Here is a number that deserves more attention than it is getting in most boardrooms between Sydney and Auckland: 61%.
That is the share of Australian companies that Deloitte's 2026 State of AI in the Enterprise report found are seeing improved efficiency and productivity from artificial intelligence. It sounds encouraging - until you read the next number. Only 30% are using AI to deeply transform how they work. Most of the rest are deploying it to automate existing processes, make minor optimisations, and leave their business model essentially unchanged.
Meanwhile, only 65% of Australian respondents said they intend to raise AI investment next year. Globally, that figure is 84%. The gap with international peers is not narrowing. Deloitte's own Australian team put it plainly: "Australian organisations are making strides in AI adoption, but the data shows they are still trailing global peers when it comes to genuine transformation. Many are seeing productivity gains, but few are using AI to go beyond productivity and into growth."
That is the shape of the problem for Australian and New Zealand business leaders in 2026. Not a failure to start. A failure to go deep enough.
Here is what needs to happen.
The majority of Australian businesses that are using artificial intelligence are still essentially running experiments. A ChatGPT account here, a Copilot licence there, a pilot project someone in marketing kicked off six months ago that nobody has quite found time to evaluate. According to Deloitte's research, a full third of Australian organisations are using AI at surface level, with little or no change to existing processes. That is a lot of subscription fees for a lot of marginal improvement.
The organisations getting genuine returns are doing something different. They are making deliberate strategic choices, typically at the leadership level, about where AI can create real competitive advantage — and they are moving from pilot to production. The distinction matters because pilot mode is comfortable and production mode requires genuine decisions about process, people, and accountability.
This week, your task is to identify two - not 20, two - parts of your business where AI can deliver a clear, measurable improvement within ninety days. The criteria are simple. You are looking for processes that are high-frequency and repetitive, where the bottleneck is human processing time rather than human judgement, and where your data is reasonably clean and accessible. Invoice processing. Contract review. Customer query triage. Sales report generation. First drafts of client communications. Demand forecasting from existing data.
Ask three questions about each candidate. Where are our decisions slow, expensive or inconsistent - and could AI make them faster, cheaper or better? Where are skilled people spending time on work that does not genuinely require them - and could AI give those hours back? Where does poor or slow information cause us to make worse decisions - and could AI fix that?
Pick the two candidates that answer yes to all three, with the cleanest data and the most motivated manager. Define success specifically before you start - "reduce the time from customer enquiry to quote by 60%" rather than "improve efficiency." Set a ninety-day clock. Report the outcome honestly, including what did not work.
That is the move from experimentation to transformation. It is not glamorous. It delivers results.
EY Australia's research, published in August last year, found that half of Australian workers — 54% — were not confident using AI in their work, and two thirds said they wanted their employer to provide more formal training. Yet only a third had received any. The finding that CSIRO's Dr Claire Mason described as most striking was not about job losses but about a different kind of insecurity: people who are using AI tools without enough guidance to know whether they are using them well, and without enough cultural permission to admit when they are struggling.
There is also a parallel finding from research covering Australia and New Zealand specifically. A 2025 survey of 300 professionals across the region found that 63% of A/NZ organisations believe they cannot afford not to adopt AI - but 40% of SMB respondents said they felt anxious about it precisely because they had no way to manage unsanctioned AI use spreading through their organisation. That is not technophobia. That is a reasonable response to a real governance vacuum.
Your employees are not waiting for your permission to use AI. They are already using it. The question is whether they are using it well, consistently, and in ways that serve the business rather than create risk. The answer, at most organisations, is that nobody actually knows.
This month, gather your team and have a direct conversation about what is actually happening. Acknowledge that AI is changing how work gets done. Explain what you are piloting and why. Ask them where they think AI could help — people who do the work every day have information that leadership and consultants do not. And then do something that most Australian employers have not got around to doing: give them actual guidance, not just a policy document, on how to use these tools effectively.
CSIRO's research is useful here as reassurance: their study of more than 4,000 Australian firms found that companies adopting AI were hiring at a significantly faster rate than those that were not. The apocalyptic version of the AI employment story is not what the Australian data shows. But that reassurance only means something if your people hear it from you directly.
The single largest barrier to AI adoption in Australian business - identified consistently across the Department of Industry research, the Deloitte report, and multiple independent surveys - is skills and knowledge gaps. Not budget. Not technology availability. Skills.
Over 50% of the SMB workforce, according to industry estimates, has only basic or novice AI literacy. What that means in practice is that organisations are giving people access to powerful tools without giving them the competence to use them effectively. It is roughly equivalent to buying your team expensive professional equipment and then providing no instruction on how it works. The equipment sits there. People use it cautiously, or not at all, or in ways that do not get close to the potential.
What role-specific AI training looks like in practice: your finance team needs to understand which analytical tasks AI now handles reliably and what excellent financial analysis looks like in that context. Your legal team needs to understand the limits of AI-generated document review under Australian privacy law and the obligations under the Privacy Act. Your customer service team needs to learn how to work alongside AI tools without feeling replaced by them. Generic awareness sessions generate attendance records. Role-specific training generates capability.
This quarter, you also need to do something that almost no Australian employer has done: update your performance review framework. If you are asking people to adopt AI and work differently but still measuring the same inputs — hours worked, volume of output, tasks completed — you are sending contradictory signals about what you actually value. The people who embrace AI earliest and most effectively will either be rewarded for doing so, or they will take those skills somewhere they are.
EY Australia's research found a clear generational divide in AI proficiency: Generation Z is significantly ahead of older cohorts. If your organisation's AI adoption strategy does not include a deliberate plan to transfer that capability across your workforce, you are sitting on an internal resource you are not using.
There is a consistent finding across every major piece of research on what separates organisations that are getting real returns from AI from those that are not: the involvement and visible commitment of senior leadership.
Deloitte's global research is unambiguous - enterprises where senior leadership actively shapes AI governance and adoption achieve significantly greater business value than those where it is delegated to technical teams. McKinsey's State of AI 2025 found that high-performing organisations are three times more likely to have senior leaders who actively demonstrate commitment to AI initiatives - not in strategy presentations, but in how they actually work.
In Australia and New Zealand, there is an additional dimension to this. EY's Australian research found there is a trust crisis in many workplaces around AI. Workers are using these tools but lack confidence because leadership is not providing clear guidance, training, or cultural permission. When a chief executive or managing director visibly uses AI in their own work and talks openly about it - what they tried, what surprised them, where it fell short - they grant the entire organisation permission to engage. When the same leaders talk about AI's strategic importance in every town hall but never touch the tools themselves, the message people actually receive is: this is for other people to sort out.
You do not need to become a technical expert. Pick up an AI tool today. Use it for something real. A board paper you are preparing. A set of briefing notes. A data analysis you have been putting off. Notice what it does well and where you had to correct it. Tell your executive team about it at the next opportunity. That costs you nothing and signals everything about whether this organisation is serious.
There is an important practical dimension to AI adoption in Australia and New Zealand that is easy to defer and consequential not to: the regulatory and legal environment is moving, and it is moving in a specific direction.
Australia's National AI Plan, released in December 2025, establishes the policy direction. The government's Guidance for AI Adoption, the forthcoming AI Safety Institute, and evolving expectations around high-risk AI applications in areas like healthcare, financial services, and critical infrastructure are creating a compliance environment that will increasingly reward organisations with governance frameworks already in place. Privacy Act obligations apply to AI tools processing personal information right now. Courts and tribunals are increasingly likely to view failure to oversee AI outputs - particularly in consumer-facing contexts - as a breach of existing consumer protection law. This is not a future risk. It is a present one.
Within the next thirty days, you need clear internal policies on four things: which AI tools staff may use and for what purposes; how AI-generated outputs are reviewed before reaching customers or decision-makers; how personal and business data is protected when employees use third-party AI tools; and what happens when an AI output is wrong, biased, or legally problematic. That last item is not hypothetical. It will happen.
Australia has a genuine AI opportunity. The figures economists cite - contributions to GDP potentially running into the tens of billions annually by 2030, productivity lifts of up to 8% in sectors like healthcare, SMEs achieving productivity growth faster than large firms - are real calculations, not marketing.
But they are conditional. They require that Australian businesses actually do the transformation work, not just adopt the tools. Right now, the Deloitte data suggests most Australian organisations are on the wrong side of that distinction. Globally, 34% of organisations are using AI to deeply transform - creating new products, reinventing core processes, or restructuring business models. In Australia, that figure is 30%, and the gap is widening.
Deloitte's recommendation to Australian leaders was put with unusual directness for a consulting firm: "The opportunity to catch up is real, but action must be decisive." Stuart Scotis, National Leader for Technology, Innovation and GenAI at Deloitte Australia, was more direct still: "Leaders need to move beyond incremental improvements and single use cases, and rethink how work is done in a world of digital labour."
There is also a regional dimension worth naming. Australia's position in Asia-Pacific matters for how the AI transition plays out. The competition for digital infrastructure investment, for AI talent, and for the kind of productivity gains that determine export competitiveness is running across Singapore, Japan, South Korea, India, and China, as well as in the US and Europe. Deloitte's modelling suggests AI-driven productivity gains could contribute as much as 7% of GDP across key Asia-Pacific markets by 2034 - for those that move. A failure to seize the opportunity, as Deloitte's infrastructure report put it, would be "a generational economic fumble."
Australia and New Zealand have shown before that when they move with purpose on new technology - digital payments, cloud adoption, renewable energy - they can move faster than the prevailing assumptions suggest. The AI transition requires the same thing: not optimism, but decisions.
The tools are here. The opportunity is clear. The question is whether there is a decision in your organisation this week or a further discussion about making one.
