For years, modernization programs often assumed that applications and data would progressively move away from the mainframe and toward the cloud. That assumption is giving way to a more selective approach to workload placement.

IBM Institute for Business Value research found that nearly three-quarters of executives are moving away from cloud-first defaults toward outcome-based workload placement. The same research found that 72% of organizations experienced higher-than-expected cloud costs after applications reached production.

These findings support a more deliberate approach to modernization, with workload placement shaped by performance, resilience, security, compliance, cost and business agility.

Place workloads around business requirements

Mission-critical, high-volume transactional workloads often remain best suited to the mainframe when availability, latency, security, compliance and transactional integrity are priorities. Cloud platforms complement those strengths by providing the agility needed for digital services, scalable applications and innovation.

Modernization increasingly depends on how effectively enterprises combine those strengths.

In-place modernization makes mainframe capabilities more accessible through technologies such as z/OSMF, Zowe and z/OS Connect, while integrating cloud-based storage and bringing GenAI and Agentic AI closer to the core.

Architects should assess each capability against its business and technical requirements, moving workloads only where relocation delivers measurable gains in agility, scalability or cost. Where those benefits are limited, secure integration may provide greater value while preserving the strengths of the existing environment.

This selective approach preserves decades of trusted business logic while making it available to new digital services and AI capabilities.

Bring AI closer to enterprise data

AI adds another dimension to workload placement.

Moving large volumes of sensitive transactional data increases latency, cost and governance complexity. Processing AI closer to the mainframe supports real-time analysis while keeping core data within established security and compliance controls.

Technologies such as IBM Telum II and the Spyre Accelerator support this approach.

The benefits are particularly relevant in areas such as fraud detection, payment processing, customer service, risk management and operational optimization, where decisions depend heavily on context and timing.

IBM's work with z17 reflects the growing convergence between transaction processing and AI.

As intelligence becomes part of the transaction flow, architects gain greater flexibility over where processing takes place and how much data needs to move between environments.

Connect the enterprise core

APIs, event-driven integration and common governance form the foundation of a broader hybrid ecosystem:

• APIs expose trusted mainframe services without requiring changes to underlying business logic
• Event-driven architectures enable mainframe, cloud, AI and enterprise systems to exchange information in real time
• Common governance provides a consistent approach to security, compliance, observability and data quality across environments

Observability is also becoming more consistent across platform boundaries. Mainframe telemetry can be converted into OpenTelemetry-compatible formats and integrated with platforms such as Splunk, Datadog and Dynatrace.

This gives operational teams a more unified view across mainframe and distributed environments, helping close a longstanding visibility gap in hybrid architectures.

With these capabilities in place, the mainframe becomes a more active part of the wider enterprise technology environment, with data, services and operational signals accessible across connected systems. 

Move from connectivity to orchestration

Over the next three to five years, intelligent hybrid architectures will evolve from connecting systems to coordinating decisions across them.

AI will continuously analyze operational data, recommend actions and automate routine processes within defined governance controls. As data, applications and AI become more tightly integrated, decisions will flow more efficiently across the enterprise and closer to the point of action.

Enterprises should build toward this incrementally, starting with high-value business capabilities. From there, teams can expose services through APIs, strengthen observability and expand automation as operational confidence grows. This phased approach reduces risk while delivering measurable value without disrupting the critical systems the business relies on.

The effectiveness of a hybrid architecture will increasingly depend on how well its different environments work together. Mainframe resilience, cloud agility and AI-driven intelligence each bring distinct strengths, and the architecture must align them around business requirements.

An intelligent enterprise core brings these elements together through deliberate workload placement, trusted services that work across environments and AI positioned close to the data and processes where it delivers the greatest value.