1. Your architecture will become more agent-centric
Most IT leaders today still think in a familiar stack: users interact with apps, apps talk to APIs, and APIs work with data. In an agentic world, that mental model evolves. You start thinking in terms of humans and agents working together, orchestrated through workflows spanning multiple systems to deliver specific business outcomes.
In practice, this means AI agents will begin calling your APIs, databases, and services much like users and applications already do. They will read data, trigger actions, and coordinate processes across your ecosystem. To make that safe and sustainable, you’ll need consistent processes to manage permissions, set guardrails, and ensure visibility into what agents can access and what they’re allowed to do.
Platforms such as OutSystems’ Agent Workbench will increasingly serve as orchestration layers: places where you define, deploy, and monitor fleets of agents across multiple applications and domains. Without a consciously designed layer like this, you run the risk of drifting into a shadow ecosystem of ad-hoc agents and scripts – automations nobody fully sees, governs, or understands. The more agentic your environment becomes, the more intentional your architecture needs to be.
2. Legacy modernisation will look very different
Despite years of “digital transformation”, most enterprises are still heavily dependent on legacy systems. Many CIOs are still running mission-critical workloads on COBOL, maintaining deep customisations in SAP, or supporting sprawling portfolios of bespoke applications built over decades.
Agentic AI changes how you approach these estates. Instead of rewriting everything in a single, high-risk transformation program, you can begin by wrapping legacy processes in agents. These agents might read screens, parse generated reports, or call existing APIs, effectively automating work around the system without having to replace it on day one.
Over time, some workflows can be taken over entirely by agentic automations that no longer require a traditional user interface. Generative AI can assist in analysing and documenting legacy logic, helping teams gradually decompose monolithic systems into more modern, modular services.
In this model, modernisation stops being a multi-year “big bang” and becomes a series of targeted, AI-augmented upgrades. You transition from a static, fragile legacy core to a more flexible, agent-wrapped architecture that can evolve incrementally, with far less disruption to the business.
3. SaaS won’t disappear – but its role will change
SaaS was, for a long time, the fastest way to deliver new capabilities to the business. SaaS made it simple for line-of-business leaders to adopt powerful tools without waiting for long internal build cycles.
But most enterprises now find themselves using only a fraction of what each SaaS platform offers. They pay for the entire product yet routinely use only half its capabilities. Over time, they accumulate dozens or even hundreds of overlapping applications, each with its own data model, user directory, workflow engine, and reporting layer.
Agentic AI, combined with modern development platforms, starts to change that equation. It becomes possible to analyse which parts of a SaaS product you actually use, capture that structure, and feed it into an AI-augmented platform such as OutSystems. From there, you can generate a right-sized, custom solution that replicates the capabilities you rely on – without all the extra weight you don’t need.
We’re still early in this trend, but the direction is becoming clear: the balance between buying generic SaaS and building tailored software is shifting again, particularly for large enterprises with enough scale to benefit from more precise fits. SaaS vendors anticipate this, which is why so many are racing to add their own agent-based features and AI layers.
What lies ahead?
The future won’t be a simple choice between “SaaS or agents.” It will be a hybrid world where SaaS platforms become agentic themselves, while cross-system agents orchestrate work across multiple technologies from a neutral, IT-controlled layer. For CTOs and IT leaders, the challenge is deciding what to buy, what to build, and where to orchestrate intelligence for maximum impact. That’s where PhoenixDX comes in. We help organisations design this balance, cut through the complexity, and move forward with confidence – turning emerging agentic capabilities into a clear, governed, and scalable path to value.
