By the editorial team airassvet.ru. This guide was prepared on September 29, 2026. The emergency dispatch service requirements are quoted from the text Rules for Managing Multi-Unit Apartment Buildings, approved by Decree No. 416 of the Government of the Russian Federation dated 05/15/2013, as amended on January 26, 2026 (GARANT system); the document is also published on the Government of the Russian Federation website. Legal sections are for informational purposes only and do not replace consultation with an attorney. The editorial team has no commercial relationship with the services mentioned, and there are no affiliate links.
Short answer. An AI dispatcher for a property management company is an AI agent on the phone and in messaging apps that immediately responds to a resident, logs the request, determines its type and urgency, escalates an emergency to the on-duty dispatcher and crew, and tells the resident the estimated response time. This is directly tied to the requirements of Government Decree No. 416: the emergency dispatch service must answer a call within no more than 5 minutes, and the resident must be informed of the estimated timeline within 30 minutes of the request being logged. AI does not replace emergency diagnosis, crew dispatch, or decision-making in an emergency—it takes routine inquiries off staff members’ plates and helps ensure deadlines are not missed.
Below: which Decree No. 416 requirements apply to request intake, what can be delegated to AI, how to set up emergency escalation, and how to measure results in a pilot.
Contents
- What Decree No. 416 requires from the emergency dispatch service
- Where the dispatcher bottleneck is
- What an AI dispatcher is and what it does
- What to keep with people
- How to set up emergency escalation
- Building knowledge base: without it, AI makes mistakes
- Channels: phone, messaging apps, app
- How to launch an AI dispatcher pilot
- Pre-launch checklist
- Frequently asked questions
What Decree No. 416 requires from the emergency dispatch service
Section IV of the Rules approved by Decree No. 416 is devoted to emergency dispatch services. The management company, homeowners’ association, or cooperative must organize an emergency dispatch service (EDS), including through a contract with a specialized organization (para. 9). The EDS operates 24/7, keeps logs, including in electronic document form (para. 12), and ensures round-the-clock registration of requests and monitoring of their completion within the established deadlines (para. 10).
The deadlines in para. 13 that matter for request intake:
| What the EDS must provide | Deadline under para. 13 |
|---|---|
| Answer to a resident’s phone call | No more than 5 minutes |
| If an answer within 5 minutes is not provided | Contact the caller by phone within 10 minutes after the call, or provide an option to leave a voice and/or electronic message, which must be reviewed within 10 minutes after receipt |
| Informing the resident of the estimated request completion time | Within 30 minutes from the time the request is logged |
| Containment of emergency damage to water supply, wastewater, heating, and power systems | No more than 30 minutes from the time the request is logged |
| Clearing blockages in the building’s wastewater system | Within 2 hours from the time the request is logged |
| Clearing trash chute blockages | Within 2 hours, but not earlier than 8:00 a.m. and not later than 11:00 p.m. when requests are received 24/7 |
| Repairing emergency damage to these systems | No more than 3 days from the date of the emergency damage |
| Minor repairs | 24/7, within the timeframe and scope agreed with the resident |
In addition, in the event of emergencies involving cold and hot water supply, wastewater, and heating systems, the EDS informs the local government authority about the nature of the damage and the estimated repair timeline.
Most of the deadlines under Decree No. 416 are counted from the moment the request is logged, so fast and accurate logging is where an AI dispatcher has the biggest impact on compliance.
Where the dispatcher bottleneck is
Editorial observation based on the structure of the requirements: the EDS has two streams at once. The first is urgent emergencies, where every minute counts and a crew is needed. The second is high-volume routine inquiries: questions about bills and recalculations, planned outages, meter readings, and minor repairs. During peak hours, for example when the water is shut off throughout the building, many residents call with the same question, and the line is busy—but the 5-minute response requirement still applies.
It is precisely the second stream that AI can take over almost entirely, freeing the dispatcher to handle the first.
What an AI dispatcher is and what it does
An AI dispatcher is an AI agent connected to telephony, messaging apps, and the management company’s ticketing system that receives requests in natural language, logs the ticket, classifies it, and routes it to the right handler. Typical set of functions:
- No-wait response. The agent picks up or replies in chat immediately, even when all dispatchers are busy.
- Request logging. Address, apartment, contact information, description of the issue, time of the request—entered straight into the log in electronic form.
- Classification. Emergency, blockage, minor repair, utility service, billing question, complaint.
- Routing. Plumber, electrician, elevator service, accounting, property manager.
- Notification. The resident receives the ticket number and estimated timeline, and in the case of a widespread outage, the same confirmed message about the cause and timing.
- Deadline monitoring. The agent monitors whether the deadline in para. 13 is approaching and reminds the responsible staff member.
How AI agents work in voice channels and what to expect from them, we covered in detail in the article voice AI for call centers.
What to keep with people
The allocation below is an editorial recommendation based on risk level.
| Task | AI | Human |
|---|---|---|
| Receiving the request, logging it, assigning a ticket number | Performs | Checks selectively |
| Classification and routing of routine requests | Performs | Handles edge cases |
| Answers to questions about planned outages, payments, schedules | Answers from the knowledge base | Approves knowledge base content |
| Signs of an emergency: smell of gas, flooding, sparking, no heat in winter | Immediately transfers it to the on-duty dispatcher | Makes the decision and dispatches the crew |
| Incident localization and resolution | Does not perform | Emergency response crew |
| Notifying the local government authority about the incident | Can prepare a draft | Sent by the responsible employee |
| Complaints and conflicts | Registers and forwards | Manager |
The main rule: if there is any sign of a threat to life, health, or property, the AI does not continue the conversation but immediately connects the caller with the on-duty dispatcher and simultaneously creates an emergency request. Speed matters more here than having all the collected data.
How to set up incident escalation
Escalation is the most critical part of the project. The workflow is:
- Stop words and warning signs. A list of phrases that cause the agent to stop collecting data and hand off to a person: "smells like gas," "neighbors are flooding us," "sparking," "stuck in the elevator," "no power in the whole building," and similar expressions.
- Parallel logging. An emergency request is created the moment it is recognized — the logging time is recorded immediately, because the deadlines in section 13 are counted from it.
- Backup channel. If the on-duty dispatcher does not respond within the specified time, the agent calls the next person on the list and notifies the supervisor.
- Message to the resident. The resident hears that the request has been accepted as an emergency and that the on-duty dispatcher will contact them — with no promises of timelines the agent cannot guarantee.
- Decision log. Every action taken by the agent and the time of handoff are recorded: this will help both with incident review and with responding to complaints.
If gas, elevator, and life-threatening issues at the building are handled by separate emergency services or contractors, their phone numbers and handoff process must be in the knowledge base and in the agent script.
Building knowledge base: without it, AI makes mistakes
The AI dispatcher performs well only when it knows the building. Minimum knowledge base contents:
- addresses, entrances, and apartment number ranges; mapping buildings to crews and contractors;
- the building's engineering systems: heating type, trash chute availability, elevators, and related contractors;
- on-duty and emergency service contacts, shift schedules;
- current and planned outages with reasons and deadlines;
- standard answers about payments, recalculations, and meter readings approved by accounting;
- rules for classifying requests and deadlines for each type.
It is convenient to build this kind of base on the RAG principle: the agent searches for the answer in the management company's documents and relies on what it finds, rather than on the model's general knowledge. We discussed how a corporate knowledge base for an AI agent is structured in the article corporate knowledge base with an AI agent.
Channels: phone, messengers, app
- Phone — the main A&D service channel, and the 5-minute response requirement applies to it. A voice agent reduces queue time, but it requires high-quality speech recognition for addresses and last names.
- Messengers — convenient for photos of leaks or damage and for mass notifications. About connecting an assistant to different messengers — in the article AI assistant for WhatsApp, VK, Telegram, and MAX.
- App or personal account — if one already exists, the agent can be integrated there so requests go into the same log.
Important note: Government Decree No. 416 allows that if it is impossible to answer within 5 minutes, the resident may leave a voice or electronic message, which is reviewed within 10 minutes. The Rules do not explicitly say whether an AI agent's response counts as an "answer to a phone call" within the meaning of section 13. The safest position is to treat AI as a way to quickly log and forward the request, while compliance with the section 13 deadlines is still monitored as a service obligation.
How to launch an AI dispatcher pilot
A process you can repeat in any management company:
- Choose part of the flow. For example, one or two buildings, or just messengers, or only off-hours.
- Record baseline metrics for 2–4 weeks. The share of calls answered within 5 minutes; the share of requests where the resident was informed of the timeline within 30 minutes; how many requests were missed; the distribution of requests by type. These metrics directly match the requirements of section 13.
- Build a test set. 100–200 past requests, anonymized, including mandatory emergency cases and unclear wording.
- Test the agent on the set before launch. For each case: whether the type was identified correctly, whether the address was recognized accurately, whether emergency escalation was triggered, whether the agent promised a deadline that did not exist.
- Launch with oversight. During the first few weeks, the dispatcher reviews all requests created by the agent, and emergency scenarios are regularly checked with test calls. Compare results with the baseline and expand coverage when classification and escalation errors are isolated and understood.
Pilot size is our recommendation; for a large management company, the test set should be built across different building types. More on metrics — in the article metrics for AI agent effectiveness.
Launch checklist
- The deadlines in section 13 of Government Decree No. 416 have been translated into control rules for each request type.
- A list of emergency signs and stop words has been compiled and checked for immediate handoff to the on-duty dispatcher.
- A backup escalation channel is configured if the on-duty dispatcher does not respond.
- The request logging time is recorded automatically in the log.
- The knowledge base contains addresses, engineering systems, contractors, on-duty contacts, and current outages.
- Standard answers about payments are approved with accounting, and texts about widespread outages are approved with the manager.
- It is defined where resident data is processed and which model is used.
- Baseline metrics and a test set of requests have been recorded.
- A person responsible for sample request review and test emergency calls has been assigned.
Frequently asked questions
What can an AI dispatcher do in a management company?
It answers residents by phone and in messengers, logs requests, determines their type and urgency, forwards them to the person responsible, tells the resident the request number and planned timelines, answers standard questions from the knowledge base, and reminds responsible staff about deadlines that are approaching.
How many minutes does the emergency dispatch service have to answer a call?
Under section 13 of the Rules approved by Government Decree No. 416, no more than 5 minutes. If an answer is not provided within that time, you must contact the resident by phone within 10 minutes or allow them to leave a voice or electronic message, which is reviewed within 10 minutes.
Within what time must the resident be informed of the request completion timeline?
Within 30 minutes from the moment the request is logged. Other deadlines in section 13 are also counted from the time of logging, for example emergency damage localization within no more than 30 minutes and clearing sewer blockages within 2 hours.
Can AI fully replace a dispatcher?
No. AI handles routine incoming requests and speeds up registration, but decisions in emergency situations, dispatching the crew, localizing and resolving incidents, and notifying the local government authority are handled by people.
Where should you start when implementing an AI dispatcher?
Start with baseline metrics for Section 13 timeframes, a test set of past inquiries, and one channel or several buildings. The agent is tested on the dataset before launch, and during the first few weeks the dispatcher reviews every request it creates.
How the airassvet.ru team helps implement an AI dispatcher in a property management company
If the dispatch office cannot keep up with the volume of requests and the Section 13 deadlines are hard to meet, the team airassvet.ru can:
- audit the request intake process: channels, types of inquiries, where deadlines are missed, and how registration time is recorded;
- build a knowledge base for the buildings — addresses, building systems, contractors, on-call staff, outages — so the agent can answer using it;
- develop an AI agent for phone systems and messengers, integrate it with your request tracking system, configure emergency escalation, and test it against a set of inquiries.
The first step is to choose one process, for example after-hours request intake, document how it works now, what data and constraints it has, and what criterion will determine whether the pilot is successful. If the process is chosen, discuss the project.
Conclusion
AI in housing and utilities delivers value where a property management company has a bottleneck — in accepting and registering requests. Government Decree No. 416 requires answering a call to the dispatch service within no more than 5 minutes, informing the resident of the timeline within 30 minutes from registration, and using that registration time to count incident resolution deadlines. An AI dispatcher answers without delay, registers and classifies requests, communicates deadlines, and prevents them from being missed, while emergencies are immediately handed off to people. Start with baseline metrics for Section 13 timeframes, the building knowledge base, and a pilot on part of the request flow.