Modernization remains an unfinished job
Organizations have been trying to modernize old technology estates for decades. Mainframes and other legacy systems have been moved, adapted or surrounded by newer infrastructure through web services, microservices, virtualization, cloud platforms, and x86-based Windows and Linux servers. Yet modernization still has a familiar reputation: costly, slow and perpetually incomplete.
Now vendors are making a new promise. Agentic AI, they argue, could give modernization programs the coordination and automation that earlier waves of technology failed to deliver. But the evidence cited in a ZDNET analysis suggests that the proposition is still early, and that the business case is far from automatic.
The core question is not whether AI can participate in modernization work. It is whether AI can change outcomes that have long frustrated technology leaders: delayed programs, cost overruns, limited operational improvement and unclear business value.
Adoption is still limited
ZDNET cited a survey of 2,000 senior IT decision-makers conducted by IT services specialist Kyndryl. Only 10% of respondents said they were applying agentic AI as a modernization tool. That figure indicates that, despite the intensity of AI marketing, agentic systems have not yet become a standard part of modernization programs.
There are plausible reasons for caution. Modernizing a legacy environment is not a single technical substitution. It can involve applications, operating processes, dependencies and long-running systems that organizations cannot simply turn off. New tooling may help with parts of that work, but it does not automatically resolve the choices involved in changing critical systems.
The low reported level of use also makes broad claims difficult to validate. If only a minority of IT leaders are currently applying agentic AI to modernization, the field has limited real-world evidence about whether the technology can reliably improve delivery, control costs or create measurable innovation gains.
The existing results are uneven
Kyndryl’s findings, as reported by ZDNET, describe a mixed record for modernization efforts overall. About half of organizations undertaking modernization reported better IT operations. Fewer than half reported innovation gains. Nearly one in five respondents, 18%, said their organizations saw limited or unclear value from modernization work.
Schedule and cost performance were also a problem. Almost half of respondents said they were behind schedule and experiencing cost overruns. These figures matter because AI is being introduced into a category of work that already struggles to demonstrate consistent returns.
That history creates a high bar for agentic AI. A new AI layer may make some tasks more coherent or help teams coordinate work, but its value should be assessed against the persistent shortcomings of modernization programs. A successful demonstration needs to show more than a technically impressive agent. It needs to show better operations, a clearer route to innovation, or an improvement in schedule, cost and value realization.
AI is being pitched as a unifying layer
The analysis describes the possibility that AI agents could add cohesion to modernization efforts. That is distinct from treating AI as a simple replacement for older systems. The promise is that agents might help connect work across a modernization program that can otherwise be fragmented among applications, infrastructure and teams.
But cohesion alone does not settle the economics. Modernization programs are often judged by the benefits they produce after implementation, not just by the tools used during the project. The survey results cited by ZDNET underline that previous approaches frequently failed to turn spending and elapsed time into clear operational or innovation outcomes.
The question, then, is whether agents can produce results that were not achieved through earlier strategies. The article frames that uncertainty directly: AI could eventually make a difference where earlier modernization efforts have fallen short, or it could become another promised remedy in a long-running search for a cure.
A practical test for technology leaders
The immediate lesson is restraint. Agentic AI may become useful in legacy modernization, but the reported adoption level shows it is not yet proven at broad scale. Organizations considering it should distinguish between a demonstration of AI capability and evidence that a program’s economics have improved.
- Has the modernization effort produced better IT operations?
- Has it delivered innovation gains rather than only a newer technical architecture?
- Is the organization reducing schedule pressure and cost overruns?
- Can it identify clear value, rather than limited or unclear results?
Those questions are not unique to AI, but they are especially important when AI is presented as the next answer to a decades-old challenge. The Kyndryl survey figures show both the scale of the opportunity and the risk of overclaiming. AI agents may help create cohesion, yet the economics must be demonstrated in results rather than assumed from the technology’s arrival.
This article is based on reporting by ZDNET. Read the original article.
Originally published on zdnet.com
