IBM expands agentic AI and infrastructure automation to bridge software, cloud and mainframe systems

IBM Corp. is using its annual TechXchange conference in Orlando this week to announce a slate of software and infrastructure updates aimed at helping enterprises put artificial intelligence into operation across hybrid environments.

Spanning agentic AI orchestration to mainframe automation and infrastructure intelligence, they underscore IBM’s intention to embed AI across its software and hardware lines. The new offerings are designed to help customers transition from experimentation to scalable deployment, said Bruno Aziza, IBM’s vice president of data, AI, and analytics strategy.

“Customers are struggling to get value from their investment,” he said. More than 300 customer sessions at the event exhibit deployments of agentic systems in production.

Hybrid AI

Aziza said IBM’s vision differs from competitors by emphasizing interoperability and hybrid operations. “What makes our approach different is that while everyone is providing a siloed, vertically integrated solution for building agents, we think about how all these agents that have been built across multiple platforms are going to be orchestrated and operated at scale,” he said.

Headlining the slate of IBM’s AI announcements are new features for watsonx Orchestrate, a platform encompassing more than 500 tools and customizable, domain-specific agents from IBM and its partners. Designed to be tool-agnostic and adaptable to any environment, Orchestrate is meant to enable scalable deployment and governance of AI agents. New capabilities include AgentOps, a governance and observability layer that provides lifecycle monitoring and policy-based control for AI agents in production. IBM said the tool makes agent behavior more transparent and secure through real-time monitoring.

Aziza said AgentOps extends IBM’s orchestration framework by enabling customers to consistently test and govern agents. “You can get real-time monitoring and policy-based controls so you can not only build and orchestrate these agents, but also run them in production,” he said.

To simplify development, IBM is introducing Agentic Workflows, which enable teams to reuse and scale multi-agent processes, as well as integraton with Langflow, a drag-and-drop visual application builder. The company said the enhancements help both developers and business users build and deploy agents quickly.

Mainframe agents

Six months after announcing its latest generation of mainframes, the watsonx Assistant is bringing agentic capabilities to the Z platform. The goal is to make AI accessible across all mainframe users, from system administrators to developers, said Tina Tarquinio, chief product officer for IBM Z and LinuxONE, the Linux-based version of Z.

The new version of Watsonx Assistant for Z moves “from question and answer to question and action,” she said, enabling agents to perform multi-step workflows, maintain conversational context, and make informed decisions.

The platform features an agent builder with a low-code interface, enabling users to design custom agents tailored to specific enterprise processes. “It can evaluate tradeoffs and understand the pros and cons of any action,” Tarquinio said, adding that the framework delivers the security, compliance and scalability features that are central to regulated industries.

IBM also announced the general availability of the IBM Spire Accelerator, a purpose-built AI processor for mainframe and LinuxONE systems that debuted in April. Designed as a collaborative effort between IBM Research and IBM Infrastructure, Spire supports generative and agentic AI workloads with low latency and power consumption. An IBM Z17 can house up to 48 Spire cards, each equipped with 32 cores and drawing about 75 watts of power.

Unified knowledge graph

In the first major product announcement to emerge from IBM’s $6.4 billion acquisition of HashiCorp in February, IBM introduced Project infragraph, a new capability within the HashiCorp Cloud Platform that provides a real-time knowledge graph for enterprise infrastructure observability. It’s “a unified knowledge graph for infrastructure, connecting application workloads with their related components and preparing enterprises for their AI-driven automation,” said Kyle Ruddy, senior director of product marketing at HashiCorp.

Ruddy said the offering is aimed at organizations struggling with fragmented tooling and reactive operations. “Cloud estates are larger and more distributed than ever, making it impossible for teams to maintain visibility with traditional tools,” he said. “Project infragraph collapses that timeline to minutes,” he said.

The new platform ingests data from cloud providers, Kubernetes clusters, and self-hosted environments to create a live relational view of dependencies across services. It integrates with ServiceNow’s digital workflow platform and security tools from Palo Alto Networks Inc. and Wiz Inc., Future plans are to connect to IBM’s Red Hat Ansible, OpenShift, watsonx Orchestrate, Concert, Turbonomic and Cloudability products.

“Project infragraph is more than just a new capability; it’s a foundation for managing infrastructure in the AI era,” Ruddy said. The product is currently in beta test and no availability date has been announced.

Quantum cryptography

Also new is Guardian Cryptography Manager, a platform designed to help enterprises manage cryptographic keys, certificates and algorithms across complex hybrid environments. The product addresses the growing challenge of cryptographic sprawl and the coming shift to quantum-safe algorithms.

“Only 45% of organizations say that they have full visibility into their cryptographic estate,” said Vishal Kamat, vice president of data security at IBM. “Most companies today manage certificates on a spreadsheet, and that’s just not scalable.”

The new platform provides automated discovery, risk assessment and lifecycle management of cryptographic assets to support “crypto agility,” or the ability to update encryption methods without disrupting operations.

Kamat noted that regulations are tightening. The frequency with which companies will need to update their certificates is expected to drop from the current 398 days to just 47 days by 2029. Quantum computers are also expected to render most current cryptographic algorithms obsolete. Enterprises need automation to keep up, Kamat said.

Software development assistant

For software developers, IBM is previewing Project Bob an AI-based tool system with task generation capabilities for enterprise software development lifecycles.

IBM said Project Bob goes beyond AI coding assistants to work alongside developers to write, test, upgrade and secure software. It works with the Claude, Mistral, Llama and IBM Granite large language models.

Features include support for the Java, JavaScript/TypeScript, Python, RPG, Go, C#, Rust, PHP and Kotlin programming languages; automated upgrades to Java code, framework migrations and multistep refactoring across large codebases. Project Bob can coordinate coding, testing and remediation tasks while preserving context across sessions. Security features include “shift-left” vulnerability scanning, accelerated FedRAMP hardening and migration to quantum-safe cryptography.

Finally, IBM and Anthropic PBC said they have formed a strategic partnership to advance enterprise-ready AI by integrating Anthropic’s Claude LLMs into IBM’s software portfolio. The first integration is within IBM’s new AI-first integrated environment for enterprise software development and modernization. Currently in private preview, the IDE is being used by more than 6,000 IBM early adopters, who report average productivity gains of 45% without compromising code quality or security, IBM said.

The partnership reflects growing enterprise demand for AI solutions that can move seamlessly from experimentation to production. By combining Claude’s generative AI capabilities with IBM’s strengths in hybrid cloud, enterprise software delivery and regulated industry compliance, the companies hope to accelerate secure AI deployment in large-scale IT environments.

Image: IBM