For developers, AI is most useful where buyers are currently being lost: in the first response to an inquiry, in qualifying prospects under the updated Family Mortgage rules, in matching an apartment to a real budget, in re-engaging a “cold” lead database, and in monitoring sales team calls. When sales are down and the terms of subsidized mortgages from October 1, 2026 depend on the number of children and the region, every lead costs more and calculation mistakes by managers happen more often. An AI agent responds in seconds and asks the right questions, but mortgage approval still belongs to the bank, and deal terms still belong to the developer.
Prepared from open sources by AIrassvet and verified on October 6, 2026. It is intended for commercial directors, heads of sales and marketing at developers, as well as real estate developers selling apartments under shared construction agreements (DDA). This article is not financial or legal advice: rates and limits are based on media publications as of the verification date; final terms are determined by the bank and the program. AIrassvet is not affiliated with the organizations mentioned, and the section at the end describes the publisher’s services.
Contents
- What’s Happening with New-Build Sales in Fall 2026
- Family Mortgage from October 1, 2026: What the Sales Team Needs to Know
- Quick Take: Where AI Helps a Developer’s Sales Team
- First Response: Lead Reply and Qualification
- Matching an Apartment to the Budget: Mortgage and Installment Plans
- Re-engaging the Database and Monitoring Calls
- Deal Documents and DDA-Related Responses
- Risks: What AI Should Not Be Trusted With in Apartment Sales
- Q4 Implementation Plan
- Which Metrics to Track
- Frequently Asked Questions
- Primary Sources and Verification Method
What’s Happening with New-Build Sales in Fall 2026
The primary housing market entered the fall with weakening demand. According to the DOM.RF Analytics Center (weekly digest for September 18–25, 2026), new-build sales in August fell 16% year over year to 1.8 million sq. m worth ₽357 billion. Since the start of the year, 14.6 million sq. m have been sold (+2% YoY), but without one-off spikes in January and June, sales were 10% below last year’s level (DOM.RF digest as paraphrased by Smart-Lab, September 29, 2026).
From the same digest, the indicators that directly affect the sales team are:
- sales pace vs. construction readiness ratio — 68% as of September 1, 2026, compared with 73% at the start of the year;
- unsold housing absorption period — 3.2 years;
- installment plans — in Q2 2026, 25% of DDA transactions were completed with installment plans, and 94% of developers offer contracts with installment payments;
- project debt coverage by escrow accounts — 65%, with 28 regions below 70%.
The outlook is cautious. Based on an estimate cited by TASS, new-build sales in 2026 could decline by 30–35% year over year if borrowing remains expensive; in Moscow and the Moscow region, DDA sales from January through August fell by nearly 25%, and in August alone by 31% (TASS as paraphrased by Smart-Lab, September 11, 2026). On September 11, the Bank of Russia kept the key rate at 14% (ALLE, September 16, 2026).
What this means in practical terms: lead flow is getting more expensive, while the share of buyers who can actually close is shrinking. The sales team has to respond faster, screen more accurately, and nurture those not ready to buy today more patiently. This is exactly the kind of work where AI removes routine tasks from managers.
Family Mortgage from October 1, 2026: What the Sales Team Needs to Know
As of October 1, 2026, the Family Mortgage program is no longer a single 6% program. According to RT, the rate and loan limit now depend on the number of children and the region, subsidies are limited to 15 years, and one of the requirements is a child under seven years old; the new terms apply only to loans originated on or after October 1 (RT, October 1, 2026).
| Number of children | Moscow, Moscow Region, St. Petersburg, Leningrad Region: limit / rate | Other regions: limit / rate |
|---|---|---|
| 1 | up to ₽12 million / 12% | up to ₽6 million / 10% |
| 2 | up to ₽15 million / 10% | up to ₽8 million / 8% |
| 3 | up to ₽18 million / 8% | up to ₽10 million / 6% |
| 4 | up to ₽18 million / 6% | up to ₽10 million / 4% |
| 5 or more | up to ₽18 million / 4% | up to ₽10 million / 2% |
Table source — RT, October 1, 2026. According to Lenta.ru, the 6% rate also remains available to borrowers making a down payment of at least 50% from their own funds, regardless of the number of children (Lenta.ru, October 1, 2026). Before closing a deal, the terms should be checked with the bank: media publications are not a regulatory document.
It’s easy to calculate how much the payment changes. AIrassvet calculation using the annuity formula over 30 years: a ₽6 million loan at 6% results in a monthly payment of about ₽35,973, while at 10% it is about ₽52,654. For ₽12 million at 6%, it is about ₽71,946, and at 12% — about ₽123,434. This is an illustration for the subsidized period, not a bank offer: after 15 years, the rate may change.
For the sales team, this leads to three conclusions:
- The question “what rate do you have” no longer has a single answer. You need to know the region, the number of children, the age of the youngest child, and the size of the down payment.
- Some buyers have lost access to the subsidy or ended up with a payment that is one-and-a-half to two times higher. They need a different path: a smaller apartment, an installment plan, or a delayed purchase.
- Managers make mistakes under complex terms. A mistake in the first conversation can cost trust and, at times, the deal.
In Brief: Where AI Helps a Developer’s Sales Team
| Task | What AI Does | How It Helps Right Now | Who Makes the Decision |
|---|---|---|---|
| Answering an Incoming Lead | Responds in chat, messenger, or by phone at any time | The buyer doesn’t go to a competitor while waiting for a call | Manager After Handoff |
| Qualification | Clarifies budget, region, family size, timing, and payment method | The manager gets a “warm” buyer with a clear plan | Sales Manager |
| Apartment Matching | Selects units from the current price list based on budget and monthly payment | Fewer calls asking, “What do you have under 8 million?” | Buyer and Sales Manager |
| Preliminary Estimate | Calculates an estimated mortgage payment or installment plan | The buyer immediately sees the real monthly burden | Bank and Developer |
| Re-Engaging the Database | Finds “cold” buyers whose situation has changed | Brings back part of the database without ad spend | Sales Manager |
| Call Monitoring | Transcribes and analyzes conversations against a checklist | Shows where deals are lost and what is being promised outside the rules | Head of Sales |
| Documents and Questions About the Shared Equity Construction Agreement (DDU) | Answers from the knowledge base: escrow, timelines, document package | Fewer repetitive questions for lawyers | Developer’s Attorney |
Table — editorial classification by AIrassvet; the order roughly follows the implementation sequence.
First Line: Lead Response and Qualification
A first-line AI agent is a program built on a language model that talks with the buyer according to the developer’s rules, logs answers in the CRM, and hands the qualified lead to a sales manager. Unlike a button-based bot, it understands free-form text: “a two-bedroom under nine million, we have two kids, the younger is three.”
New-build buyers message in the evening and on weekends, comparing several residential communities at once. The one who responds first and on point gets the conversation. For more on channels, see the article on an AI assistant in WhatsApp, VK, Telegram, and MAX and on voice AI for call centers.
What Questions the Agent Asks
Qualification after October 1 is built around financing terms:
- purchase region and the residential community of interest;
- budget and comfortable monthly payment;
- payment method: cash, mortgage, installment plan, or sale of another apartment;
- for a family mortgage — number of children and whether there is a child under seven;
- size of the down payment;
- the time frame when the buyer is ready to make a decision.
The agent does not ask for documents or collect extra data: only what is needed for matching and handoff to the manager. Answers are saved in the deal card — there is a separate breakdown of the CRM connection in the article on an AI agent for Bitrix24 and amoCRM.
Where the First Line Ends
The agent passes the conversation to a human when the buyer is ready for a showing or reservation, asks a question outside the knowledge base, argues about price, or complains. The rule is simple: the agent can state terms approved by the developer and cannot change them — discounts, special conditions, and timeline promises stay with the sales manager.
Apartment Matching by Budget: Mortgage and Installments
After changes to the family mortgage program, many buyers need not an “ideal apartment,” but an apartment that fits the payment. An AI agent connected to the current inventory grid and price list can:
- match units that fit the program limits for a specific family;
- show an estimated mortgage payment at a given interest rate;
- show an installment plan option according to the developer’s rules;
- suggest alternatives: smaller size, a different building, a later phase.
Installments have become a mass-market tool: according to a survey by DOM.RF, IRSO, and VCIOM, they accounted for 25% of DDU transactions in Q2 2026. But in the same DOM.RF digest, the risk is noted: as of September 1, 17% of the value of active DDUs had not been paid, and in August 2026 a bill was introduced that, among other things, proposes limiting installment sales to 25% of a project’s floor area without bank consent (DOM.RF digest, September 29, 2026). That is why the agent takes installment terms only from the developer-approved table and does not invent “custom” schedules.
Important: all agent calculations are estimates. This must be stated clearly to the buyer, and the final terms are confirmed by the bank or the contract.
Re-Engaging the Database and Call Monitoring
Every developer has a database of buyers who inquired but did not purchase. As of October 1, terms worsened for some of them and improved for others: for example, for families with three or more children in the regions, the limit increased from 6 million to 10 million rubles, and for families with four or more children, the rate also dropped — to 4% and 2%. AI can review the CRM interaction history, identify those for whom the change in terms is a reason to reach out again, and prepare a personalized message for the sales manager.
There is a legal boundary here: promotional mailings require the recipient’s consent. More on this in the article about fines for AI spam mailings.
The second scenario is call quality monitoring. AI transcribes sales conversations, checks them against a checklist, and finds:
- missed steps: budget not clarified, no showing offered;
- incorrect terms: the old 6% rate for everyone, wrong limits;
- promises outside the policy: discounts, move-in dates, “approval is guaranteed”;
- unhandled objections and lost leads.
The sales manager doesn’t get 5% of calls listened to — they get a summary of all of them. Learn more in the article on AI Sales Quality Control; for training new managers, AI training simulator.
Deal documents and answers on the equity participation agreement (DPA)
Buyers ask the same questions over and over: how escrow works, when they can get the keys, what documents are needed to reserve a unit, and what happens if the project is delayed. Lawyers and managers answer these questions dozens of times a day.
The solution is a RAG system: the AI answers only from the developer’s approved documents — the standard equity participation agreement, project declaration, reservation policy — and shows which document the answer is based on. If the answer isn’t in the database, the question goes to a lawyer. How these systems work and where they fail is covered in the materials on a RAG system for business and RAG accuracy and hallucinations.
The second document-focused scenario is package verification: OCR of passports, children’s birth certificates, certificates, and checking that the package for reservation or a mortgage application is complete. This saves time, but it requires special care with personal data.
Risks: what AI should not be trusted with in apartment sales
| Risk | How it shows up | How to reduce it |
|---|---|---|
| Incorrect program terms | The agent quotes an outdated rate or someone else’s limit | Terms come only from a table updated by the responsible employee; the response includes the update date |
| Promise of approval | “You’ll definitely be approved” | Prohibited in the agent guidelines; only the bank can approve |
| Invented deal terms | A discount, installment schedule, or completion date that doesn’t exist | The agent does not change prices or terms; anything unusual goes to a manager |
| Personal data | Family documents are sent to an external service without a legal basis | Consent for processing, minimum necessary data, model in a closed environment, or anonymization |
| Pushiness | Repeated messages without consent | Mailings only with consent, clear unsubscribe option |
Buyers’ personal data is a separate issue: family structure, income, children’s documents. Its processing must comply with Federal Law No. 152-FZ. If the data cannot be sent to an external cloud service, a model in a private environment or anonymization is used — the options are covered in the article “Can Customer Data Be Sent to AI?”.
Q4 implementation plan
Fall is a good time: the new mortgage terms have just taken effect, buyers are asking many questions, and by the December deal season the sales team can be working in a tested mode.
- Week 1. Funnel audit. Count how many inquiries come in outside business hours, how many minutes it takes to respond, and how many are lost without a reply.
- Week 2. Knowledge base. Compile a table of mortgage terms by partner banks, installment rules, pricing, and unit availability, plus answers to common questions about the equity participation agreement. Assign someone to keep it updated.
- Weeks 3–4. Frontline pilot. The agent responds in one channel for one residential complex, and the manager reviews every conversation.
- Weeks 5–6. CRM qualification. The agent’s answers are recorded in the customer card, and the manager receives the lead with fields filled in.
- Weeks 7–8. Call monitoring. Add transcription and a checklist, and compare the AI’s scores with the sales manager’s scores on a sample.
- December. Scale-up. Add channels and residential complexes, and launch consent-based database outreach.
What metrics to track
| Metric | How to calculate | Why it matters |
|---|---|---|
| First response time | Median minutes from inquiry to the first substantive response | The key frontline performance indicator |
| Share of inquiries without a reply | Inquiries without a reply per day / all inquiries | Shows how many are lost at night and on weekends |
| Share of qualified leads | Leads with budget, payment method, and timeline filled in / all leads | Quality of handoff to the manager |
| Conversion to showing and reservation | Showings and reservations / qualified leads | Connection between the agent’s work and sales |
| Errors in terms | Calls and chats with an incorrect rate or limit / reviewed conversations | Control of the most serious risk |
| Share escalated to a person | Conversations escalated to a manager / all agent conversations | Whether the agent is too independent or too helpless |
Capture baseline values before the pilot — without them, the impact can’t be proven.
Frequently asked questions
What can AI do in a developer’s sales department?
AI responds to inquiries in chat, messaging apps, and by phone at any time, qualifies the buyer by budget, payment method, and family mortgage terms, selects apartments from the current price list, estimates the monthly payment, reviews managers’ calls against a checklist, and answers standard questions about the equity participation agreement and escrow using the developer’s knowledge base.
How did family mortgage change as of October 1, 2026?
According to RT as of October 1, 2026, the rate and loan limit now depend on the number of children and the region: from 2% to 12% annual interest, with limits from 6 million to 18 million rubles. Subsidies are capped at 15 years, and a child under seven is required. The new terms apply to loans originated on or after October 1, 2026. Final terms should be confirmed with the bank.
Can an AI agent calculate a mortgage for a buyer?
It can estimate an approximate payment based on a given rate and loan amount and match an apartment to the budget. The bank makes the approval decision and sets the final loan terms, so the agent must clearly state that the calculation is preliminary and never promise approval.
Will AI replace new construction sales managers?
No. AI takes over initial responses, lead qualification, and routine work for managers, while a person handles showings, price negotiations, nonstandard terms, and closing the deal. The goal is for the manager to spend time on buyers who are ready to purchase.
Is it safe to share buyer data with AI?
It depends on where the model runs and what data is shared. Family composition, income, and documents are personal data, and their processing must comply with Federal Law No. 152-FZ. For sensitive data, a company-hosted model or data anonymization is appropriate; the agent should collect only the data needed for matching buyers with properties.
Where should a developer start when introducing AI in the sales department?
Start with a funnel audit: how many inquiries come in after hours and how long it takes to respond to them. Then build a knowledge base with up-to-date mortgage and installment plan terms and launch a pilot first-line response in one channel for one residential development.
Primary sources and verification method
Verified on October 6, 2026:
- DOM.RF analytical center digest for September 18–25, 2026, retold by Smart-Lab — sales, sell-through rate, installment plans, escrow, installment plan bill. The primary source is listed on the page as DOM.RF's Telegram channel.
- TASS, retold by Smart-Lab, September 11, 2026 — sales forecast for 2026, DDU sales in Moscow and the Moscow region.
- RT, October 1, 2026 — family mortgage rates and limits by number of children and region.
- Lenta.ru, October 1, 2026 — keeping the 6% rate with a down payment of at least 50%, applying only to new contracts.
- ALLE, September 16, 2026 — the Bank of Russia's decision on the key rate on September 11.
The DOM.RF and TASS figures are taken from retellings that match the cited primary sources; family mortgage terms come from media publications as of the verification date. Payment calculations use the standard annuity formula and are for illustration only, not a bank offer. The scenario, risk, and metric tables and the implementation plan are AIrassvet's editorial analysis; the article does not include its own testing or client results.
How AIrassvet helps developers with new construction sales
If the sales team can't keep up with inquiries, gets confused by new mortgage terms, or can't see where deals are falling through, the problem can be broken down step by step. On the AIrassvet website, which also uses the name "Vladimir Lomtev's AI Studio," the following applicable solutions are described:
- An AI agent for messengers and telephony, integrated with the developer's CRM, that responds to inquiries and qualifies buyers according to the company's rules.
- A RAG system based on the developer's documents — mortgage and installment terms, standard DDU, reservation rules — to answer buyers and managers with source citations.
- Deployment of a language model within the company's infrastructure to work with buyers' personal data.
- A predictive model based on CRM history that identifies buyers worth re-engaging.
The first step is to choose one process, such as responding to after-hours inquiries, and define its current metrics, data sources, constraints, and the criterion for considering the pilot successful.
Bottom line: every inquiry counts
In fall 2026, developers are selling in a market shaped by weakening demand, expensive market-rate mortgages, and a new, more complex family loan program. In this environment, the winner is not the one with the most advertising, but the one that responds to every buyer faster and more accurately.
AI handles the first line, qualification under the new terms, matching apartments to payments, call review, and standard DDU questions. The manager gets pre-qualified buyers, and the sales lead gets a clear view of every conversation. The best place to start is with inquiry responses and an up-to-date knowledge base, while leaving loan approval and deal terms to the bank and people.