The Real Estate Industry Still Runs on Coordination Labour

A lot of the AI conversation in real estate still sits at the visible end of the business... listing copy, prospecting emails, portals, property search, marketing automation. That makes sense, because it is the part people can see. But the more I look at how agencies actually run, the more I think the bigger shift is building a layer or two further back, in the coordination work that sits underneath the visible job.

That is part of why Anthropic’s recent labour-market research caught my attention. Not because it says the professions in the firing line are about to vanish overnight, but because it points to where AI is already biting first. In its March 2026 report, Anthropic says Office & Admin roles have theoretical LLM scope across about 90% of tasks, while Data Entry Keyers already show 67% real-world task coverage in its usage data. McKinsey is pointing in much the same direction. Their recent work says more than 70% of today’s skills still matter, but people are likely to spend less time preparing documents and doing basic research, and more time framing questions, interpreting results and making decisions. (Labor market impacts of AI: A new measure and early evidence)

That feels very familiar if you have spent any time inside a real estate office. A lot of the work is not really the front-stage version of the job at all. It is information coming in from a vendor, details being checked, marketing being prepared, portals being updated, buyer enquiries being answered, notes being logged, reports being prepared, documents moving between agent, client, buyer, conveyancer and back again. Then once a deal is underway, the amount of handling usually increases rather than decreases.

So when people ask whether AI will replace agents, I still think that is the wrong place to look first. The more immediate pressure seems likely to land in the work around the agent... the drafting, summarising, updating, checking, chasing and moving information from one stage to the next. Anthropic’s earlier Economic Index made a similar point in a different way. It found that AI use is still mostly diffuse across tasks rather than whole occupations, with roughly 36% of jobs seeing AI used for at least a quarter of their tasks, but only about 4% seeing use across three-quarters of them. That reads much more like task compression than wholesale replacement. (The Anthropic Economic Index)

You can see the same pattern outside real estate as well. McKinsey’s view is that AI will not make most human skills obsolete, but will change how they are used. Anthropic’s own data leans more toward augmentation than straight automation, and that also feels right to me. The tools are very good at the structured, repetitive, language-heavy parts of work. They are much less convincing at the parts that rely on judgement, context, trust and accountability. (The Anthropic Economic Index)

That distinction matters because coordination-heavy businesses can get awkward very quickly when you speed up one part of the machine without fixing the rest of it. Asana’s recent AI productivity research makes that point very clearly. Their report says 65% of workers think AI creates more coordination work between team members, which is a useful reminder that faster output is not the same thing as a better workflow. In real estate terms, that could mean more AI-generated drafts, more follow-up prompts, more automated notes... and still the same messy handoffs, unclear approvals and duplicated admin sitting underneath it all. (The AI Super Productivity Paradox: More output, same bottlenecks)

That is also why I keep paying attention to what the larger property platforms are doing. Zillow is now talking about AI not as a standalone feature but as something embedded across an integrated transaction system, spanning discovery, touring coordination, pricing guidance and transaction oversight. REA is signalling something similar locally. In February, it expanded natural-language search to half of realestate.com.au’s visitors and rolled out conversational search in beta, while also talking publicly about AI-driven productivity inside the business and in Mortgage Choice. That does not mean agencies should copy portals, but it does suggest the direction of travel is moving beyond isolated tools and toward AI sitting inside the workflow itself. (The next chapter of AI in real estate: What Zillow engineers see coming)

For smaller agencies and solo operators, that could be where the change becomes most obvious. If the coordination layer starts to shrink, small teams get disproportionately stronger. A lot of what once required extra hands can start to be handled by systems that draft, sort, remind, summarise and push work along. McKinsey’s broader point about AI-human work partnerships fits neatly here as well: the gains do not come from removing people altogether, they come from changing how much low-level process work needs to sit around them. (Agents, robots, and us: Skill partnerships in the age of AI)

That is why workflow keeps coming back into it for me. Not AI in the abstract, and not whichever app happens to be fashionable this month... workflow. Where information enters the business, how many times it gets re-entered, where handoffs happen, what still depends too much on memory, what gets rewritten that probably should not need rewriting, and which parts of the process genuinely require judgement as opposed to just repeated coordination. If those things are clear, AI can reduce friction. If they are not, it usually just helps the clutter move faster. Asana’s work on coordination costs, and McKinsey’s warning that organisations need to redesign workflows rather than just sprinkle AI on top, both point in that direction. (The AI Super Productivity Paradox: More output, same bottlenecks)

None of that means the human side of real estate suddenly stops mattering. Trust still matters... judgement still matters... negotiation still matters... local knowledge still matters. But the labour wrapped around those strengths looks much more compressible than many people assume, and that layer is larger than many people realise. So my guess is that the first serious AI pressure in real estate does not show up most clearly at the front of the business. It shows up underneath it, in the coordination layer that has been quietly holding the whole machine together for years.