I acknowledge the Wurundjeri Woi‑wurrung and Bunurong Boon Wurrung peoples of the Eastern Kulin Nation, and pay my respects to all First Nations people present.
Thank you for the invitation to join you this morning, and to Deakin's SME Research Centre for bringing us together.
I'm especially pleased to be here because this is exactly the right audience for a conversation about artificial intelligence. Too much of the AI debate has been conducted as though the only people who count are big technology firms, big consulting firms and big corporates with procurement teams large enough to form their own netball comp.
Australia's economy is built in workshops, clinics, studios, restaurants, building sites, local accountancy practices, family businesses, exporters, repair shops and spare bedrooms.
So my theme this morning is 'AI for the underdog'.
The question for Australia is not whether artificial intelligence can write another Henry Lawson poem, generate a picture of a kangaroo in a spacesuit, or summarise a Productivity Commission report into a podcast. Useful, yes. Occasionally alarming, also yes. But not enough.
The real test is whether AI helps a small business owner quote faster, roster better, market more cheaply, serve customers more personally, detect errors sooner or spend fewer Sunday nights doing paperwork while the rest of the household is entertained with something better than MYOB.
For a small business, the killer app may look like getting a decent night's sleep.
In September last year, I spoke at a symposium in Parliament House on 'Seizing the Opportunities of AI While Protecting the Fair Go', organised by the Australian Council of Trade Unions, the Centre for Future Work, and the Centre for Employment and Labour Relations Law. The argument I made then was that technology does not determine our future by itself. AI can empower workers, lift expertise and improve job quality. Or it can be used to monitor, deskill and squeeze. The difference lies in the choices made by employers, workers, unions, policymakers and communities. In that speech, I framed the challenge around 3 principles. Voice - giving workers a say on AI. Skills - providing the right AI training. And fairness - using AI to raise standards rather than cut them.
Similar principles should apply to small businesses AI adoption. Here, the question is whether AI helps a small startup compete with a big monopolist.
At the end of June 2025, Australia had 2.73 million actively trading businesses. Around 97 per cent of all Australian businesses are small businesses, and almost two‑thirds are self‑employed or non‑employing.
That should shape how we think about productivity. Productivity growth will come when better tools spread across the economy, not just in frontier firms. Productivity is made in the daily decisions of millions of companies: how to organise work, how to adopt new tools, how to learn from customers, and how to trim wasted effort.
Recent NAB research suggests that AI adoption among SMEs is already moving from curiosity to practice. It found that 42 per cent of SMEs are using AI, with another 14 per cent planning to adopt it. The top uses include administration, marketing and decision‑making, and the biggest benefit reported is time back for business or family.
Time back. For a small business, time is the scarcest input. A large firm might have a legal team, a marketing team, a finance team and a dozen people whose business cards include the word 'strategy'. A small firm often has one person doing all of that after dinner, while also remembering to order printer toner, organise a birthday morning tea, chase an invoice and ask why the website has decided to stop accepting bookings.
So AI can be a leveller. It can give a small firm some of the back‑office capability that used to be available only to big firms.
A suburban builder can use AI to draft customer updates and prepare first‑cut quotes.
A restaurant can use AI to look across bookings, weather and past sales to improve ordering.
A small manufacturer can use AI to spot bottlenecks in production or prepare plain‑English instructions for staff.
A local retailer can use AI to test marketing copy, translate product descriptions, respond more quickly to customer questions or make its website simpler for customers who rely on assistive technology.
A sole trader can use AI to tame the inbox, prepare a grant application, or turn a messy set of notes into a simple business plan.
None of that requires a moonshot. It requires tools that are affordable and easy enough to use before the next BAS statement is due. Preferably tools that do not require a 47‑minute onboarding webinar hosted by someone who says 'solution' when they mean 'button'.
But if AI can be a leveller, it can also become a toll road.
Big firms have advantages that small firms do not. They have more data, bigger IT budgets, specialist lawyers and teams who can spend months testing new systems. They can negotiate with software vendors. They can absorb mistakes. They can hire consultants to tell them what the consultants told the previous client, perhaps with the slides rearranged and a new stock photo of someone pointing at a screen.
Small businesses face a different world. The product claims are confusing. The privacy risks are real. The training data may be opaque. The subscription model may be sticky. The promised gains may be hard to verify. And once a firm has built its customer records, invoices, marketing and workflows into one platform, switching can feel like moving house during a thunderstorm, while the removalists keep asking you to reset your password.
That is why competition policy has such a central role in the age of AI.
Competition policy can sound abstract, but at its heart it is about freedom and fairness. The freedom for workers to move. The freedom for consumers to choose. The freedom for entrepreneurs to enter a market without being blocked by incumbents. The fairness that comes when firms win by offering better products and better prices, rather than by locking people in or wearing them down.
That is the thread running through the government's competition agenda. We have enacted major merger reforms to put competition first. We are revitalising National Competition Policy. We are working to remove barriers that make it harder for new firms to enter and grow. And we are reforming non‑compete clauses because people with ideas should be free to use their skills, change jobs and start businesses. We are banning subscription traps and drip pricing, so that consumers aren't duped, and businesses are not undercut by firms that turn frustration into a business model.
Similar principles should guide how we think about AI.
First, small businesses need capability. Adoption is not automatic. We should not assume that every cafe owner, accountant, tradie or family business has the time to decode vendor claims or test every AI tool. Practical demonstrations count. Trusted advice counts. Peer examples count. And as my colleague Andrew Charlton has argued, if your firm has a close choice between buying a foreign AI product and an Australian one, lean Australian.
Second, small businesses need choice. AI tools should make it easier to compete, not harder to leave. Data portability and interoperability count. If a small business puts its customer relationships and product catalogue into a system, it should not find itself trapped behind a digital turnstile.
Third, small businesses need confidence. Many owners are not anti‑technology. They are anti‑wasting‑money. They are wary of cyber risks, privacy problems and tools that produce plausible nonsense with the confidence of a first‑year student who has not done the reading. Trustworthy adoption means knowing what the tool can do, what it cannot do, and when a human must remain firmly in charge.
AI should help small firms become better versions of the businesses they already are.
That means we should measure AI adoption in practical terms. Hours saved. Errors reduced. Invoices paid faster. Stock managed better. Customers served sooner. Workers trained more effectively. Decisions made with better information.
There is also an Australian opportunity here. Our productivity challenge is often discussed in national accounts language, but it is lived in small moments. These frustrations include a form that takes 20 minutes instead of 2. A customer inquiry answered tomorrow instead of today. A machine idle because the schedule is wrong. A good employee stuck doing repetitive admin instead of higher‑value work.
Multiply those moments across millions of businesses and you have an economy‑wide productivity challenge. Fix enough of them and you have a productivity agenda.
That is why I think the next stage of AI model development should be less about the spectacular and more about the useful. Less HAL 9000, more helping hand.
For SMEs, boring AI may be the most beautiful AI.
Boring AI reads the invoice.
Boring AI drafts the first version.
Boring AI checks the roster.
Boring AI spots the anomaly.
Boring AI translates the manual.
Boring AI reminds the owner that the customer who bought 3 months ago might be ready to buy again.
None of this removes the need for human judgement. In fact, it raises the value of judgement. The best small business owners know their customers, their workers and their market in ways no model can fully capture. AI can help process information, but it cannot replace the trust built in a workshop or on a job site.
A chatbot can suggest the words. It cannot look a customer in the eye and know they need reassurance. It cannot smell when the sourdough is ready, hear when an engine is misfiring, fix a sports injury or sense when a client is asking the question underneath the question.
So the goal is not artificial intelligence instead of human intelligence. It is artificial intelligence in service of human enterprise.
That is why I'm focused today on underdogs. Australia should want AI to help challengers challenge, entrants enter, innovation unfurl and good ideas travel further. We should want it to help a small manufacturer compete with an importer, a regional business reach new customers, a local accountant serve more clients, and a start‑up test an idea before the cash runs out.
The danger is a two‑speed AI economy: big firms racing ahead, while small firms are left with locked‑up data and another monthly subscription. The opportunity is an economy in which AI spreads widely enough to lift productivity from below, not only from above.
Let me finish with this.
The history of technology is a story of choices: rules, institutions, skills, norms and markets.
The printing press spread ideas because people built schools, publishers and libraries around it.
Electricity transformed factories only when firms redesigned production around it.
Computers lifted productivity when businesses learned to reorganise work, not merely when they bought beige boxes and hoped for the best.
AI will be no different.
If it is captured by the largest firms, it will widen moats.
If it is wrapped in confusing contracts, it will become another source of lock‑in.
If it is imposed on workers without trust, it will breed resentment.
But if it is adopted well, AI can give small businesses capabilities that once belonged only to big firms. It can reduce admin, lift capability, encourage innovation and improve decisions. Used well, it can give entrepreneurs more room to do what they do best: solve problems, serve customers and build something of their own.
Australia does not need every small business to become a tech company.
We need technology that helps every good small business become a stronger competitor.
That is AI for the underdog. And if we get it right, the underdog may surprise us all.
