Hybrid IT’s AI problem starts with workload placement

The cloud-first default is giving way

For years, modernization meant moving everything off legacy platforms and into the cloud. The data now says otherwise. 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 saw cloud costs run higher than expected once applications reached production.

The cloud isn’t the problem. Deciding where a workload runs before asking what it needs is.

AI raises the stakes

AI is only as useful as the data it can reach. In most mature enterprises, the most valuable data is transactional, sensitive, and already governed where it sits. Moving it out for every model call adds delay and exposure.

So AI is moving closer to the core. Processing near the data supports real-time decisions in fraud detection, payments, risk, and customer service, while keeping sensitive data inside existing security controls. A fraud flag that arrives after settlement is a report, not a control.

Placement is a pipeline

Treat placement as a repeatable process, not a one-time migration plan.

Assess. Score every capability on what failure costs, where its data lives, what it must comply with, and what moving it would actually gain.

Place. Move workloads only where relocation delivers measurable gains in agility, scale, or cost. Where it doesn’t, integrate instead of rebuild.

Connect. Expose trusted services through APIs, link systems with event-driven integration, and apply one governance model across every environment.

Observe. Standardize telemetry so operations teams see mainframe, cloud, and distributed systems in one view, then expand automation as confidence grows.

Where it usually breaks

• Placement decisions made by momentum, not requirements.
• Cloud cost models built on pilots, not production behavior.
• AI models chosen first, with data movement figured out later.
• Each environment reports in its own format, so incidents hide in the gaps.
• Ownership split between infrastructure, data, and AI teams, so no one owns the whole picture.

What a working hybrid foundation looks like

• A placement framework applied to every capability, with clear owners.
• APIs that expose core services without touching the logic behind them.
• Event-driven integration across mainframe, cloud, and AI systems.
• Common governance for security, compliance, and data quality.
• Unified observability, so orchestration is built on a picture teams can actually see.

Start with your highest-value capabilities

Pick the three to five capabilities where speed, risk, or cost matter most, such as payments or fraud. Run each through the placement test, expose it as a service, make it observable, and measure the result. Then expand. This phased path lowers risk without putting critical systems on a single cutover.

How Saguna helps

Saguna helps enterprises design hybrid architectures around business requirements. We assess workload placement, connect core systems through APIs and event-driven integration, put AI where the data lives, and set up the observability and managed services that keep it running. The intelligent core isn’t a destination. It’s a placement decision, made one capability at a time