TL;DR

On 2 June, Realtor.com switched on RealAssist AI, built with Google — a buyer's concierge that lives inside the portal. It runs side-by-side comparisons of homes and neighbourhoods. It turns a buyer's income and outgoings into a personalised affordability picture. And it remembers where someone left off so they can pick a search back up instead of starting over. It's American and early. But the direction is impossible to miss. The portal is becoming the adviser. The noticeboard era is over. For a small UK agency that has lived on portal leads for two decades, that prompts an uncomfortable question — if the portal can advise the buyer, what is actually left for you to own? My honest answer is “more than you'd fear, but not the bits you might assume.” Here's how I'd think it through.

For most of the portal era, the deal between an agency and a portal has been simple and slightly grim: the portal owns the audience, you pay for access to it, and you do the human work of turning a click into a completion. The portal was a very expensive noticeboard. You didn't love it, but you understood it.

RealAssist AI is the moment that deal starts to change shape. Realtor.com — one of the big US portals — has put an AI concierge directly in front of buyers. It compares homes and neighbourhoods side by side. It takes a buyer's income and expenses and turns them into a personalised what-can-I-afford picture. And it remembers where someone was in their search so they can resume without starting over. Right now it's a select rollout with a wider launch promised. But read what it does and the intent is obvious: the portal wants to be the thing that helps the buyer decide, not just the thing that lists the houses.

I don't cover this because UK agencies use Realtor.com — they don't. I cover it because Rightmove and Zoopla read the same press releases you and I do, and “AI buyer concierge inside the portal” is exactly the kind of feature that crosses the Atlantic in a year or so. The smart move is to use that gap to decide what you'll own before the question is forced on you.

This is a portal power move, not a buyer gift

It's tempting to read a buyer-facing AI tool as a nice service for buyers. It is — but that's not why it's being built. Every minute a buyer spends being advised inside the portal is a minute they're not on your site, not in your inbox, not forming a relationship with your branch. The concierge deepens the portal's grip on the part of the journey that used to be your opening to add value: the “help me understand my options” phase.

If you're an agency, the strategic risk isn't that the AI gives bad advice. It's that it gives good enough advice, early enough, that the buyer arrives at you later in the process, more committed to a view, and treating you as the key-holder rather than the adviser. The portal captures the relationship; you get handed the admin.

What the concierge is genuinely good at — and where it's hollow

I've spent enough time building these things to be unromantic about what they can and can't do. An AI concierge built on portal data is strong at the structured, data-rich stuff: price history, local sold prices, school catchments, commute times, an affordability sum based on numbers the buyer typed in. That work is real and it's not coming back to you. Trying to out-compute the portal on data is a losing game and always was.

Where it's hollow is everything that isn't in the dataset. The concierge does not know that the “quiet street” backs onto a pub garden that's bedlam every Friday. It doesn't know which vendor is motivated and which is testing the market. It can't read a chain, can't sense when an offer is about to wobble, can't tell a nervous first-time buyer the one true thing that gets them over the line. It produces confident, generic, plausible advice — and plausible-but-generic is precisely the gap a good local agent has always filled.

The way I'd frame it to a branch: the portal's AI will win the information layer of the buyer journey — it has more data and can present it faster than you ever could. You win the judgement layer — the local truths, the read on people, the negotiation, the hand-holding through a wobbly chain. Stop competing on the first. Get ruthless about being unmistakably better at the second.

What a small agency should actually own

So when the concierge arrives here — and I think it's a when — what do you double down on? Four things, in rough order of how defensible they are.

  1. The vendor side. An AI concierge is built to serve buyers, because that's where the portal's engagement metrics live. Your instructions come from sellers. The relationship that wins you the listing — valuation, trust, track record on this street — is the part the portal is least interested in automating. Pour energy here.
  2. Local truth that isn't in any database. The flood that wasn't, the development that's rumoured, the genuinely quiet cul-de-sac the data can't distinguish from the noisy one. Write it down, say it out loud, make it the reason a buyer wants you and not just the listing.
  3. The negotiation and the chain. This is judgement under pressure with real money on the line, and it's the single hardest thing to hand to a generic model that has no stake in the outcome. It's also where buyers and sellers feel the value most viscerally.
  4. The relationship that outlasts the transaction. Referrals, repeat business, the local reputation that means your name comes up at the school gate. A portal concierge is transactional by design. You can be the opposite, and that's a moat.

The compliance wrinkle nobody's discussing yet

There's a quieter signal buried in the same week's news. In the US, the Colorado AI Act takes effect on 30 June, and one of its requirements is impact assessments for AI used in housing decisions. The detail isn't what matters to a UK reader — the pattern is. The moment AI starts shaping who can afford what and which homes get surfaced to whom, regulators start asking who's accountable when it gets it wrong.

If a portal's concierge tells a buyer they can comfortably afford a house and the affordability maths was off, the fallout doesn't land neatly on the portal alone — the agent in the chain can get pulled into it too. UK agencies already carry consumer-protection duties around how property is represented. I'd assume those duties stretch, not shrink, as AI-generated advice enters the journey. The practical version: keep your own records of what you told a client, in your own words, so that “the app said so” never becomes your problem to disprove.

What I'd do this quarter

You don't need to build anything to respond to this. You need to decide where your value is going and lean into it deliberately.

  1. Audit your buyer journey for “information vs judgement.” Mark every touchpoint as something a portal AI will eventually do well, or something only a person on the ground can do. Quietly stop selling the first as if it's your edge. It isn't anymore.
  2. Capture your local knowledge somewhere it compounds. Even a simple internal note per street or development turns scattered agent intuition into an asset the branch owns — the one thing a portal's national dataset structurally can't replicate.
  3. Get your own house machine-legible. Clean, structured, accurate listing data on your own site. When buyers start asking assistants “who's the best agent for this street,” the agencies with clear, well-structured information are the ones that get surfaced — and the ones whose own records are tidy if a representation question ever arises.

None of this is a panic response, because it isn't a panic. The portal getting an AI concierge doesn't make a good local agency redundant. It makes the commodity parts of the job redundant and throws the human parts into sharp relief. The agencies that struggle will be the ones who were quietly competing on information all along and never noticed. The ones that thrive already knew their value was judgement. They'll spend this head start making that judgement impossible to ignore.