Don’t Buy the Agent.Own the Architecture.

What decides the next decade is a sharper, more revealing question: does a portfolio get better to live in, to work in and to run as it grows, or does it just get bigger? That is an operating-leverage question, and it is where AI either earns its place or becomes an expensive distraction.

Based on conversations with operators, industry surveys and the research behind this piece

A review of AI adoption across build-to-rent, in Australia and internationally, finds it far thinner than the noise suggests. Where it does appear, it takes two forms, and both are real. One is bolting a point tool onto the side. The other is personal productivity.

Each is useful, and neither is the point. Drafting a sharper email with Copilot helps one person; it does not change how the building runs. An agentic workflow that onboards a new resident end to end, collecting documents, running the checks, scheduling the move-in and setting up the account, is a different thing. It is an agent built into the work itself, governed by the owner, and it changes the operation rather than one person’s day. A point tool can improve a single workflow; on its own it does not establish capability across the portfolio.

Personal productivity is the kind that has spread fastest. Microsoft and LinkedIn’s 2024 Work Trend Index reported three in four knowledge workers already using AI at work.² That figure measures individual use, well short of enterprise deployment, and further still from change inside Australian build-to-rent operations. It gives individuals more capacity. There is little evidence yet that it changes how a portfolio is actually run, or who keeps the saving.

Deeper, operational adoption is rarer still, and the reasons we heard are practical rather than philosophical:

AI itself is not the distraction. The risk is stopping at personal productivity, where the quick wins are visible but the operation underneath is untouched. The same work, hand-offs and service failures remain. A faster reply is easy to point to. Whether it ever reaches net operating income depends on how the operation is built.

A workflow can pay within one building. Economical reuse across the next is the additional test of portfolio leverage.

Take a live example. Mirvac’s LIV fund is reportedly weighing two existing Brisbane buildings, some 354 apartments.³ Buy them, and the real question is not the extra rent. It is which of the fund’s standards, workflows and systems those buildings can actually adopt, and what integration will cost.

The answer starts with the experience you want for residents and staff, not with the tool. Architecture turns that intent into an operating system. Without it, you can own the technology and still automate the wrong work.

Only part of the cost base is addressable at all.

In BDO’s sample of 35 valued projects, operating costs run 25 to 30 per cent of gross revenue, about $13,000 per apartment a year.⁴ Only part of that is addressable: the specific workflow, the controllable spend inside it, and the cost of changing it. How large that part is, these figures do not establish.

Here is how the saving actually appears. On a turnover, an agent can run the make-ready sequence end to end: booking trades, chasing quotes, ordering the clean and the inspection, re-marketing the unit and pre-screening applicants, so it is let-ready and re-let sooner, with fewer coordination hours lost. On a repair, a workflow can be designed to triage the request, narrow the likely fault and dispatch the right trade with the right part, then measured on whether it actually cuts repeat visits and wasted call-outs.

But does the saving beat the cost of the agent?

It is the first question a sceptical operator asks, and it is the right one. The honest answer is that it depends on three things:

Who keeps that cut is a separate question, set by the contract and by who owns the data and the workflow underneath. A faster make-ready only lifts income if it fills a vacancy sooner. Freed-up hours only become cash if the costs actually change. And the reusable capability built along the way (the configuration, the data, the exception history) compounds for whoever holds the rights to it.

Few owners will hold every layer, and they do not need to. What matters is which rights they keep: data portability, audit access, a share of the gains, and the freedom to move between operators. Architecture can be owned, governed or contracted for. The one real mistake is surrendering it by default.

And the saving is only the first dividend. The larger return is what the work teaches.

But a finished job is not automatically organisational learning. Closing a work order records that the job is done. It does not explain what caused the fault, whether the repair lasted, or where the same solution would work again. Someone has to capture that evidence, check its relevance and test its reuse. The advantage builds when another team can act on a dependable answer instead of starting from scratch. This is why an organisation can collect data for years and still not get better: a record is not yet knowledge. Learning creates further value when it is reliable, transferable and used, and it can then compound across the portfolio in a way a single saving cannot.

So the agent decision is really an architecture decision, and it has to be made before anything is bought.

Prove AI where proof is cheap. Aim it at the workflow that actually builds advantage.

Start with maintenance triage: frequent, measurable, auditable, with an obvious human boundary. But the proof is not the prize. The prize is the turnover, the sequence from move-out to re-let that touches every role and, reused across buildings, becomes real platform advantage. Technology pays when it is built into a redesigned operating model, not bolted on as a tool.

Transparency makes oversight more important.

People create value through relationships, judgement and persuasion, so keeping tenancy, safety, vulnerability and non-standard spend with accountable people is a deliberate choice. Separately, the law is moving on transparency. From 10 December 2026, Australian privacy rules require firms to explain, in their privacy policy, how personal data feeds automated decisions that could significantly affect someone, a tenancy application included.⁶ It is a transparency duty, not a rule about who decides. But you can only explain a decision if you own, or can see, the workflow and data behind it.