Scaling AI in Australia: Governance, Trust and the Rise of ‘Digital Employees’
In this written Q&A article, Richard Davies (pictured), Field CTO in APAC for leading Artificial Intelligence (AI) development platform OutSystems, discusses how Australian companies are investing in AI, what roadblocks are currently hindering their progress, and what needs to be done to really start seeing major benefits from the next wave of technology.
Australian organisations are investing heavily in AI, but many remain stuck in pilot mode. What’s holding them back from scaling?
The biggest barrier isn’t the technology, it’s more about trust and readiness. Many organisations are experimenting with GenAI and Agentic AI in isolated use cases, but scaling AI requires clean, well-governed data, clear ownership, and confidence in security and compliance. In Australia especially, regulatory expectations are high for laws such as APRA CPS-230 and 234 that place more accountability on organisations to secure their data, so businesses are cautious about moving beyond pilots without having robust guardrails in place. What we’re seeing is that organisations that treat AI as a core capability rather than a side project, are the ones breaking out of pilot mode.
There’s growing discussion around ‘agentic AI’. How do you define it in practical enterprise terms — beyond the hype?
Agentic AI is essentially AI that can take action, not just generate outputs. In enterprise terms, that means systems that can interpret intent, make decisions within clearly-defined parameters, and execute tasks across multiple applications. The key distinction is autonomy with accountability; these agents operate within pre-defined policies, workflows and governance frameworks, rather than acting independently in an uncontrolled way.
To support this shift, established platforms give enterprises structured ways to design, deploy and govern AI agents at scale. OutSystems Agent Workbench, for example, is designed to help organisations orchestrate how agents interact with systems, data and users within defined guardrails. Rather than building agents in isolation, this approach enables enterprises to embed them into existing workflows with visibility, control and auditability — which is critical as AI moves closer to core operations.
What fundamentally changes when AI moves from generating insights to actually taking operational action inside business systems?
The shift is significant. When AI starts taking action, it moves from being a productivity tool to becoming part of the core operational fabric of the business. That introduces new requirements around auditability, control, and resilience. Organisations need to know not just what the AI is doing, but why, and be able to intervene when needed. That means a need for visibility and again, policy guardrails that allow humans to intervene when required. It also means systems need to be designed for collaboration between humans and AI, rather than simple automation.
Do AI agents require the same identity, access and governance controls as human employees? Why?
Absolutely. They require even stricter controls. AI agents can perform set tasks at scale and speed, so the potential impact of an error or misuse is amplified. They need defined identities, role-based access, and clear boundaries around what they can and cannot do. Treating AI agents as “digital employee-assistants” is a useful mindset; they should be governed, monitored and audited with rigor.
How can Australian organisations accelerate AI-driven development without creating new security, compliance or technical debt risks?
The key is to embed governance into the development process itself. That’s where AI-powered development platforms come in, they provide a structured environment where security, compliance and integration are built in from the start. Part of what makes this hard in practice is that it’s not purely a technology problem — it’s an organisational one. Data, UI frameworks, integration layers, and business logic are often owned by different teams, using different tools, with different governance models. AI tooling alone doesn’t resolve that fragmentation. Platforms that consolidate these capabilities in a single environment give enterprises a structural advantage: when the context for data, workflows, and application logic all live in the same place, agents and developers can work with a shared understanding of the system rather than stitching together an approximation. That’s what reduces shadow IT and long-term technical debt in a meaningful, durable way.
Is AI replacing developers — or redefining what it means to be a developer in 2026?
It’s definitely redefining the role. Developers are becoming orchestrators – designing systems, defining logic, and governing how AI operates, rather than writing every line of code manually. AI is therefore accelerating development, but human expertise is critical for architecture, security, planning and business alignment. In many ways, the demand for skilled developers is increasing, or at least evolving – but not decreasing.
How does AI-powered low-code help organisations modernise legacy systems faster, particularly in sectors like government and financial services?
Legacy modernisation has always been constrained by time, cost and risk. AI-powered development changes that by enabling rapid application development on top of existing systems, rather than requiring full replacement. It allows organisations to progressively modernise by integrating AI and agentic capabilities, improving user experiences, and automating processes while at the same time maintaining stability in core systems. This is particularly important in highly regulated sectors where disruption must be minimised.
What are the most immediate, real-world use cases you’re seeing for agentic AI in Australian enterprises?
We’re seeing strong momentum in areas like customer service automation, IT operations, and internal workflow orchestration. For example, AI agents handling service requests end-to-end, triaging incidents, or automating compliance checks. There’s also growing interest in using agentic AI for application development itself, in ways such as generating, testing and refining code within controlled environments.
With growing regulatory scrutiny around AI, how should businesses balance speed of innovation with governance and oversight?
It’s not about choosing one over the other but rather, the organisations that succeed will build governance into innovation from day one. That means clear policies, strong data management, and platforms that enforce compliance by design. If governance is treated as an afterthought, it becomes a bottleneck. If it’s embedded, it actually enables faster, safer innovation within the guardrails that have been defined by government agencies such as the Australian Signals Directorate (ASD) and Australian Prudential Regulation Authority (APRA).
In Australia, this is where many organisations are still catching up. While AI adoption is accelerating, the challenge is moving from experimentation to controlled, repeatable deployment. Without the right governance in place, innovation often slows as teams pause to address security, compliance or data gaps. The priority now is building that foundation upfront, so teams can scale AI quickly without needing to retrace their steps later.
Looking ahead to 2027, what will separate Australian organisations that succeed with AI from those that struggle?
The winners will be those that operationalise AI by embedding it into core business processes, not just experimenting at the edges. They’ll have strong data foundations, clear governance models, and platforms that allow them to scale quickly and safely. Those that struggle will remain stuck in fragmented pilots, often with disconnected tools operating in isolation across different parts of the business. A telling pattern we already see is that AI is being used productively in upstream activities — drafting requirements, synthesising inputs, planning work — but those outputs rarely connect to the development and delivery pipeline in any structured way. The result is that AI efforts compound rather than compound on each other. Organisations that bridge that gap, connecting the full arc from specification through to production, will have a meaningful structural advantage over those treating each AI initiative as its own closed loop.
Ultimately, success will come down to whether organisations can effectively turn AI into a disciplined, clearly defined enterprise capability.
