Kyndryl, a provider specialized in technology services for mission-critical environments, has taken another step forward in its legacy modernization strategy with the launch of a new framework of agentic AI services for mainframes. The announcement, made from New York, aims to position the IBM z/OS platform at the forefront of the intelligent automation wave that is already transforming the rest of enterprise infrastructure.
The company, spun off from IBM in 2021 and with decades of experience managing large systems, presents these services as a way to accelerate application development, gain operational agility, and extract more value from data residing on the mainframe, without the need to move it to other environments.
Agentic AI serving IBM Z
The new services are specifically designed for IBM z/OS clients, the operating system for IBM Z mainframes. The proposal combines three elements: Kyndryl’s mainframe expertise, agentic AI capabilities (a set of agents that collaborate with each other and with human teams), and a hybrid approach that integrates the mainframe with the rest of the IT ecosystem.
According to Hassan Zamat, Global Practice Leader for Core Enterprise and zCloud at Kyndryl, the goal is clear: “to deliver agentic AI-enabled services that reinforce the mainframe’s role as an innovation engine in large enterprises.” In practice, this translates into faster decision-making, automation of complex tasks, and smarter support for operations teams.
These services can be integrated with Kyndryl’s Agentic AI Framework, an orchestration layer managing workflows and AI agents. This allows clients to connect existing automation within other environments with what occurs inside the mainframe, avoiding traditional “silos” between platforms.
More automation, less operational risk
Kyndryl highlights several key benefits for organizations adopting this approach:
- Proactive issue resolution and better decision-making, enabled by agents that monitor events, correlate information, and suggest corrective actions before disruptions occur.
- Increased reliability and adherence to software lifecycle management, automating change, testing, and deployment tasks where errors can have significant financial impact.
- Closing the skills gap, a growing challenge in the mainframe world, through assistants that encapsulate expert knowledge and make it accessible to less experienced or newer teams.
- Faster response to opportunities and threats, from load spikes to new regulatory requirements, empowered by more observable and automated platforms.
All this is supported by Kyndryl Bridge, the company’s open integration platform, which provides intelligent automation capabilities, predictive analytics, and centralized reporting on infrastructure and applications, including the mainframe.
IBM watsonx Assistant for Z and AI ecosystem
Kyndryl’s strategy is not developed in isolation. The company is incorporating technologies from partners like IBM and other AI providers to enrich its offerings. Among them is IBM watsonx Assistant for Z, a conversational assistant specialized in mainframes that helps automate operational and support tasks on the IBM Z platform.
The integration of these assistants with Kyndryl Bridge and the agentic AI framework enables the creation of workflows where humans, AI agents, and legacy systems collaborate in a coordinated manner. The goal is for tools not only to respond to questions but also to execute actions, orchestrate processes, and learn from the history of incidents and changes.
Kyndryl AI Assistant for Z: consolidating mainframe knowledge in one place
Among the new offerings announced is Kyndryl AI Assistant for Z, an assistant trained with decades of accumulated experience from the company’s global mainframe projects. This specialized “brain” aims to solve two critical issues:
- The shortage of mainframe experts, a challenge acknowledged by most large organizations today.
- The dispersion of knowledge across manuals, internal documentation, and tacit team experience.
Integrated into Kyndryl Bridge, the assistant can help interpret error messages, suggest resolution procedures, propose optimizations, or guide maintenance tasks—reducing dependency on a limited number of specialists.
AI close to the data: less latency, more business value
Kyndryl emphasizes that a significant part of its value proposition lies in bringing AI inference as close as possible to mainframe data. Many critical applications—from banking systems to insurance and large retailers—are still running on IBM Z, and massively replicating this data on other platforms incurs costs, risks, and compliance issues.
By moving AI models nearer to the z/OS environment, several objectives are pursued:
- Reducing latency in time-sensitive use cases such as real-time fraud detection, risk assessment, or payment authorization.
- Enhancing the performance of analytical and decision-making applications by avoiding unnecessary data movement.
- Ensuring compliance and security, leveraging existing controls within the mainframe ecosystem.
Market outlook: nearly 9 out of 10 companies want AI on their mainframe
Kyndryl’s move arrives at a time when organizations are viewing the mainframe not just as a stable system but as a key asset for AI-driven transformation. According to Kyndryl’s “2025 State of Mainframe Modernization Survey,” 88% of respondents have already implemented or plan to implement AI—including agentic and generative AI—on their mainframes.
Expectations are high:
- 37% expect to increase business agility.
- 32% aim for faster, more repeatable operations with less human error.
- 31% anticipate significant cost savings.
The study estimates that AI usage on mainframes could generate $12.7 billion in cost savings and $19.5 billion in new revenue over the next three years, driven mainly by resource optimization, fraud detection, security enhancements, and the ability to extract new customer behavior patterns.
However, the report also highlights a major obstacle: 70% of organizations admit difficulty finding talent with hybrid mainframe and AI skills. Kyndryl seeks to address this gap by combining its extensive experience with new AI agentic tools.
Mainframe and AI: a strategic alliance
In a landscape where many companies wrestle with whether to decommission or modernize their mainframe, Kyndryl’s message is clear: Far from fading away, the mainframe is becoming a strategic platform for enterprise AI, especially in scenarios where resilience, security, and data integrity are paramount.
The combination of managed services, advanced automation, AI agents, and tools like watsonx Assistant for Z sketches a future where IT teams can continue leveraging the robustness of IBM Z but add an intelligence layer that enables innovation at the pace of digital business.
FAQs about Kyndryl’s agentic AI for mainframe
What exactly is agentic AI applied to the mainframe?
Agentic AI relies on sets of software agents that collaborate with each other and with human teams to perform complex tasks, make decisions, and automate workflows. In the mainframe context, these agents can analyze events, prioritize alerts, propose actions, and in some cases execute them automatically on IBM z/OS systems.
What are the advantages of using AI on the mainframe instead of moving data to the cloud?
Bringing AI inference directly to the z/OS environment reduces latency, maintains data under existing security and compliance controls, and avoids costs and risks associated with data replication elsewhere. This is especially critical in sectors like banking, insurance, or telecoms, where mainframes remain the primary “system of record.”
In what use cases is AI already being applied on IBM Z?
Common use cases include real-time fraud detection, performance and resource optimization, customer purchase pattern analysis, automation of critical business processes, and improved observability and infrastructure support.
How does Kyndryl AI Assistant for Z help reduce the mainframe talent gap?
This assistant integrates Kyndryl’s accumulated mainframe expertise and exposes it through conversational interfaces and guided workflows. Less experienced personnel can diagnose issues, follow recommended procedures, and execute operational tasks more safely, decreasing reliance on a small group of senior specialists.