IT pros in AI implementation face challenges with legacy systems—and opportunity

Money keeps pouring into AI, but the investments are running up against the reality of difficult integrations with legacy systems.

That’s one of the main takeaways from new research by IT services firm GFT Technologies. The study, entitled The Future of AI, Unfiltered, was released on September 29 and delves into the complexity of putting AI into practice.

GFT CEO Rishi Chohan told IT Brew that legacy applications can range in age from two to 50 years old, paired in many cases with complex mainframe architectures.

“They have a lot of silos, they have a lot of disintegrated stuff going on,” Chohan said. “That poses many limitations in terms of scaling AI, in terms of implementing AI.”

Run the numbers. The survey, which included 945 CIOs and CTOs across 19 countries, shows a near-universal belief (93% in favor) that AI implementation without legacy system updates is likely to lead to an organizational security crisis. Concern goes beyond systems: 89% said they were afraid that global investment in AI could be leading to a bubble.

Executives reported taking action to manage the uncertainty, with 84% of respondents saying they canceled at least one AI implementation project due to legacy system limitations. When it came to delays, numbers were even higher: 50% said they saw a moderate delay, 39% a minor delay, and 6% a major delay—leaving only 4% with no delay reported at all.

Chohan warned IT Brew not to look too deeply into those numbers, because the survey findings are open to some interpretation.

“It’s a number where, the way I’m reading it, some legacy projects would be on track,” Chohan said. “The legacy ecosystem might be very limited.”

Irrespective of the details, legacy systems come up as a concern when it comes to enterprises deploying AI. In a July survey of 214 IT pros, IT Brew found that 34% reported that legacy systems were an impediment to implementation.

“You see those older systems that are rigid, and you want a fluidity when you use AI,” Brian Luckey, CIO and CTO at Integris, told IT Brew at the time.

Moving on. Despite the very real worries about how to quickly implement AI, there’s an opportunity provided by revamping. To Chohan, making sure core systems are up to date is a good way to make sure that you’re using AI in the most capable way.

“You have to start in the right way, make sure you’re designing them in the right way,” Chohan said. “You have a complete, holistic view of how you’re designing and building that; the short answer is you will modernize the ecosystem and foundations to scale AI.”