial intelligence, there are a few clear pointers that can be helpful to navigate AI adoption and implementation in 2025: 

Select the most relevant use cases: Understanding the use cases that can have the most impact on a business can be an avenue toward developing and applying AI models correctly, making sure that its application is cost efficient, at the same time that it delivers measurable results. 

Open innovation is key: As we move forward, it's clear that open-source innovation in AI is not just a trend but a strategic necessity. By embracing this approach, businesses can enhance their AI capabilities, improve operational efficiencies, and ultimately, create new avenues for growth.

Design your AI strategy: As companies advance in their artificial intelligence journey, and as they test the use cases that are most relevant for them, they also need to look beyond and begin designing a roadmap that includes the correct storage and management of their data, the process and workflow automation that derives from analyzing said data correctly, and finally, governing that data in a way that builds trust for all its users, within the company and externally as well.

According to the latest AI Adoption Index study, companies accelerated their AI adoption by 67% in Latin America over the past year. It is the right time to move from the hype of AI to the ROI.  This is a prime time for companies in Mexico to begin taking these steps and formulating a tangible route toward AI-powered value generation. 

The future of AI is open, efficient, and accessible. And at IBM, we're dedicated to making that future a reality.