I have spent a great deal of the past year reading disciplinary decisions. Tribunal decisions, complaints committee findings, the occasional appeal. It is not glamorous reading. Most of them run to a familiar rhythm: a buyer says the agent told them something. The agent says they didn’t, or that they said it differently, or that the buyer misunderstood. The property has already settled. The conversation happened at an open home fourteen months ago, or on a phone call, or in a text thread that one party has since deleted.
And then the tribunal does what tribunals do. It weighs credibility. It reconstructs, from fragments, what was probably said in a lounge on a Sunday afternoon in front of a heat pump that may or may not have been described as “brand new.”
Read enough of these and you stop seeing individual cases. You see a system with a missing organ. The disputes are not really about dishonesty — some are, most aren’t. They are about the fact that nobody, anywhere in the transaction, was keeping a reliable record of what was actually said. The largest purchase of most people’s lives runs on claims made out loud, in passing, by someone paid on commission, and the official memory of those claims is whatever both sides can plausibly assert two years later.
That is the problem I wanted to work on. So before I committed to building anything, I did what any sensible person does before quitting their evenings and weekends for a product: I went looking for who else was already solving it.
What I found instead
Short answer: nobody. Longer answer: a very crowded market solving a completely different problem, wearing a costume that occasionally says “compliance” on it.
AI in real estate is genuinely booming.
PwC and MetaProp’s latest proptech research describes AI moving from experiments into practical adoption, and the framing is consistent throughout — efficiency, performance, streamlined operations, better decisions. Buildium’s industry report found AI adoption among property managers jumped from 20% to 58% in a single year. Scroll through Y Combinator’s proptech portfolio and you can watch the money concentrate: lease abstraction in minutes, AI property managers that schedule tours and chase vendors, HOA back-office automation, banking layers for landlord cash flow.
Every one of these products makes somebody in the industry faster. Valuation models price the property faster. Lead tools qualify the buyer faster. Generative tools write the listing faster. Transaction coordinators chase the signatures faster.
Even the products that market themselves as compliance tools turn out, on inspection, to be speed tools in a high-vis vest. They check disclosure packs for missing signatures. They track regulatory changes so the brokerage doesn’t have to. They validate lease data across a portfolio before an audit. Useful work — I say that without irony, I’ve built a document-completeness reviewer myself and there is real value in it. But notice what all of them have in common: they are sold to the agent, the brokerage, or the operator, and their job is to protect that customer from fines and delays.
Not one of them remembers what the agent said to the buyer.
The gap is structural, not accidental
For a while I assumed this was an oversight — a corner of the market nobody had noticed yet. I no longer think that. I think the gap exists because of who pays.
Proptech’s paying customers are agents, brokerages, developers, landlords, investors. So proptech optimises their workflows, and it always will. A feature that says “we keep a durable, cross-channel record of every claim your salespeople make, and we will flag it when this week’s version contradicts last month’s” is not a feature that market can sell to its own customers. It is not that anyone sat down and decided against building it. It is that the incentive structure of the entire industry points the other way, quietly and permanently.
The buyer — the person carrying the mortgage and the risk — is not anyone’s customer. So the buyer gets chatbots.
And here is the part that made me stop deliberating
The efficiency wave is not neutral to this problem. It is actively making it worse.
Think about what generative AI in real estate actually produces: more claims, faster.
More listing copy, more marketing emails, more chatbot answers, more “AI-enhanced” photography. In November 2025, New York’s Department of State had to warn agents and buyers about a surge in AI-generated listing images that misrepresent properties — including images with window views that do not exist. Around the same time, American brokers started writing candidly about the licence risk of pasting hallucinated listing descriptions straight into the MLS: invented features, wrong floor areas, amenities the property has never had. One put it well — AI amplifies whatever sloppiness already existed in your process.
So the industry is scaling up the production of representations while the accountability layer stays exactly where it was in 1995: two people’s recollections and a tribunal’s judgment about who seems more credible. The volume of statements is going vertical. The memory of them is flat.
I find it hard to look at that chart and get excited about shaving another forty minutes off transaction coordination.
What Resaido actually is
Resaido is a compliance memory.
It records what was represented to a buyer across channels — what was said at the open home, what was written in the follow-up email, what appeared in the listing, what changed between the first conversation and the third — and it notices when those things stop agreeing with each other. Not sentence-by-sentence gotcha flagging; I tried that version first and it was the wrong shape. Contradiction over time is where the real signal lives, because that is where the real disputes live. Fourteen months later, the question is never “was this sentence compliant.” It is “what was this buyer actually told, by whom, and when did the story change.”
I want to be precise about something, because it matters: this is not an anti-agent product. The agents who end up in front of tribunals fall into two groups, and the honest ones are the larger group. An honest agent with a clean record has something they have never had before — proof. Right now, when a buyer’s memory drifts (and buyers’ memories drift too; motivated recall is nobody’s monopoly), the diligent agent has nothing to stand on either. A shared record protects whoever told the truth. That is the whole point of records. It is why we invented writing.
Why New Zealand, why me
New Zealand is a strange and excellent place to build this. One statute, one regulator, one licensing regime, and — crucially — a tribunal that publishes its decisions. I have a corpus of real disputes, with real fact patterns, showing exactly where representations break down and exactly what evidence would have resolved them. Most markets can’t offer that. I have also spent extensive years inside these transactions, which means I know where the bodies are buried in a sale and purchase file, and I know which of the tribunal’s patterns are quirks and which are load-bearing.
There is a version of this article that ends with a growth thesis and a market-size number. I am not going to write that version, because it would be dishonest about my motivation. I did not start building Resaido because I found a gap in a market map. I found the gap because I was already staring at the problem — file after file of decent people, on both sides, litigating the contents of a conversation no one recorded.
The entire industry is building tools to make the machine run faster. I am interested in the part of the machine that fails.
That gap in the market map? It is not an opportunity that nobody spotted. It is a problem that nobody was paid to care about.
I intend to be the exception.
Resaido is in development. If you work in NZ real estate, conveyancing, or compliance and this problem is familiar to you, I would genuinely like to hear your worst story — reply to this article or find me at @ellenbuilds.