For on-premises/private cloud, the following are all client-managed: application, data, runtime, middleware, OS, virtualization, servers, storage, networking, and facilities. For infrastructure as a service, the following are client-managed: application, data, runtime, middleware, OS; and the following are provider-managed: virtualization, servers, storage, networking, and facilities.

Notably, these improvements came not from massive investments in a structured IT transformation program but from surgically applying intelligence across workflows: a forecasting model here, a dashboard there, a middleware workaround where needed. Each step enhanced decision velocity and business alignment. Gulf Marine was transforming itself by using IT to amplify performance in ways that mattered.

Today’s operating model relies on humans as middleware; AI agents replace humans as the main process coordinators; tomorrow’s AI agents access humans when needed. Figure 3

Translation management systems (TMSes) provide advanced controls and orchestration; but as localization AI improves, lighter-weight middleware and integration platform as a service (iPaaS) may become sufficient. The magic of LLMs is in workflows, not just translation. Unlike NMT, which simply translates text, LLMs’ ability to respond to prompts can automate planning, creation, personalization, editing, and production.

As a new entrant in SCF, Indian Bank needed support across origination and platform build-out; it also needed partners to help tailor solutions to emerging use cases and integrate with enterprise systems like its digital lending platform and middleware. In the first year, the bank focused on anchor program creation; in the second, it matured programs and scaled acquisitions.

Decoupled architectures use APIs and middleware, support continuous modular updates, enable faster innovation, provide data access through independent layers, and allow incremental replacement of system components. The comparison highlights how decoupled architectures increase flexibility and reduce modernization risk.

Finally, connectors include: middleware, IT distribution, web, and integration platforms as a service (iPaaS). Source: 535 integrations listed by evaluated Forrester Wave vendors on their public websites in July 2025. Figure 2 Customer Feedback Reveals Tension Between Innovation And Execution As part of our evaluation, we interviewed 25 reference customers in multiple roles across a range of industries (see Figure 3).

Use middleware to support a hybrid approach. An application can be composed of a mix of microservices and a monolithic core — especially if it’s in the middle of a modernization effort. Middleware, such as API gateways, event buses, and integration platform as a service (iPaaS), are what tie these together (see Figure 3). Embracing architectural diversity while maintaining system cohesion opens up powerful possibilities. Consider how Uber approaches its architecture.

This pattern prioritizes scalability by disaggregating functionality between different platforms, effectively distributing OASIS capabilities across multiple elements of infrastructure and middleware. By deliberately disaggregating OASIS capabilities across storage, data platforms, and control layers, this approach enables independent scaling and performance optimization of each element.

While current tools like APIs and middleware offer a starting point, they lack the depth, clarity, and control needed for agentic operations. Emerging protocols such as model context protocol (MCP) and Agent2Agent (A2A) are being explored to expose platform functions to AI agents, though both are still in early development. OASIS is designed to be flexible across infrastructure types — whether that’s physical hardware, cloud ecosystems, or SaaS-based business fabrics.