Henry Harrison
The opportunity AI presents is still real. McKinsey has put a number on it, somewhere between $430 billion and $550 billion of value across real estate, construction and development, though most of that is expected to come from redesigned workflows rather than one-off prompts.
There's a commercial angle too: 58% of the respondents to our 2026 AI in Real Estate survey see the time-based fee model as the part of the business most exposed to AI. It’s worth thinking about that one, because it points to a harder question about where value actually sits once a task stops taking as long as it used to.
There's a temptation to treat AI as a fix for anything. Slow reporting, messy data, inconsistent processes: the instinct is to point AI at the problem and expect it to sort itself out. It doesn't work that way.
Many of you will already have seen the AI in Real Estate survey we published earlier this year. It's been quoted a fair bit since, and I won't repeat the headline numbers in full; however, it was identified that access to AI is now widespread, but only 7% of respondents said it was fully integrated into how their organisation actually works. That gap, not the technology itself, is the real story.
Most current AI use in real estate is still what it was when we first ran the survey: drafting, summarising, research, document review. Useful work. Over half of respondents using AI said it saves them more than 10 hours a month, and almost a third put the figure above 20 hours. We shouldn’t dismiss that, but speeding up an existing task is not the same as changing how a business operates, and the two keep getting treated as if they were interchangeable.
AI is only as good as the process it sits inside. If nobody can say where information lives, which system is the source of truth, or who owns a decision, AI won't clean that up. It will just help someone reach the wrong answer more quickly.
The survey also highlighted the emergence of what we termed "Shadow AI", where people use their own personal tools instead of whatever their firm has approved, because they find them faster or simply better. That's not really a technology problem. It's a sign that the governance conversation hasn't kept pace with how people are already working, and it carries obvious risk once client data and confidentiality are involved.
There's also a clear trust gradient worth remembering. People are comfortable letting AI handle low-risk, verifiable work like transcription or a first draft. That trust drops away fast once judgement or commercial consequence is involved: only 16% said they'd trust AI to estimate rental value or yield. Real estate runs on context, negotiation and knowing when something doesn't feel quite right, and none of that gets replaced by a model, however good it is at drafting an email.
None of this is an argument for caution. It's an argument for sequencing. Firms that give people a tool and stop there will pick up some time savings and not much else. The ones that treat this as a question about process, data and governance first will be the ones actually able to use AI properly, rather than just having access to it.
If you would like to discuss this topic or need to talk about integrating AI into your business, contact Henry Harrison.