Blog

Answering real-estate inquiries automatically: how agents use AI to qualify, prioritise and book viewings in under 5 minutes (2026)

By Daniel Pietschmann and Bilton Taheri, founders of Realto. Last updated: 20 September 2026
Methodology: The guide draws on publicly cited studies (see table), on the product practice of an AI operating system for German agencies, and, once added, on original founder interviews. Source links sit in the table.

Why response time decides the mandate

In real-estate marketing, the prettier exposé rarely decides who wins the viewing or the mandate, the first serious reply does. Anyone who inquires about an apartment on ImmoScout24 or Immowelt has often written to several agencies at once. The first coherent response structures the conversation: availability, next step, appointment. Everything that arrives later competes with a relationship that is already underway.

The observation is not a new PropTech talking point. Harvard Business Review described as early as 2011 that a lead contacted within five minutes is about 21 times more likely to be qualified than after 30 minutes. Mapped onto the German agency market, that means: if you only empty the inbox in the morning and evening, you give away most of the qualification chance before the working day starts.

One figure often used in the industry as evidence of the same effect: widely cited according to NAR, 78% of buyers hire the agent who responds first.

Service quality still lags that window. The DISQ service study on real-estate agents shows that about one third of email inquiries to German agents are never answered, and the average wait for a reply is about 40 hours. In a market where attractive listings generate 40-150 inquiries in the first 48 hours, that is not a cosmetic flaw, it is a systematic leak in the pipeline.

The table below collects only metrics with a publicly reachable source. It is deliberately short: no extra, rounded or internally “felt” figures.

To quote or double-check the same figures, use Real-estate agents & AI: figures, studies, sources.

For daily work that implies a tight loop: qualification likelihood drops steeply once minutes become hours. A reply on the next working day is not “solid follow-up” under that model, it is often the end of the case. Taking that seriously means treating the first response as a system job, not a heroic act by whoever happens to be online.

Why most inquiries arrive at night

Portals are not an office-hours channel. 62% of all online inquiries arrive after 6 p.m. exactly when a classic agency shuts the laptop, wraps viewings or clocks off. Prospects do the opposite: they scroll after work, compare listings at the kitchen table, send three messages at 10 p.m. and expect “someone to still be there”.

In practice many offices describe the reality as “80 inquiries per listing at 2 a.m.”: a listing goes live, the algorithm distributes it, and the wave does not spread across the week, it arrives in one night. Anyone who then answers with copy-paste at 9 a.m. is talking to people who already held two other appointments or paused the search.

The economics are simple, without adding another study row: a typical market commission on the order of about €5,000 per closing sits against the cost of a lost inquiry, lost not because the property was weak, but because the reply was late, incomplete or missing. A single extra qualified viewing that closes funds a software subscription for years. Conversely, every night without a system is a quiet decision to leave part of the pipeline unread.

That is why “we reply quickly when we are in the office” is not enough. Inquiry handling has to be decoupled from staff being awake, without the prospect feeling fobbed off with an empty autoresponder line.

Night shifts do not solve it. They postpone burnout and make quality depend on who is on duty. An operating system uses the same rules at 2 p.m. and 2 a.m.: assign the property, ask for missing fields, offer a slot, flag exceptions. Humans join when judgement is required, not to type the first polite line.

Handover inside the office changes too. Instead of opening an unsorted inbox in the morning, the team starts from a prioritised list. That is the real culture shift: fewer hero stories about “I was already in the mails at six”, more visible cases that kept moving overnight.

Four ways to handle inquiries: an honest comparison

There is not “the one AI”, but four operating models agencies actually use. The difference is not the marketing label, but response time, depth of qualification, and whether a calendar entry exists at the end.

Manual email plus Excel works as long as the office stays under about ten inquiries a month. Beyond that, quality drops: messages are batched, tone gets shorter, follow-up stalls. Autoresponders look fast but do not qualify and annoy people who asked a concrete question about the property. Website chatbots only help the share that already lands on your own site, not the portal inbox where most German agent inquiries arrive.

An AI operating system starts at the inbox: the same case across ImmoScout24, Immowelt, email and messenger, property-specific rather than generic, aiming to leave a qualified appointment in the calendar. That is not a guarantee of every closing, but it is the only of the four paths that scales response time and qualification at once.

An honest comparison also names switching costs. Staying manual costs no licence, but it costs calendar time, nerves and lost first contacts. Introducing templates is cheap and fast, and trains prospects that the office is not listening. A website chatbot helps if you truly have your own traffic; for the typical portal-led agent it stays a side stage. The inbox approach needs a one-time connection and clear rules, and then delivers the same standard on every channel.

How AI inquiry handling works in 5 steps

Technically, an operating system differs from a chat widget because it runs cases, not just sentences. The flow breaks into five visible steps, the same steps an office can use as an internal playbook.

1. Connect the portal inbox

   ImmoScout24, Immowelt and other inbound channels (email, WhatsApp) are connected to one system that monitors the inbox around the clock.

2. Understand the inquiry and assign the property

   The AI reads the message, recognises the listing and channel, and assigns the case to the right property and thread.

3. Qualify the prospect

   Missing details are collected: budget or financing, timeline, self-disclosure form and other property-specific criteria.

4. Book the viewing

   Qualified prospects receive time slots and are booked directly into the agent’s calendar.

5. Hand over to the agent

   The agent only sees prioritised cases and handles exceptions; follow-ups continue in the system.

Step 1 is organisationally the most important: as long as messages sit in five mailboxes, there is no shared truth. Step 2 stops the AI answering “an apartment in Frankfurt” when the inquiry is about the penthouse in Sachsenhausen. Step 3 is the real lever against unqualified viewings. Step 4 turns a dialogue into an appointment. Step 5 keeps the human in the loop without turning them into a typist.

In practice: an inquiry arrives at 02:14. Within seconds a property-specific reply goes out, including the next missing piece of information. Once budget, timeline and self-disclosure are in, free slots are offered. The next morning the agent does not see 80 raw emails, but a list: qualified, appointment booked, exception (special request, clash, unclear financing).

Order matters. Anyone who books viewings first and only then checks financing or the self-disclosure form fills the calendar with trips the property cannot support. Anyone who first understands and assigns saves the whole office journeys. The five steps are therefore not a marketing funnel, but the minimum discipline without which automation only produces faster mess.

Handover does not mean the agent never sees the thread again. It means they see it when a human must decide: price cut, a pet despite a ban, a clash with a second prospect, an unclear owner brief. Everything else can stay in the system, including a polite nudge when a document is missing.

The qualification question set for agents

Qualification is not an interrogation and not spam. It is the minimum information without which a viewing wastes time, for both sides. Modelled on BANT (budget, authority, need, timeline) plus property fit, a short, polite dialogue is usually enough.

Budget and financing

Sales: price range, equity, whether a financing confirmation or at least a bank conversation exists. Lettings: household net income versus rent, type of employment, whether a guarantor is needed. If you do not ask, you schedule people who cannot carry the property, and block slots for those who can.

Timeline

“From when?” separates window-shoppers from decision-makers. Someone who must move in 14 days is a different case from someone looking “sometime in 2027”. The AI can stage follow-ups: move-in date, notice period on the current home, whether an interim solution exists.

Pets and smoking (lettings)

In many house rules these points decide yes or no before a key is handed over. Clearing them early saves everyone a trip. They belong in the catalogue because they are property-specific, not to exclude people, but because the listing and the owner’s brief require them.

Self-disclosure form

In lettings the Selbstauskunft is the standard filter. Automation here means: send the link or form, watch for arrival, follow up politely, do not close the case while the document is missing. Without the form step, calendars stay full and files stay empty.

Seriousness

Seriousness shows in concrete answers, willingness to book a time, and whether the prospect even knows the property (rooms, location, price). An AI can flag contradictions (“looking for 2 rooms but asking about the 5-room house”) without sharpening the tone. The agent decides who still gets a chance. Judgement stays with the agent, not with algorithm dogma.

The catalogue belongs in writing in the office, not only in a senior colleague’s head. Otherwise the AI (or whoever is covering) answers differently every time. A shared set: budget or financing, timeline, property house rules, self-disclosure, seriousness. Cover becomes possible and the tone stays polite, because nobody has to improvise on the tenth night-time email.

What does not belong in the catalogue: idle curiosity, questions the listing already answers, or filters the office cannot legally stand behind. Qualification is the opposite of arbitrariness: a few explainable criteria that fit the property.

What 100+ conversations with German agencies showed

Realto’s founders interview German agencies. Method, interview count and original findings will appear under the heading once they are ready, in the visible body, not only internally.

GDPR, the AI Act and legal certainty

Automated inquiry handling processes personal data: name, contact details, often income, household size, schedule. For a German agency three building blocks are non-negotiable.

First: EU hosting. Personal data should be processed in the European Union, not in an opaque third-country pipeline without appropriate safeguards. Second: a data-processing agreement (AVV/DPA) with the provider. Without it, the legal basis for processing on behalf of the agency is missing. Third: transparency. Prospects must understand who they are writing to.

Since 2 August 2026, the transparency obligations in Art. 50 of the EU AI Act apply: chatbots and comparable systems must disclose that they are AI. Pretending that the owner is personally typing at 2 a.m. does not only break trust, it breaches a duty under Union law. A clean system states in the first reply that AI is assisting, and offers a path to a human.

Legal certainty also means: no invented property promises, no silent discrimination via opaque filters, logging of cases, and a way to escalate a conversation to the agent. Technology does not replace the controller’s responsibility under the GDPR.

In practice every office should have three documents to hand: the DPA/AVV, a short internal description of data flows, and the wording of the AI disclosure in the first message. Art. 50 AI Act is not a footnote for the legal team, it changes the visible dialogue with the prospect. Hiding it does not save work; it buys an explanation later. More on hosting, DPA and Art. 50: AI in the brokerage, GDPR and the AI Act.

Cost & ROI

The market for agent software is wide: depending on the stack, publicly available comparisons put costs at €39-700+ per month. Simple AI tools often start in the lower band (around €39-99); a full stack with CRM, portal connection, calendar and automation can reach or exceed the upper end.

ROI is not “AI hype”; it is avoided idle time. If a typical closing commission in the market is about €5,000, one additional held case (a night in which 80 messages are not left unanswered) can carry software costs for a long period. The second lever is qualification: fewer viewings without property fit, more calendar time for advice and closing.

What should not pay off: five tools that never unify the same inbox. An operating system saves not only licence lines, but context switching.

Count cases internally, not gut feel: how many messages sit untouched after 6 p.m.? How many viewings end without property fit because budget or timeline were never asked? Those questions need no new industry studies, they need your own inbox. The table above frames the market; the office does the arithmetic.

FAQ