The Seller Asked My Bot: “Who Are You?”

A property owner somewhere in central Israel got a WhatsApp message from a broker last month. Nothing unusual about that — a polite note asking whether she’d thought about selling her apartment, because a buyer was looking on her street. She answered. Within seconds, she got a reply. She mentioned the apartment was renovated; the reply asked how many rooms. She said she wasn’t sure she wanted to sell at all; the reply asked, gently, whether she’d like to hear a price estimate before deciding.

Then she typed the question that made me sit up: “Who are you, anyway?”

And the system I built answered, without hesitating, with the broker’s own name.

I build WhatsApp automation for Israeli small businesses for a living (my studio is called Achiya Automation), and this bot — a lead-qualification assistant for a real-estate broker — is one of mine. The broker reviews every conversation, sets every rule, and personally handles everything after the first few questions. But in that moment, a machine introduced itself as a person. The exchange has been sitting with me ever since, and I think it’s a question the whole Israeli services economy is about to face — sooner than most people think.

Why brokers are doing this

Start with the un-romantic reason: speed, because speed is where deals live or die.

Harvard Business Review published a study that salespeople have been quoting for years, and it has aged remarkably well. Researchers audited 2,241 U.S. companies by sending each a web lead and measuring the response. The average response time was 42 hours. Twenty-three percent never responded at all. And the firms that made contact within one hour were nearly seven times more likely to qualify the lead than those that waited even an hour longer — over sixty times more likely than those who took a day.

Now map that onto an Israeli broker’s actual afternoon. He is showing an apartment on the third floor with no reception. A seller he messaged two weeks ago suddenly replies at 21:40: “Actually, maybe. What do you think it’s worth?” By the time the showing ends and the phone comes out of the pocket, the seller has talked herself out of it, or a competitor has answered first. The lead didn’t die because the broker was lazy. It died because he was working.

The bot exists for exactly that window. It answers in seconds, at any hour. It asks the three or four questions a broker would ask first anyway — which street, renovated or not, how many rooms, how many square meters — and it can offer a ballpark figure (“around two million; I’d need to see the apartment to be precise”) built from current market data. It never schedules a visit and never negotiates; the moment a conversation turns real, a human takes over. It is, in effect, the broker’s front desk — one that happens to never sleep.

What the machines still get wrong

Before anyone panics about being replaced: the current generation of these assistants is impressive and also reliably, sometimes hilariously, dumb.

Early on, this one asked the owner of an industrial property — a logistics warehouse — how many rooms it had. Three times. The question ladder was built for apartments, and the software marched down it with total confidence. A human agent would never ask a warehouse owner about rooms; software has no embarrassment reflex. We fixed it, but only after a real conversation exposed it.

We also learned that speed itself can betray the illusion. Real people don’t reply in 900 milliseconds, and when a seller fires off two quick messages — “no” and then “actually, make me an offer” — a naive bot answers each one separately and contradicts itself. We ended up teaching the system to wait several seconds before responding, like a person finishing reading. There is something almost comic about engineering a machine to be slower so that it feels more honest.

That word — honest — is the real subject here.

The question with a deadline attached

When the seller asked “who are you?”, the bot gave the broker’s name because that is what the broker wanted, and because nothing in Israeli law says otherwise. Our spam amendment governs consent — who you may message and when — not identity. Nothing in the statute requires software to admit it is software.

Europe has just decided differently. On August 2, 2026 — nine days ago — Article 50 of the EU AI Act became applicable. It says, among other things, that a person interacting with an AI system must be told they are talking to a machine, clearly, at the first point of contact. Not in a terms-of-service page. Fines run up to €15 million or 3% of global turnover. Israeli brokers are not bound by Brussels, but anyone who remembers how GDPR quietly became the default expectation of Israeli customers can guess where this goes.

And beneath the regulation is a more interesting human question: where exactly is the line? If a secretary answers the broker’s WhatsApp and signs with his name, nobody calls it deception. If software does the same thing — asking the same opening questions, under the same supervision, with the same human taking over for anything that matters — is it different? I genuinely hear both answers from business owners, and both sound reasonable. My own discomfort in that “who are you?” moment tells me something, though. The technology has gotten good enough that disclosure is no longer a performance cost — the seller kept chatting happily either way — which means hiding the bot is becoming a choice rather than a necessity. Choices, unlike necessities, get judged.

My bet: within two or three years, “you’re talking to my digital assistant, I’ll jump in personally in a moment” will read as professionalism, not as an admission of weakness — the same way “this call may be recorded” stopped scaring anyone decades ago. The brokers who adopt that sentence early will get the trust dividend. The ones who wait for an Israeli version of Article 50 will adopt it anyway, just without the credit.

The seller from the beginning of this story, by the way, got her price estimate, thought about it for a day, and scheduled a visit — with the human. That part, no one is automating.

So here’s what I want to know, from both sides of the conversation: if you found out that the “broker” who answered your late-night WhatsApp in twelve seconds was software — and a human closed the deal beautifully afterward — would you feel served or deceived? And if you’re a broker or business owner already running an assistant like this: have you tried disclosing it upfront, and did your reply rate actually drop? I read the comments; the machine doesn’t.