Every CRM in UK estate agency now has an AI story. Street.co.uk markets itself as the most comprehensively AI-integrated CRM in the country, with AI threaded through onboarding, lead handling, sales progression and reporting. Its latest release, Cortex, was unveiled this spring as an assistant capable of holding individual conversations with all of an agent's customers at once. Reapit is rolling out RAI Copilot. Alto pushes Alto Intelligence. The pitch is the same everywhere: the AI lives inside the platform, it knows your data, and it does the work.
Much of it is genuinely useful. I want to say that plainly before I say anything else, because this is not a piece arguing that in-CRM AI is bad at its job. An assistant that drafts listings, answers calls at 11pm and summarises a sales chain saves real hours for real agents, and the vendors building these tools employ serious people doing serious work.
My argument is about something the product demos never show: what happens to the relationship between an agency and its software when the AI moves in. Because the structural effect of these assistants, whatever the intention behind them, is to make leaving harder. The AI is not just a feature sitting on top of the platform. It is becoming the strongest lock the platform has ever had.
Lock-in, stated as a feature
You do not have to take my word for the lock-in, because the vendors say it themselves. When Cortex launched, Street.co.uk's co-founder was clear in trade press coverage that the assistant works only with the Street CRM, so agents need the platform in place to use it. That is not a hidden clause. It is the headline of the strategy, and Street is simply the most candid about a model the whole sector is converging on. Notably, Street Group also owns Spectre, the prospecting tool, so the gravitational pull of that ecosystem extends well beyond the CRM itself.
Think about what an in-platform assistant accumulates. Every workflow it learns, every automation an agent configures, every prompt refined and every process the team comes to rely on deepens the agency's dependence on that one system. Under the data laws I wrote about recently, an agency can always extract its records when it leaves. What it cannot extract is the assistant, the configurations, or any of the working intelligence built up around them. The records come with you. The machine does not.
That was always true of CRMs, of course. The difference is one of degree. A conventional CRM holds your data and your workflows. An AI assistant absorbs your ways of working, and the more capable it becomes, the higher the wall gets. Every month of use raises the cost of exit. In economic terms the assistant is a switching cost that compounds, and it is being given away enthusiastically because the vendors understand that arithmetic perfectly well.
The pattern across the sector: integration flows inward, control stays with the vendor.
The sector will reply that it is open, and point to its integration marketplaces. Look closely at the shape of that openness. Approved partners can plug in, and data flows inward to the platform. But the vendor decides who gets approved. Reapit's AppMarket, to take one public example, has been embargoed to new entrants since June 2025, with Reapit controlling what enters. That is openness on the landlord's terms. The door opens inward, and the vendor holds the handle.
The three questions that actually matter
So here is the test I would put to any agency evaluating an AI assistant, whether from their CRM or anyone else. Three questions, all answerable before you sign anything.
Can I change the model underneath? The AI models powering these assistants are improving every few months, and they are not improving at the same rate. If you cannot choose or change what sits under the hood, your assistant is permanently only as good as your vendor's last procurement decision, and you will have no idea whether you are getting this year's intelligence or last year's at this year's price.
Can I connect my own tools? Open standards now exist, MCP being the obvious one, that let an AI assistant use any tool or data source its owner chooses. An agent working in an open setup who finds a useful third-party tool on a Tuesday can be using it by Wednesday. An agent inside a closed assistant can use what the vendor has approved, when the vendor approves it. One of those is your roadmap. The other is theirs.
Can I take it with me? The decisive question. If every workflow, skill and configuration you build dies the day you migrate, then the assistant is not an asset you are building. It is a dependency you are renting, and the rent is your future freedom to choose.
The alternative already exists
What makes 2026 different from every previous round of platform lock-in is that the open alternative is no longer hypothetical. An agency principal with no engineering team can today sit a general-purpose AI assistant, on their own subscription, on top of their own data, build custom skills for the tasks their business actually repeats, and connect it to the systems they already run through open standards. I have watched non-technical operators do exactly this in an afternoon. The result is not always as polished as a vertical product on day one. But it is theirs. It improves as the underlying models improve, at no extra cost and nobody's permission. And it survives every migration, because it was never inside anyone's walls to begin with.
That is the honest comparison agents should be making. Not “is the CRM's AI clever?” It usually is. The comparison is between a capable assistant you will never own and a capable assistant you will never lose.
I should declare my interest, because it is real: I co-founded Propalt, which supplies property data through exactly this kind of open layer, and Street is a competitor of ours. Readers should weigh my argument accordingly. But the test I have set out does not depend on believing me, or on buying anything from anyone. It depends on three questions any vendor can be asked in a sales call, and on noticing which ones they answer and which ones they talk around.
The CRMs have decided that AI is the future of their lock-in. Agents should decide whether it is the future of theirs.
Daniel Morgan is the co-founder of Propalt, a UK property data platform, and has been a property investor and developer for 20 years, specialising in HMO and student accommodation. This article represents his personal view.
Sources
— Street.co.uk releases “game-changer” AI tool (Cortex) — The Negotiator
— Street.co.uk AI positioning and platform comparison — Street.co.uk
— Alto, Street and Reapit AI compared, AppMarket status — Estate Agent Today
—
AI agents and open APIs in UK proptech — TopTenAIAgents
Co-founder of Propalt and a property investor and developer for 20 years
