Meanwhile, 73% of organizations expect AI to deliver positive benefits within three years, and most expect AI to be central to business processes by 2028.
However, only 10% of companies are implementing AI strategically across their organization.
There is a contradiction here. We see the value and know it’s real, so why isn’t adoption widespread?
This is not primarily a technology issue. It’s an organizational thing.
More efforts needed to close the gap
Many organizations point to skills shortages. More than two-thirds believe their workforce lacks the ability to maximize AI return on investment (ROI), and nearly three-quarters are reskilling or upskilling their workforce. But skills alone cannot explain the gap. You cannot train your way out of structural inefficiencies.
Fragmented workflows, unclear ownership, and slow decision-making processes remain obstacles. Even with a well-trained team, AI cannot scale if the underlying processes are not designed for it or are lacking.
In some cases, employees are solving problems themselves.
Approximately 69% of organizations report that their staff uses unapproved AI tools at least sometimes. This shadow AI is often a signal that demand is not being met when the company’s official systems don’t provide the speed or intelligence needed.
next wave
The next wave of technology is likely to widen this gap even further.
Agenttic AI can plan, execute, and collaborate throughout the process and is expected to deliver an additional 10% return within two years. But only 6% of Australian organizations feel ready to do so.
Able to perform autonomous tasks, AI is reshaping how operations, finance, supply chain, and customer functions work.
Without a coherent strategy, these systems risk being unevenly deployed and increasing fragmentation rather than productivity.
2 speed model
So how do organizations move from individual pilots to enterprise-wide value?
The most effective organizations employ a two-speed model.
In the first example, users leverage the already available embedded AI built into their core business systems for quick wins. This requires minimal change and delivers consistent benefits through widespread adoption, strong governance, and predictable business results. Embedded AI expands the breadth.
Second, they pursue targeted, high-impact innovations, applying custom AI models, deterministic automation, and redesigned workflows to areas where friction and human intervention are still highest.
These solutions require clear process ownership and human oversight and yield the highest marginal benefits. Combining these approaches will enable AI to be composite rather than fragmented.
Reframe AI conversations
This model reframes management’s agenda.
AI strategy can’t just stay within IT. It should be based on your business model, cost structure, customer path, and long-term capability needs.
Similarly, you can’t keep your data in your pocket. It needs to be a shared infrastructure.
Additionally, risk frameworks need to be designed for adaptive rather than static systems, and boards need to treat AI as economic infrastructure rather than a technology item.
AI rewards organized people
Australia has already achieved belief and an early return. And the policy foundation is in place through a new National AI Capability Plan. The next thing you need to do is align.
Organizations that combine embedded AI for broad, reliable benefits and targeted solutions to capture the most valuable opportunities can further improve productivity.
Companies that don’t will continue to face the same challenges, but competitors with the discipline and structure to act decisively will turn the promise of the pilot into a lasting advantage.
Lee Marshall is ANZ’s AI Business Solutions Specialist SAP.
