Article: The rise of agentic AI and why Australia must avoid sidecar chaos

Authored by Brad Stratton, Head of Digital Transformation, Data and AI

Australia is confronting long-running productivity stagnation while also slipping behind global peers in the pace and quality of artificial intelligence adoption.

Labour productivity has flatlined for more than a decade, real wage growth has stalled and organisations across the economy are being asked to deliver more with fewer people amid persistent skills shortages. At the same time, a new wave of artificial intelligence known as agentic AI is reshaping how work is organised and executed globally.

This convergence matters as agentic AI offers a potential pathway out of Australia’s productivity malaise, but only if it is adopted with discipline. Without it, the technology risks reinforcing fragmentation, governance failure and rising costs. This pattern is already visible in many organisations experiencing what might be called “sidecar chaos”, with AI tools bolted onto brittle, siloed systems and little coordination or accountability.

Why agentic AI changes the equation

Agentic AI changes how artificial intelligence operates compared with earlier approaches. Rather than simply responding to prompts or analysing data in isolation, agentic systems are designed to pursue goals. They can plan multi-step actions, draw on context from across an organisation, execute decisions across systems and learn from outcomes over time. In practical terms, this is AI that can interpret a situation, decide what needs to be done, coordinate across platforms and follow through.

This shift matters because it moves AI from task-level assistance to workflow-level execution. When AI begins to operate across end-to-end processes, weaknesses in data quality, system integration and governance are no longer contained within individual tools. They surface quickly and are often amplified where organisations lack the foundations to support them.

Australia’s productivity challenge meets an adoption lag

Globally, agentic AI is being positioned as a productivity multiplier, particularly in service-heavy economies like Australia’s. Local projections suggest it could deliver material labour-productivity gains and contribute tens of billions of dollars to economic output over the coming decade.

Yet Australia is not keeping pace with leading adopters such as the United States and Singapore. While many enterprises report “using AI”, most deployments remain shallow, experimental and disconnected from core operations. Skills shortages compound the challenge as demand for advanced digital capability continues to outstrip supply, with many organisations struggling to recruit people who can combine technical fluency with business judgement.

Over time, AI advantage compounds. Organisations that move early and build capability systematically learn faster, attract scarcer talent and reduce marginal costs, while lagging adoption creates cumulative disadvantage.

The paradox of AI momentum

Despite rapid acceleration in AI adoption globally, results have often disappointed. A large share of enterprise AI initiatives never progress beyond pilot stage, and many organisations report limited or no bottom-line impact despite growing investment.

The problem is often due to the failure of execution.

In practice, AI adoption in many organisations has taken the form of tool proliferation. Chatbots, copilots, automations and point solutions are layered onto existing platforms, often adopted independently by teams because they make local tasks easier. The cumulative effect is duplication, inconsistent outputs, unclear accountability, rising integration debt and growing security and compliance risk.

As AI becomes more autonomous, these issues intensify. Fragmentation that was manageable when tools merely assisted humans becomes destabilising when systems begin to make decisions and act across multiple platforms.

The widening execution gap

Most organisations still operate on fragmented systems, siloed data and governance models designed for slower, more predictable technology cycles. Even simple workflows often require staff to navigate multiple disconnected tools, while data is scattered across platforms with inconsistent standards and limited contextual linkage.

At the same time, AI innovation is moving far faster than institutional readiness. Product teams can prototype in weeks, while governance, risk and decision-making processes often operate on quarterly or annual cycles. This mismatch increases risk and slows value realisation.

For Australian organisations and businesses already behind the adoption curve, this execution gap risks hardening into a structural competitiveness gap.

Avoiding “sidecar chaos”

The lesson emerging globally is that successful AI adoption is less about tools and more about foundations.

First, governance must be designed to enable speed rather than suppress it. Traditional gatekeeper models are too slow for AI-driven environments. Organisations need adaptive guardrails such as clear boundaries within which teams can experiment safely, supported by closer collaboration between innovation, risk, legal and security functions as well as visible executive ownership.

Second, integration must come before proliferation. Adding more AI tools to fragmented environments increases cost and complexity without delivering the benefits of scale. Disciplined platform strategies that consolidate where data consistency and integration matter most, while preserving modularity, are important.

Third, data architecture becomes strategic infrastructure. Agentic AI depends on real-time access to high-quality and accurate data, shared context and reliable orchestration. This requires treating data not merely as something to control, but as something to connect with strong metadata, knowledge layers and observability built in by design.

Finally, change management and workforce engagement matter as much as technology. Capability building, transparency about how roles evolve and early engagement with employees determine whether AI is trusted, adopted and sustained.

From ambition to execution

Agentic AI offers Australia a genuine pathway to lift productivity and competitiveness but only if adoption is deliberate. The organisations and economies that succeed will be those that resist uncoordinated tool proliferation, invest early in foundational capabilities, and align governance, architecture and culture to the speed and complexity of AI-enabled work.

Those that do not will find autonomous systems magnifying existing weaknesses rather than resolving them.

Australia’s productivity challenge is real and pressing, and the technology will continue to advance regardless. Whether agentic AI delivers value will depend on whether Australian enterprises make the foundational choices required before the execution gap widens beyond reach.