Legacy tools gain agentic AI support and redefine enterprise automation
Companies still relying on legacy systems are about to experience a pretty significant turning point. Between March and April 2025, some of the biggest enterprise technology vendors in the world announced updates that place agentic AI at the center of tools that have supported critical operations for decades, from ERPs to mainframes.
It’s no exaggeration to say we’re talking about one of the most sensitive layers of global technology infrastructure. And what’s happening right now could change the way these layers function for years to come.
Workload automation, as it’s known in the industry, has been around for so long that many people don’t even realize how present it is in everyday corporate life. It’s the thing that connects systems from different eras, keeps processes running across platforms that were never designed to talk to each other, and ensures data gets where it needs to go without failures or surprises.
According to Dan Twing, an analyst at Enterprise Management Associates (EMA), workload automation is basically the glue that holds processes together as they move between different application domains and distinct environments. Every large organization has this layer in some capacity, and over the past 20 years the cloud has only grown because it had this support and automation behind it.
Now, with AI entering the game, this layer is gaining a new role: serving as the bridge between the deterministic world of traditional automation and the more unpredictable universe of artificial intelligence agents. 🤖
Broadcom, BMC, and IBM are at the center of this movement, each with its own strategy, pace, and vision of what the future of enterprise orchestration will look like. It’s worth understanding what each one is doing and, more importantly, what it means for anyone using or planning to use these tools in the real world.
What changed in enterprise automation with AI
For a long time, workload automation operated as an invisible gear inside companies. It was there, running in the background, making sure processes happened in the right order, at the right time, without anyone needing to manually monitor every step. This model worked extremely well for decades, especially for mainframe environments, where predictability and stability were always valued more than flexibility.
The problem is that the corporate world has changed quite a bit. These systems now need to communicate with cloud platforms, modern applications, and more recently, with AI agents that make decisions autonomously.
The arrival of agentic artificial intelligence changes the equation in a very concrete way. Unlike AI models that just answer questions or generate text, AI agents execute tasks, make chained decisions, and interact with external systems to achieve a goal. This creates an interesting challenge for traditional orchestration tools, which were designed to handle predictable and well-defined workflows. Now, these tools need to accommodate dynamic behaviors, where the next step may depend on a decision that hasn’t been made yet, rather than a rule carved in stone in the system.
This is exactly where the major vendors are investing. The idea isn’t to throw away what already exists, especially since mainframes and legacy systems still carry critical workloads that can’t simply be migrated to the cloud overnight. What’s happening is an intelligence layer being added on top of these infrastructures, creating a kind of more sophisticated nervous system that can react, adapt, and even anticipate problems before they become incidents.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
As Twing put it quite directly: workload automation connects the old world to the new world and allows them to coexist. And there isn’t just one old world and one new world. There are 15 layers of different ages. There’s still stuff running on client-server architecture out there, there are still first-generation cloud architectures. Workload automation is what makes the enterprise backward-compatible.
Broadcom organizes its portfolio and bets on Model Context Protocol
Broadcom took a concrete step with the launch of version 26 of Automic Automation on April 8, 2025. The update introduced a new Job type called Agentic AI Job, which allows the automation tool to function as a Model Context Protocol (MCP) server, connecting traditional IT orchestration directly to AI agents.
Automic is structured around Jobs, which are software components responsible for executing commands across different environments, such as operating systems, databases, enterprise applications like SAP and Oracle E-Business Suite, file transfers, and web services. Version 26 adds the AI Agent Job type, which orchestrates AI agents and integrates them into existing workflows. Additionally, the new Job type applies Automic’s role-based access control, logging, and audit protocols to AI agent activities, ensuring governance over what those agents do.
Another new feature is a natural language-based workload execution interface, added alongside a Python-based Code Assist tool for building data pipelines.
Rajeev Kumar, who leads workload automation products at Broadcom, explained that users can now type a simple prompt and, with the LLM connected to Automic along with the grounding configurations and rules that are part of the product, the system generates a workflow plan.
To give a practical example: imagine a business analyst wants to pull data from Salesforce every day at 6 AM, move that data to BigQuery for analysis, then to Looker to generate a report, create an AI summary of that report, and send it by email to the CEO. Automic can now generate a workflow plan for this, identifying the Jobs and elements that need to be created, presenting everything to the user for review, and then deploying it with the user’s approval.
Kumar emphasized that the focus isn’t on software engineers but rather on business analysts who were already building these workflows in the past but relied on makeshift tools to do so.
Broadcom has accumulated a broad set of hardware and software businesses over the past decade, which positions it among the most relevant AI infrastructure vendors for enterprises to consider. Stephen Elliot, an IDC analyst, pointed out that people don’t realize how much internet traffic passes through Broadcom hardware. VMware is just one part of this massive group of infrastructure software, and you can’t forget the pieces that came from CA, Symantec, networking software, and chips.
BMC takes a cautious and strategic approach
BMC, with its Control-M, took a different but equally ambitious path. The March 2025 update brought support for AI agents from partners like CrewAI, LangGraph, and Snowflake Cortex, along with an AI assistant and workflow builder called Jett AI. Support for multi-agent AI orchestration is still in development, as explained by Ram Chakravarti, CTO of BMC.
According to Chakravarti, the company is approaching the topic on two fronts. In the core product, it’s already possible to call individual agents based on pre-built integrations and incorporate them into workflows. In parallel, BMC is co-innovating with some of its most relevant customers on significant use cases, where custom agents are being orchestrated with the Control-M core or even with additional capabilities like Managed File Transfer for federated data exchange with AI.
Federated data exchange is a process where query instruments access potentially sensitive data within a partner’s infrastructure, or vice versa, without exporting the information outside the company’s network. This can be an important part of the early stages of working with a new partner. A Control-M pilot customer was able to reduce the federated data exchange process using AI agents from 30 days to less than 12 hours. Chakravarti didn’t reveal the customer’s name or the exact size of the company, saying only that it’s an extremely large organization.
He also offered a relevant warning: unless your AI use cases are aligned with your broader digital business strategy, your AI pilots will wither as science experiments. 🧪
On April 8, BMC also published a statement of direction for AI agent support in its Automated Mainframe Intelligence (AMI) product and expanded zAdviser Enterprise’s AI-generated mainframe reports to include distributed systems applications.
BMC has also undergone significant portfolio rationalization in recent years, separating its IT service management and operations management businesses from its workload automation and mainframe business into distinct companies last year.
The race to modernize mainframes with AI
Broadcom began integrating generative and agentic AI into mainframe management by adding MCP servers to its agile development software Rally and its change management software Endevor, which support mainframes alongside distributed systems. The company also supports the open-source Zowe framework for hybrid cloud mainframe integration, including a Zowe MCP server. Additionally, IBM’s WatchTower observability tool includes AIOps capabilities for mainframes.
BMC’s April updates to the AMI tool included enterprise application analysis reports for zAdviser, its AI-based development productivity monitoring tool. The existing AI assistant in AMI gained integrations with the mainframe Knowledge Hub and a Knowledge Expert chat, which pulls information from sources like runbooks, tickets, log files, and previous incident resolutions.
BMC’s statement of direction for AMI envisions an evolution that goes beyond explanations and recommendations, moving toward autonomous AI agent-driven workflows for system diagnostics and performance, development workflows, security validation, and operational recovery, all learning from past incidents.
Steven Dickens, CEO of HyperFrame Research, assessed that with this statement of direction, BMC is taking a more holistic and thoughtful approach to AI in mainframe modernization than Broadcom. According to him, Broadcom put an MCP server on the mainframe and connected it to a bunch of legacy applications, which lets you interrogate them via the MCP server, but that seems like table stakes rather than a holistic AI deployment. BMC, in Dickens’ view, is looking at things like support data ingestion, Redbooks, knowledge bases, code explanation, and operations automation with a broader scope of thinking.
IBM, Arm, and a parallel with the past
In Dickens’ view, BMC has the most ambitious mainframe strategy, but IBM also has its Concert AIOps software, which supports System Z automation, plus control over mainframe hardware, which was put into play in the recent deal to support Arm chips.
On April 2, IBM announced a deal with semiconductor manufacturer Arm that will bring cloud and mobile applications running on low-power processors to IBM Z and LinuxOne environments through virtualization. This partnership has an interesting historical parallel: IBM made a similar effort to integrate x86 chips into its zBX systems over a decade ago. These systems now support most major enterprise workloads but have some known issues in areas like storage resource management and, in some cases, third-party application support.
Opening the platform to Arm chips could offer another path to third-party application compatibility, and there are strong incentives on both sides to make the integration work, according to Dickens. He pointed out that regardless of what anyone says about the mainframe, it’s highly available, highly resilient, and highly performant, being the fastest commercially available processor. Arm, in turn, gains access to this instruction set collaboration with hundreds of developers and chip architects, and IBM has plenty of experience in this space.
However, Dickens doesn’t expect to see delivery-ready results from this collaboration until the launch of the next generation of System Z, likely in 2028, considering IBM’s typical three-year release cadence. The latest z17 systems were launched in April 2025.
The competitive landscape of workload automation
On the workload automation side, the October 2025 EMA Radar Report for Workload Automation and Orchestration placed IBM Workload Automation in the strong value category. This ranking fell below Control-M and Automic, which were among the tools in the top value leader category, alongside Stonebranch, HCLSoftware, Beta Systems, and Redwood.
But overall, IBM and Red Hat have a solid set of agentic AI tools for hybrid cloud to compete with, in Dickens’ view. He pointed out that when you look at Red Hat and the OpenShift integration done on the mainframe, IBM isn’t just having a mainframe tools conversation but rather a more holistic hybrid IT conversation. 🔍
What this means in practice for companies
For those on the user side of these technologies, the most important question isn’t which vendor is winning the race, but rather what changes in the day-to-day of IT operations. And the most honest answer is: quite a lot, but not overnight. Adopting agentic AI within enterprise automation environments is a gradual process that starts with well-defined use cases, like anomaly detection in data pipelines, automatic reprioritization of jobs during peak hours, or generating smarter alerts that reduce the volume of false positives that operations teams need to investigate.
The most immediate impact tends to show up in reduced operational effort. When orchestration starts working more autonomously, teams can handle a larger volume of demands without needing to scale proportionally. This is especially relevant for companies operating with lean IT teams that need to sustain complex environments, often mixing mainframes, on-premise systems, and multiple clouds at the same time. AI doesn’t solve complexity, but it helps manage it in a smarter way, identifying patterns that a human eye would take much longer to spot.
Another point that deserves attention is governance. As AI agents gain more autonomy within automation workflows, companies need to ensure there’s a clear audit and control mechanism in place. No large organization will accept a system making critical decisions without traceability and well-defined boundaries. Broadcom, BMC, and IBM platforms are each, in their own way, investing in explainability and control mechanisms that allow IT teams to understand why a particular decision was made and, if necessary, reverse or adjust the system’s behavior. This is fundamental for trust in AI to grow sustainably within corporate environments.
What’s happening in the enterprise automation market in 2025 is, at its core, the beginning of a rewrite of how companies operate their most critical infrastructure. Mainframes aren’t going away, traditional orchestration isn’t being thrown out, and AI isn’t going to solve every problem on its own. What will change is how these pieces connect and complement each other, creating systems that are more resilient, more adaptable, and hopefully easier to maintain in the long run.
