Meanwhile, 73 per cent of organisations expect positive AI returns within three years, and most expect AI will become central to business processes by 2028.
Yet only 10 per cent of businesses are adopting AI in a strategic, whole-of-organisation way.
Herein lies the paradox: we can see the value and know it is real, so why isn’t adoption broader?
This is not primarily a technology problem; it is an organisational one.
More is needed to bridge gap
Many organisations point to skills shortages. Over two thirds believe their workforce lacks the capabilities to maximise their AI return on investment (ROI), and almost three quarters are reskilling or upskilling workers. But skills alone don’t explain the gap. You can’t train your way out of structural inefficiency.
Fragmented workflows, unclear ownership and slow decision pathways remain obstacles. Even well-trained teams cannot scale AI if the underlying processes aren’t designed for it or are missing.
In some cases, employees are taking matters into their own hands.
Some 69 per cent of organisations report that staff are using unauthorised AI tools, at least occasionally. This shadow AI is often a signal of unmet demand when internal official systems are not delivering the speed or intelligence needed.
The next wave
The next wave of technology could well broaden this divide.
Agentic AI – capable of planning, acting and collaborating across processes – is expected to deliver an additional 10 per cent return in two years. Yet only 6 per cent of Australian organisations feel ready for it.
AI capable of autonomous task execution will reshape how operations, finance, supply chains and customer functions work.
Without coherent strategy, these systems risk being deployed unevenly, accelerating fragmentation rather than productivity.
The two-speed model
So how do organisations move from isolated pilots to enterprise-wide value?
The most effective organisations are adopting a two-speed model.
In the first instance, users seize quick wins leveraging already available, embedded AI built into core business systems. Because of this, minimal change effort is required and it provides consistent returns through widespread adoption, strong governance and predictable business outcomes. Embedded AI gives you breadth.
Second, they pursue targeted, high-impact innovation – applying custom AI models, deterministic automation and redesigned workflows to areas where high friction and human intervention remain highest.
These solutions demand clear process ownership and human oversight and generate the greatest marginal gains. Combined, these approaches allow AI to compound, not fragment.
Reframing the AI conversation
This model reframes the executive agenda.
AI strategy cannot sit inside IT alone. It must be anchored in business models, cost structures, customer pathways and long-term capability needs.
Likewise, data cannot sit in pockets. It must be a shared infrastructure.
In addition, risk frameworks must be designed for adaptive systems, not static ones, and boards need to treat AI as economic infrastructure rather than a technology line item.
AI rewards the organised
Australia already has the belief and early returns. And through the new National AI Capability Plan, we also have the policy foundations. What we need next is alignment.
Organisations that combine embedded AI for broad, dependable gains with targeted solutions for the highest-value opportunities will unlock compounding productivity growth.
Those that do not will continue to face the same challenges, and competitors with the discipline and structure to act decisively will convert the promise from their pilots into lasting advantage.
Lee Marshall is AI business solution specialist SAP for ANZ.