The Reduction Is Real. The Count Is Wrong.

The AI jobs debate keeps trying to settle on a clean number. Half the white-collar workforce disappears, maybe 20% disappears, maybe very little happens and the whole thing was overstated. I do not think real estate will fit neatly into any of those versions, because the industry has never been a clean employee-count story in the first place.

The reduction probably is real, but a lot of it will be counted in the wrong place. Official data remains useful, although real estate work already sits across too many structures to be explained by a simple employee number. Agent, assistant, contractor, property manager, principal, licence holder, franchisee, sole trader, company, outsourced supplier, offshore team, software vendor. The same economic activity can sit in several different categories depending on how the business is built around it.

As AI gets better, the first pressure is unlikely to be a clean replacement of the visible agent. It is more likely to land in the work wrapped around the agent: the drafting, checking, chasing, updating, reporting, coordinating, scheduling, preparing and following up that keeps an office functioning. Some of that work will still need people, some will go offshore, some will be carried by software, and some will disappear into better process. From the outside, it can look like a staffing reduction. Underneath, it may be labour being reallocated into systems, suppliers, capital and smaller operating units.

Deloitte’s June 2026 employment forecast is useful because it does not lean on a neat layoff story. It says there is still limited evidence of widespread AI-driven job losses in Australia, with the effect more visible in hiring than redundancies. Employment in AI-disrupted occupations has continued to rise so far, but vacancies are weakening, and Deloitte expects annual employment growth in those occupations to slow materially once AI is included in the forecast. Jobs and Skills Australia reached a similar view last year, arguing that generative AI is more likely to augment human work than replace it outright, although routine tasks remain more exposed.

That feels much closer to what I would expect inside agencies. A principal may not sack half the office. They may simply not replace two admin people over the next few years. A sales team may carry the same number of listings with one less support person. A property manager may not double a rent roll overnight, but maintenance triage, arrears reminders, owner updates, inspection reports and routine document handling may require fewer human hours per property. Nothing dramatic needs to happen in one moment. Two roles leave and only one comes back, the office keeps functioning, and the old staffing ratio begins to look heavy.

The official real estate numbers do not describe an industry falling apart. Jobs and Skills Australia has the broader Rental, Hiring and Real Estate Services workforce at 212,600 in February 2019, rising to 230,800 by November 2024, then easing back to 226,200 by February 2026. Within that, Real Estate Services is the largest sector at 147,500 workers. Real Estate Sales Agents sit at 103,900 employed people, and Property Managers at 39,200. On a simple chart, the broad industry is still larger than it was before COVID.

Even that comparison needs care. Rental, Hiring and Real Estate Services is a broad division, not a clean view of residential agency employment. It includes property operators, rental and hiring categories and other activity that does not map perfectly to a suburban sales and property management office. More importantly, it does not tell you how much of the work sits inside a traditional bricks-and-mortar agency, inside a virtual agency model, inside a franchise platform, inside a contractor arrangement, inside a one-person company, or inside an outsourced service provider.

The category problem is not small. A person can be an employee, an owner-manager of an incorporated enterprise, an owner-manager of an unincorporated enterprise, or a contributing family worker. In real estate, those categories are not just technical labels. A salesperson can leave payroll and become a licensed operator under a network. A property manager can work inside a traditional office, inside a supported independent model, or inside a smaller business built around outsourced trust accounting and software. A principal can run a ten-person office, or a two-person office with a much larger external operating layer around it.

The headcount inside the old agency may fall, while the productive person still exists in the market. ABS data shows Australia had 2.73 million actively trading businesses at 30 June 2025, and only 994,178 of those were employing businesses. In 2024-25, non-employing businesses increased by 71,633, companies increased by 54,666, and sole proprietors increased by 19,186. ABS also recorded a net movement of 32,428 surviving businesses from employing to non-employing, while Rental, Hiring and Real Estate Services businesses increased by 3.1%. ASBFEO’s small-business data adds another layer: 97.3% of Australian businesses were small businesses in June 2025, and 64% of Australian businesses were self-employed or non-employing.

AI did not cause all of that. COVID, remote work, labour shortages, outsourcing, franchise models, platform economics and cost pressure were already pushing people toward leaner structures before ChatGPT arrived. Real estate had already been through a virtual-office experiment during the pandemic, and the market never fully returned to the old model. Independent-agent networks, multi-brand models and supported operators were already part of the landscape. AI makes the next version easier to imagine and, in some cases, easier to run.

A salesperson who might once have needed a larger office around them can now buy a lot of the operating layer by the month, by the task, or by the transaction. Listing copy, campaign assets, database work, social content, call summaries, appraisal preparation, buyer follow-up, reporting and basic workflow management are all becoming easier to carry through software and structured support. Trust accounting can be outsourced. Marketing can be templated. Admin can be offshore. Some customer communication can be triaged by systems. None of that makes the person irrelevant, but it reduces the amount of fixed labour required around them.

This is the part of the reduction I think is easier to miss. It may not mean a dramatic fall in the number of people making money from real estate. It may mean fewer people sitting inside the traditional office structure, fewer assistants per agent, fewer coordinators per campaign, fewer internal admin roles per dollar of GCI, and fewer full-time staff required before a small operator can look credible in the market. Labour moves into software cost, supplier cost, platform fees, owner margin, and capacity for the agent to carry more listings without the old support base underneath them.

ABS industry data gives a crude but useful warning against using employment alone. In 2023-24, employment in Rental, Hiring and Real Estate Services fell by 1.0%, while sales and service income rose 7.7%, EBITDA rose 9.4%, and industry value added rose 7.6%. Again, that is broader than agency work and should not be stretched too far. But it is still a useful reminder that employment and output do not always travel together. An industry can have fewer counted workers in one frame and more income, more value added or more capacity in another.

You can see the same logic in smaller, more practical examples. Photography is one. A traditional sales campaign may still justify a professional photographer, especially at the premium end. But if an AI-supported photography workflow can give a rental property, a lower-value listing or a condition-update image set something close to professional presentation at a fraction of the cost, the category behaves differently. Some traditional photographers lose work while the total volume of photography-quality output grows. The industry does more of the thing, but the labour attached to the thing changes.

Property management is another. The human part of property management is not going away, and in many respects it is becoming more complex. Compliance, judgement, conflict, owner expectations and tenant stress all still need people. But a meaningful part of the workload is structured service work: maintenance triage, routine updates, arrears communication, inspection reporting, document handling, lease administration, owner statements, task routing and record keeping. If that layer becomes easier to systemise, an agency that once saw property management as too heavy may look at it differently. The barrier is not removed, but it is lowered.

Jevons paradox is useful here, provided it is kept practical. When something becomes cheaper and easier, you do not only reduce cost inside the existing use case. New use cases become viable. Professional-grade imagery becomes commercially sensible in places it previously was not. Better reporting becomes possible where the labour cost once made it inconsistent. Smaller operators can offer a level of service that used to require more staff. A country agency may consider adding a rent roll. A solo agent may look more like a small business than a salesperson with a laptop.

The tax and capital side probably deserves its own edition, although it is part of the same pattern. If more people earn through companies, contractor structures, owner-managed businesses or capital-light operating models, governments eventually care. Australia relies heavily on personal income tax, and a labour market with more business income, deductions, supplier arrangements, software, capital equipment and eventual business value is not the same fiscal machine as one built mainly around payroll. That is not a moral argument. It is just another sign that the categories are getting messier.

For agency principals, staff numbers alone are becoming a lazy measure of operating strength. A larger office with more people may still be less capable than a smaller one with cleaner workflow, better systems, sharper outsourcing and fewer places where information leaks. A leaner office may not be under-resourced if the work has been properly designed. Equally, cutting people without redesigning the work underneath is just another way to create service failure.

The better measures are more specific: revenue per operator, listings per support person, properties managed per property manager, admin hours per campaign, cost per listing produced, speed from enquiry to follow-up, quality of vendor reporting, software and outsourced support as a share of revenue, the number of times information gets re-entered, and the number of handoffs required before a task is finished. The structure of the business matters more than the raw number of people sitting inside it.

For me, the AI jobs debate misses real estate because the industry may absolutely reduce parts of its employee base without becoming a smaller industry in any simple sense. In some areas, the employee base probably should reduce. A lot of admin work has survived for years because systems were poor, software was fragmented, and someone still had to hold the pieces together manually. AI, outsourcing and better workflow will put real pressure on that labour. Pretending otherwise seems naive.

But a reduction in traditional employee count does not automatically mean a smaller industry. Work can be automated, externalised, capitalised, carried by software, or reappear through independent agents, owner-managed companies, sole traders, virtual agencies, network partners and specialist suppliers. The old office may shrink while the operator gets bigger.

So yes, I think the reduction is real, but if we only count the traditional employee base, the count will be wrong.