Legacy systems become the new obstacle to deploying AI in businesses

The race to incorporate artificial intelligence into businesses is leaving companies with a conclusion that is less eye-catching than the launch of new models, but potentially more important: an organization's ability to take advantage of AI depends, to a large extent, on how prepared its technology infrastructure is.

A new study by GFT Technologies surveyed 945 CIOs and CTOs at companies with annual revenues of more than US$500 million across 19 countries, highlighting the extent to which legacy systems have become an obstacle to AI initiatives.

The most striking finding is that 84% of the technology leaders surveyed said limitations in their legacy systems had already led their organizations to cancel an AI pilot or project. In other words, in many cases, the problem is not that AI technology does not work, but that existing infrastructure is not prepared to support it.

The problem starts before AI is implemented

Over the past few years, companies have incorporated generative AI models, copilots, automation and other AI applications into their operations. However, these tools need to connect with information, applications and processes that in many cases were built long before generative AI existed.

A bank, for example, may have customer data in one system, financial transactions in another, inventories on a different platform, and certain processes still running on software developed decades ago. Each layer can have its own access rules, data structures and security mechanisms.

"AI adoption is advancing rapidly, but our research shows that companies are increasingly recognizing the importance of the technological foundation that supports it. For many organizations, legacy infrastructure has become a real constraint, not only for innovation but also for security and scalability," said Marco Santos, Global CEO of GFT Technologies.

AI can be added on top of that architecture, but that does not necessarily mean it can make efficient use of it.

GFT's study shows that 93% of the CIOs and CTOs surveyed believe that failing to modernize infrastructure before running AI on legacy systems will eventually lead to an enterprise-scale security crisis.

This makes technology modernization more than just an efficiency project. For companies bringing AI into critical processes, it could become a condition for reducing risk.

The return on investment is also under pressure

The infrastructure problem comes at a time when companies are also beginning to question how much real value they are getting from their investments in artificial intelligence.

According to the research, 89% of technology leaders are concerned that global investment in AI is growing faster than the business value the technology can generate.

The concern is relevant because much of the conversation around enterprise AI has focused on how much companies are spending on models, computing infrastructure and tools. But technology spending alone does not guarantee results.

If a company needs to carry out multiple manual integrations, maintain incompatible systems or move information between platforms before a model can use it, the cost and complexity of each project increase.

As a result, the next stage of AI adoption may be less about adding more tools and more about creating the conditions that allow existing tools to scale.

The research points precisely in this direction. GFT argues that companies generating value from artificial intelligence are approaching the technology as part of a broader transformation that includes infrastructure, data governance, security and business strategy.

"The pressure to show results from AI is real, but moving faster than your own infrastructure simply pushes the risk down the road. Building the foundation first is what makes the value and speed that follow sustainable and scalable," Santos said. GFT

This does not necessarily mean replacing every legacy system within an organization. For large companies, doing so immediately can be too costly and risky. Modernization can involve migrating specific applications to the cloud, updating architectures, improving integrations or turning legacy systems into platforms capable of exchanging information in real time.

The goal is to reduce the gap between the systems a company needs to maintain and the technologies it wants to implement.

For Latin America, this challenge is particularly relevant. Banks, insurers, telecommunications companies and other organizations across the region have gone through years of digital transformation, accumulating different generations of technology. The arrival of AI adds another layer of complexity to that architecture.

The question, therefore, is no longer simply which AI model a company should use. It is also whether its systems can provide that model with the data, connections and controls it needs to operate securely.

GFT's research suggests that this difference could determine which companies turn their AI investments into results and which end up accumulating pilots that never make it into production.

In a market where pressure to demonstrate results from artificial intelligence continues to grow, the advantage may lie less in adopting the latest tool first and more in building the infrastructure needed to use it at scale.