I’m not an estate agent – but I know AI, I know data and I know what happens when you apply both properly to an industry that runs on relationship, trust and compliance.

That’s not a gap in my credentials. It’s exactly the lens this conversation needs right now.

The people selling AI to agents often come from agency backgrounds. They know the job intimately. What they’re less likely to challenge is whether the technology underneath their product actually does what it says. I will. Because that’s my world, not theirs.

I’m a numbers person. I like facts. I like being able to say: here’s what changed, here’s how we know, here’s what it means.

So when the industry got flooded with AI announcements from CRM providers in the last few weeks, I read them carefully. And what I found, underneath the headlines, is worth saying out loud.

A lot of what’s being sold to agents right now as AI isn’t really AI.”

A lot of what’s being sold to agents right now as AI isn’t really AI. It’s a wrapper around someone else’s large language model, marked up, handed to you with the risk attached, and a price card that doesn’t show you the full cost.

That’s not a comfortable thing to write. But I think agents deserve to know it.

Not all AI is equal

There are broadly three stages to how AI can operate in a business.

The first is generative. Write me a listing description. Draft me an email. Enhance this photo. Useful. But table stakes now. Most of what’s been launched in the last 18 months lives here, whether it says so or not.

The second is insight. AI that doesn’t wait to be asked, but learns from your data, your pipeline, your history, and surfaces the things you didn’t know to look for. Fewer tools operate here, because it requires depth of data, not just access to a model.

The third is execution. AI that doesn’t just surface the insight but acts on it. Triggers the follow-up. Books the viewing. Manages the compliance task end to end. Logs everything. This is where AI starts to genuinely change how an agency runs.

When you read recent announcements carefully, ask yourself which stage is actually being described. Then ask whether it’s live today, or scheduled for summer – or sometime in the future.

What the price card doesn’t show you

Several of the AI tools now being positioned at agents by CRM providers are built on third-party large language models. A licensed version of someone else’s AI, with a layer on top. That’s not inherently wrong. But it has two implications that agents should understand before they sign up.

What you actually pay every time the model processes a query, sits on top of that.”

The first is cost. The licence fee is what you’re quoted. The token cost, what you actually pay every time the model processes a query, sits on top of that. At the volumes being implied by some of these tools, the total cost of ownership looks materially different from the headline number. Arguably, there may be better value in just utilising the AI tool directly, whether that’s Claude, Gemini or GPT.

Ask the question. Model it out before you commit.

The second is risk. These models are non-deterministic. Ask the same question on Monday and Tuesday and you can get a different answer. In estate agency, where you’re communicating with vendors, buyers and tenants on behalf of your business, that is not a theoretical problem. It is a live one.

And when AI generates a client communication in a regulated industry and something goes wrong, the question of who carries the risk matters. Some providers are already on record saying disclosure of AI-generated communications is a decision for the agent, not the platform.

The Property Ombudsman and the ICO will have a view on that in due course. Agents should form their own view now, before they’re in front of a complaint.

The thing you can’t manufacture

Here’s where I want to be direct about what actually makes AI useful in estate agency, because it’s not the model. It’s the data underneath it.

Alto is used by more than a third of all UK estate agents. That means we hold two decades of real transaction data, real workflow data, real outcomes from agencies like yours.

When Alto makes a decision, it draws on actual transactions. Actual pipelines. What separated the deals that closed from the ones that stalled.

You cannot manufacture twenty years of estate agency workflow data.”

You can build a compelling interface in a quarter. You can write a press release in a week. You cannot manufacture twenty years of estate agency workflow data.

That distinction matters and it is worth pressing any provider on directly.

Generic AI is generic. A model trained on the internet does not know how estate agency actually works. Depth of data is not a marketing claim. It is the difference between AI that guesses and AI that knows.

The questions worth asking

Before you commit to any AI product or any CRM that claims to have AI in the platform, ask which stage of AI it is actually delivering, and whether it is live today.

Ask what the total cost looks like once token usage is factored in at the volumes you intend to use it. Ask who carries the compliance and disclosure risk when something goes wrong.

Ask what data the AI is trained on, and whether it was built specifically for estate agency or adapted from something more generic.

The agencies that perform well in the next few years will be the ones that used AI to get real work done and enact real change in their business.

For some providers, the gap between what is being announced and what is actually working inside agencies is significant. It is closing. But it has not closed yet.

At Alto, it already has.

READ MORE ABOUT ALTO