- We’re moving from Software as a Service to AI as a Service. Instead of a user moving between SaaS platforms to complete a task, with Agentic AI, an agent can call the necessary systems directly and execute the required steps itself.
- The next phase of agentic AI is not about enabling single agents to collaborate.
- Many organisations are unaware of how to safely design and implement agents. At present, many are just letting them loose and are happy with results, but without realising or planning for the underlying risks.
- The role of developers is likely to evolve. Rather than focusing solely on building integrations between systems, they will increasingly design, refine and supervise agent behaviour within defined governance frameworks.
Recent years have seen AI transform from a generative tool to a workplace colleague. Gone are the days of 2023 when generative AI was the pinnacle of innovation – we are now seeing AI systems operate autonomously within businesses, supporting workers in day-to-day tasks or taking over some entirely.
For the Software-as-a-Service industry, this is especially relevant, as Agentic AI is increasingly applied to software creation, delivery and management. Indeed, the scope for change is so profound that SaaS models and applications as we know them may disappear entirely, as a transformation brings new opportunities to integrate AI agents.
How Agentic AI changes SaaS
To put this in context, ‘traditional’ SaaS is a user-centric model: human users log in, navigate interfaces and manually execute workflows in a process which, generally speaking, is familiar to us all. Similarly, applications are designed around dashboards, configuration panels, and predefined user journeys, delivering value through access to functionality employees actively use.
Agentic AI changes this operating model because agents do not rely on dashboards or visual interfaces. Instead of a user moving between SaaS platforms to complete a task, an agent can call the necessary systems directly and execute the required steps itself. Tasks that once required manual coordination across tools can now be orchestrated end-to-end within a defined objective.
This shifts software from being a primary workspace to being an underlying capability. The interface becomes less important than how systems connect and how tasks are executed behind the scenes. Given that the role of agents is to achieve outcomes, the focus moves from “Which application does the user open?” to “Which services can the agent call?” The disruption, therefore, is not about software vanishing but about control shifting from user-driven workflows to autonomous orchestration.
The rise of agent collaboration
But where is this taking us? The next phase of agentic AI is not about making single agents smarter; it’s about enabling them to collaborate. For instance, the Model Context Protocol, developed by Anthropic, provides a mechanism for linking agents so that information does not remain trapped within a single application or stack. Rather than responding to isolated prompts, agents can pass tasks, data, and state to one another as part of a wider workflow.
In addition, Microsoft’s recent introduction of its Work IQ orchestration capability is another step towards bringing agent-to-agent collaboration into mainstream enterprise tooling and out of experimental environments.
The underlying point of all this innovation is that when agents can operate across internal and external ecosystems, the boundaries between applications begin to blur. The result is a shift from isolated automation to coordinated execution, where multiple agents contribute to achieving a single outcome.
In this environment, human users are no longer the sole orchestrators of workflow. Instead, agents can initiate, adapt and complete processes with limited intervention. This represents an architectural turning point because once context sharing and agent coordination are embedded in enterprise platforms, software shifts from something users operate to something agents consume. Interfaces become secondary to integration layers, and workflows are triggered by objectives rather than manual input.
Building for the AI as a Service era
As with every other AI-centric innovation, building new capabilities depends on robust governance and infrastructure that can support increased autonomy. An important part of the overall challenge is that many organisations are unaware of how to safely design and implement agents. At present, many are just letting them loose and are happy with results, but without realising or planning for the underlying risks.
For instance, when agents can share context and coordinate across systems, operational complexity also increases. This means businesses will need to set out extremely clear policies defining what agents are permitted to do and what they aren’t. These should be implemented as guardrails built into the orchestration layer itself, particularly where agents exchange sensitive business data.
Managing these issues appropriately is vital, not least because turning on collaborative agent capabilities without appropriate oversight risks exposing data in unintended ways. In this context, human oversight remains important, especially in environments where agents are making decisions based on incomplete or evolving information.
Infrastructure foundations must also support traceability and accountability across agent-led workflows. It is not enough to know that a task has been completed; organisations must understand how decisions were reached and ensure that policies are consistently enforced.
At the same time, the role of developers is likely to evolve. Rather than focusing solely on building integrations between systems, they will increasingly design, refine and supervise agent behaviour within defined governance frameworks.
Ultimately, organisations that take the time to strengthen their governance models and infrastructure foundations now will be better placed to manage the transition to agent-led systems. The bottom line is that as autonomy increases, so too must the structures that support it.
Mark Skelton is CTO at Node4.
Read more
Digital friction is where most AI initiatives fail – Get your digital employee experience in line to ease the introduction of AI initiatives into your organisation
From generative to agentic AI – now the real transformation begins – Node4’s Mark Skelton takes us through the move from generative to agentic AI and how to approach it in your organisation
Why ISO 42001 sets the standard for responsible AI governance – With the use of AI increasing inall areas the development of effective governance is paramount. ISO 42001 is the latest standard helping businesses build trust moving forward
