Chatbots for business: which tasks to hand over to a bot, from requests to stock control and reports

Chatbots for business are programs inside a messenger that let an employee or customer do a work task without a separate app: submit a request, record a delivery of goods, get a morning summary, approve a payment. They deliver the most value inside the company. There they replace phone calls, photos posted to a group chat, and data copied by hand into spreadsheets and 1C, the accounting and ERP platform most Russian companies run on. Customer scenarios come second and follow the same rule: a bot is useful exactly to the extent that it is connected to the system of record.

Below, we go through the internal scenarios in order: employee requests, warehouse and stock, the owner's morning report, approvals, newcomers' questions, and drivers and staff at retail locations. Then come customer tasks and the point where a bot builder stops being enough. After that, we cover what Federal Law 152-FZ, Russia's personal data protection law, requires of bots that see employee and customer data, how to calculate savings, why bots fail, and what to prepare before you start. Choosing a messenger, website bots and first-line support have their own articles, and this one only links to them.

What is a chatbot for business and why do you need one?

In short: it is the cheapest interface to a business process. Almost everyone already has a messenger on their phone, so nobody has to install a new program or learn how to use it.

Bot builders describe a bot mainly as a marketing channel: funnels, broadcasts, auto-replies. A company with a warehouse, drivers or a network of retail locations needs a bot as a way into its system of record. A warehouse clerk in the middle of receiving a delivery won't open 1C on their phone. A driver on a trip won't fill in a web form, and an owner won't log into the reporting system at seven in the morning. All three open the messenger dozens of times a day, and the bot fits into that habit. It takes a short action, checks it against reference data and writes it to where the company's data lives.

The audience for this interface is already there. According to Mediascope, in June 2026 MAX, a Russian messenger, ranked first among messengers in Russia by monthly reach for the first time. It reached 86.26 million people, or 70% of the country's residents over 12 (3DNews, Andrey Sozinov, July 21, 2026). Telegram had 75.69 million that month, although in January it was first with 95.98 million (Habr, March 5, 2026). The gap is wider in daily reach: in August, 69.9 million people opened MAX on an average day, compared with 33.8 million for Telegram (Expert, Anna Zakharova, September 18, 2026). For an employee, it is an app that stays open all day anyway.

A work bot also depends on a reliable channel. In February 2026, Roskomnadzor, Russia's communications regulator, restricted media uploads in Telegram (Kommersant, July 6, 2026). According to Expert, it officially restricted access to the messenger in April. A warehouse bot routinely receives photos of delivery notes and damaged packaging, so the channel has to suit what employees will send through it. According to a VK report of August 13, 2026, the MAX partner platform, where chatbots and mini apps are built, is open to all Russian entrepreneurs and self-employed individuals (VK press release).

People already have plenty of programs without adding another window. In a survey of 4,000 Russians by VK WorkSpace and Hi-Tech Mail, more than a quarter said they lose between 30 minutes and two hours a day switching between programs. 42% said they sometimes spend longer looking for a file or conversation than working with it (ComNews, May 15, 2026).

Two more surveys show how much time goes into manual work a bot can take over. A study by NAFI and Yandex 360 (April 2025, 1,400 working respondents) found that office employees spend up to four hours a day on routine tasks. The most common ones included collecting and storing documents (55%) and sending messages and materials in messengers (51%) (Sostav, June 18, 2025). The hh.ru survey was conducted on May 5–12, 2025 among 3,283 respondents, and the results were published in February 2026. For 25% of workers, preparing reports compiled by hand takes more than six hours a week. Another 38% spend more than six hours a week getting documents, requests and queries approved (CNews, February 26, 2026).

What sets a bot apart from a group work chat is that it knows how the work is organized:

Which tasks should a bot handle inside the company?

In short: tasks where a person does a short, repetitive action away from a computer and the result is needed in the system of record.

Every action in the table already happens without a bot: in a phone call, a message to a group chat, a notebook in the warehouse. The bot moves it to a place where it can be seen and counted. Two questions help pick the first scenario: how often the action repeats and how much a mistake in it costs. Start with a frequent action where mistakes are expensive, such as receiving goods.

Surveys confirm the demand for this kind of automation. In a study by SberAnalytics and Sber Business Soft (November 2025, 559 respondents), 70% of participants named document workflows and request processing as the areas companies automate most. Among automation tools, 42% of companies use chatbots and voice assistants, the same share as use a CRM (CNews, January 22, 2026).

Most internal scenarios run on buttons and forms. The quantity received, the amount of a request and the item number have to be exact, and checking them against reference data is more reliable here than any model. A language model comes in where a person writes free text or sends a document: a newcomer's question, a photo of an invoice, a driver's voice message. The model prepares a draft and a person confirms it. Only then does the data go into the records. The article “AI agents for business” explains how a scripted bot differs from an agent that chooses its own steps.

How does a bot handle employee requests?

In short: the bot collects the request step by step, assigns it by rule and keeps the author informed about what is happening with it.

In many companies, a request for a purchase, an equipment repair, system access or a guest pass starts as a message in a group chat. An hour later the message has scrolled out of view, the assignee has forgotten it, and the author is asking “so, any news?” in private messages. A bot asks for the details in order: the request type from a button, the site or department from a list, a description, a photo, the urgency. It assigns the request by rule, for example by the combination of request type and site.

From there the bot works in both directions. The assignee gets a task with “in progress” and “done” buttons, and the author hears about status changes without making calls. If the deadline passes, the request is escalated to a manager. After a month, the register shows how many requests came in for each site, who closed them and how many days they waited.

Agree on reference data before launch. Sites, departments, cost items and the list of assignees already live in 1C or the ERP. If the bot pulls them from the same database, requests are sorted by cost item and checked against the budget. If employees type the site by hand, the report will list “Warehouse 2”, “whse #2” and “the second warehouse” as three different places.

Can you run the warehouse and stock through a bot?

In short: yes for receiving, write-offs, transfers and stock queries, as long as the bot works with the same database where the goods are counted. It does not replace bin-location storage or bin-by-bin order picking.

When a delivery arrives, the warehouse clerk selects it, enters the actual quantity for each line and photographs the delivery note and any damaged packaging. Any gap between the delivery and the order goes to the buyer right away, while the supplier's truck is still at the gate. At a store, the sales assistant records a sale, a return or a write-off, and the system recalculates the stock itself. The bot answers “how many are left?” from the records, so nobody has to call the warehouse clerk.

This is how our project for a retail chain works. A Telegram bot records sales and returns at the stores, and a shared system calculates receipts, stock, financial transactions and payroll automatically. Together with the inventory accounting, this let the client cut an inventory controller position and keep human error in the records to a minimum.

A warehouse bot has its limits. When the product range runs to thousands of items, picking goods with buttons becomes awkward, so the bot gets SKU search or a separate barcode-scanning screen. If the warehouse uses bin locations and handheld data terminals, it needs a warehouse management system (WMS), and the bot stays alongside it as a secondary channel. The article “WMS: what it is and when 1C is enough” lists the signs that a warehouse is ready for one.

How can an owner get a morning report in a messenger?

In short: the summary arrives at the same time every day and shows five to seven numbers that tell you where to look today.

Technically, this is the simplest scenario: the bot does not write anything, it only reads. A typical set of numbers covers yesterday's revenue against plan, cash in the accounts, overdue receivables, orders and shipments, and requests past their deadline. On top of the schedule come event alerts: a large payment, stock below the minimum, a large order cancelled. These messages arrive immediately rather than waiting for the morning.

The hard part here is the data. If a number in the summary differs from the one in 1C, the owner quickly stops trusting the bot. So the summary is built on the same sources and calculation rules as the management reports. It also needs an agreement in advance on what counts as revenue: shipments, payments received or closed orders. The article “The owner's dashboard” covers which metrics belong on the first screen, and the piece on management reporting explains how to build the reports themselves.

How do you approve payments and documents through a bot?

In short: a manager approves a payment request or an internal document with a button on their phone, and the system records the approval along with who gave it and when.

A payment request arrives as a card showing the counterparty, amount, cost item, due date and the invoice as an attachment. The route depends on the amount. Below the threshold, the department head decides. Above it, the CFO or the owner joins in. Each of them has two buttons: “approve”, and “send back”, which requires a comment. After the last approval, the request goes to the payment register, where accounting sees it without anyone forwarding it. If an approver is on vacation, their deputy gets the request.

A button in a messenger has a legal limit. Under Part 2 of Article 6 of Federal Law 63-FZ, Russia's electronic signature law, a document signed with a simple electronic signature is equivalent to a paper one with a handwritten signature only in certain cases. Those cases are set by law, by regulations or by an agreement between the parties to the electronic exchange (63-FZ, Article 6, ConsultantPlus). Inside a company, this procedure is set out in an agreement with employees or an internal policy. Documents for counterparties are signed with an electronic signature, usually through an operator of EDI, the legally binding electronic document exchange between companies in Russia. The article “Contract document workflow” shows how to build a complete contract approval route. The “Document flow” page describes how we build this kind of setup turnkey.

How does a bot answer newcomers' questions?

In short: the bot answers from the company's procedures and knowledge base, which takes the recurring “where do I find” and “who do I ask” questions off the mentor's plate.

In their first weeks, new employees ask the same questions: how to request leave, where the contract template is, who signs off a business trip, how to get access to a system. A mentor or the most patient colleague answers them, and nobody accounts for that workload. A bot with a language model finds the answer in the company's documents and links to the relevant procedure. The article on a website chatbot explains how this kind of document search works, and an internal bot is built the same way.

X5, one of Russia's largest grocery retailers, has an example at scale. The AI assistant in Sfera, its mobile app for 426,000 employees, lets staff resolve up to 90% of requests without an HR specialist and has already handled more than 70,000 inquiries (CNews, September 11, 2025). A company of a hundred people gets fewer questions, but they cover the same topics.

Unlike a customer bot, an internal bot has to handle permissions. An employee knowledge base holds bonus policies, HR orders and internal prices, and a new warehouse worker has no need for the finance department's documents. So the bot knows who is asking and searches only the documents open to that role. A question the knowledge base cannot answer goes to a person, and it also shows which procedure needs to be written next.

How does a bot work with drivers and staff at retail locations?

In short: an employee outside the office sends status updates, photos and checklist items from their phone, and the data goes straight into the system.

Field staff have no computer at all, so the messenger becomes their only window into work. In our dispatch system for a logistics company, the driver gets the trip, route stops and documents in the bot, sends photos as evidence, and trip statuses update straight from the road. The dispatcher sees where the truck is without calling the driver, and the trip documents are ready the same day the trip closes. Trip rules, statuses and documents live in the dispatch system itself, and the messenger only carries the messages. So Telegram's restrictions, including the limits on photo uploads, are solved by moving the bot to MAX. Only the layer that connects to the messenger has to be rewritten.

At sales and service locations, the bot opens a shift and walks the employee through a checklist (cleanliness, displays, cash register). It asks for a photo where a checkmark alone can't be trusted. In the same retail project, shifts at the stores open automatically. The system collects the store checklist and the employee's form, and analytics shows the best and worst shifts. The chain manager no longer has to drive around the stores to see where something went against procedure.

This kind of analytics has a legal limit. Paragraph 6 of Article 86 of the Russian Labor Code says an employer may not make decisions affecting an employee's interests based on data obtained only through automated processing or only in electronic form (Russian Labor Code, Article 86, ConsultantPlus). A bonus or a disciplinary penalty based on bot data is decided by a person who has looked into the circumstances.

Which tasks does a bot handle for customers?

In short: booking and rescheduling visits, order status, taking orders and handling complaints. Each of these scenarios relies on the same data as the internal ones.

1. Booking and rescheduling a visit. The bot shows open slots from the schedule, makes the booking, sends reminders and handles rescheduling. In our project for a network of service branches, an AI administrator in messengers books and reschedules visits around the clock. Automatic reminders and confirmations reduce no-shows without anyone phoning clients.
2. Order or delivery status. The answer to “where is my order?” is in the records, and the bot takes it from there without pulling in a manager.
3. Taking an order. A wholesale customer sends an order through the messenger, and the bot checks the items against the product catalog and replies with what is in stock. In the project for a wholesale and manufacturing company, orders from email, messengers and the customer portal go into a single shared queue and immediately reserve real stock. Customers see their contract prices in the portal.
4. Complaint or return. The customer sends the order number and a photo, and the bot creates a complaint card with an owner and a response deadline.

The article “Chatbot for business in Telegram and MAX” covers which messenger to choose for customers, how to connect the bot and how to accept payments through it.

Website bot. It works with an anonymous visitor while they are choosing and knows which page they came from. The piece “AI chatbot for your website” explains how it works, which model to choose and what the law requires when collecting contacts.

First-line support. When a company gets a large volume of similar customer inquiries, the bot is treated as an investment with a payback period. The article “First-line AI agent” walks through the calculation and the signs that a bot will end up as a toy.

Where does a bot builder reach its limits?

In short: a builder is enough as long as the bot runs a button-based script and doesn't touch data in other systems. As soon as the bot writes to 1C, a CRM or inventory records, it takes on the company's reference data, access rights and exceptions.

A builder does well what it was made for: button menus, broadcasts, funnels, accepting payments, collecting requests into a spreadsheet or CRM. Its list of integrations shows this. The catalog of BotHelp, a Russian chatbot builder, for example, lists GetCourse, Bizon365, Tilda, amoCRM, Bitrix24 (a Russian CRM and collaboration suite) and Google Sheets together (BotHelp help center). As of September 26, 2026, the catalog has no ready-made 1C integration. The connection is built through intermediary services from the same list, such as Albato or ApiWay, or through an external API request.

The large platforms usually handle data storage properly. BotHelp states in its policy that it hosts and processes Russian citizens' data only in Russia and acts as a processor on the user's behalf (BotHelp policy). Puzzlebot also stores data in Russia and publishes its data processing terms (Puzzlebot policy). Where a builder falls short is business logic.

Exceptions take up the most time. A supplier delivers only part of the order, a delivery arrives with similar items mixed up, a sales assistant processes a return without a receipt, an approver goes on vacation on the day of a large payment. Every company has dozens of rules like these, written into procedures, 1C settings and employees' heads. In a builder, each rule becomes another branch of the script on top of someone else's logic, and such a script quickly becomes hard to maintain.

What does 152-FZ require when a bot sees employee and customer data?

In short: the company is already a personal data operator for both employees and customers. The bot changes details in its notification to Roskomnadzor. Databases with Russian citizens' data must be in Russia. A builder or hosting provider needs a data processing agreement, and a data leak can cost millions of rubles.

An internal bot almost always processes personal data: full names and job titles in requests, drivers' phone numbers, shift schedules, passport details in guest pass requests. A company with employees is already a personal data operator. As a rule, it has already filed a notification with Roskomnadzor under Article 22 of 152-FZ. A bot often changes information in that notification: the purposes of processing, the types of data, where the database is located. Under Part 7 of Article 22, changes must be reported no later than the 15th day of the month after the month in which they occurred (152-FZ, Article 22, ConsultantPlus). If there is no notification at all, it has to be filed before processing begins. A legal entity that fails to notify faces a fine of 100,000 to 300,000 rubles under Part 10 of Article 13.11 of the Code of Administrative Offenses (Russian Code of Administrative Offenses, Article 13.11, ConsultantPlus).

Employee data. The legal basis is usually the employment contract and the employer's legal obligations (clauses 2 and 5 of Part 1 of Article 6 of 152-FZ). The Labor Code adds its own rules. Article 86 limits the purposes of processing to compliance with the law, employees' personal safety, monitoring the quantity and quality of work, protecting property and a few other purposes it names. Article 88 requires that only specially authorized people get access to employee data, and only to the data they need for their specific duties (Russian Labor Code, Article 88, ConsultantPlus). For a bot, this translates directly into permissions: the warehouse clerk doesn't see drivers' phone numbers, and a store manager sees only their own staff.

Customer data. If consent is the basis for processing, then since September 1, 2025 it has to be obtained separately from any other documents the person confirms or signs. Federal Law 156-FZ added this rule to Article 9 of 152-FZ (156-FZ, Article 5, ConsultantPlus). Collecting contacts through a bot is covered in detail in the articles on the website bot and choosing a messenger.

Where to store it. Part 5 of Article 18 of 152-FZ, as amended by Federal Law 23-FZ with effect from July 1, 2025, prohibits recording, accumulating and storing Russian citizens' personal data in databases outside Russia (152-FZ, Article 18, ConsultantPlus). So the bot's database of requests, employees and customers has to sit on a server in Russia. If the bot runs in a builder or on third-party hosting, the company signs a data processing agreement with the provider under Part 3 of Article 6. The law explicitly requires that agreement to include compliance with Part 5 of Article 18 (152-FZ, Article 6, ConsultantPlus). The company still answers to the individual for what its contractor does.

Foreign messengers. Anna Kononenko, a Roskomnadzor representative, explained that personal data may be processed through foreign messengers in the cases listed in Part 1 of Article 6. The condition is that the operator complies with Part 5 of Article 18 and with Article 12 (Klerk, Galina Nerubashchenko, September 12, 2025). Article 12 governs cross-border data transfer. The clarification doesn't say whether bot messaging in Telegram itself counts as such a transfer, so the cautious reading is the safer one. If full names, phone numbers and photos of documents end up on a foreign company's servers, that is a cross-border transfer. Under Part 3 of Article 12, the operator must notify Roskomnadzor of its intention before the transfer begins. This rule has applied in its current form since March 1, 2023 (152-FZ, Article 12, ConsultantPlus). The simplest approach is to keep such data out of messenger messages entirely. The messages carry only the request number and status, and everything else stays in your own database in Russia.

For some organizations, the law has already settled the question. Since March 1, 2023, banks, non-bank financial institutions, companies more than 50% state-owned and a number of other organizations have been prohibited from sending payment documents and citizens' personal data through foreign messengers (149-FZ, Article 10, ConsultantPlus). Since June 1, 2025, Federal Law 41-FZ has barred government bodies, banks, telecom operators, marketplaces and several other categories of organizations from using foreign messengers to send information to citizens (41-FZ, Rossiyskaya Gazeta).

The cost of a leak. Since May 30, 2025, a personal data leak has cost a legal entity 3 to 15 million rubles, depending on how many people are affected, and a leak of biometric data up to 20 million. A repeat leak costs 1–3% of annual revenue, but no less than 20 million and no more than 500 million rubles (ConsultantPlus on Federal Law 420-FZ). A bot that handles the data of a hundred employees and thousands of customers needs the same protection as a CRM: permissions, action logs and backups. The article “AI without data leaks” explains how to build a setup that keeps data inside the company.

How much does a bot save and how do you calculate it?

In short: savings are calculated per operation. Count how many times a month the action is done by hand, how many minutes each one takes and what an hour of the person doing it costs, then add the cost of errors.

Here is an example of the math for a warehouse:

1. Count the operations. Say there are 12 deliveries a day over 22 working days, which gives 264 deliveries a month.
2. Estimate the time. The warehouse clerk spends 15 minutes checking the delivery note and entering the data into 1C, and the buyer spends 10 minutes sorting out discrepancies by phone. That makes 25 minutes per delivery, or 110 hours a month.
3. Multiply by the rate. At 600 rubles an hour, that comes to about 66,000 rubles a month.
4. Add the errors. A mix-up of similar items found a month later during stocktaking can cost more than all those minutes put together.

The bot doesn't remove those minutes entirely, because the goods are still counted by hand. It takes over the data entry and the phone calls and flags a discrepancy while the supplier is still on site. That's why it pays to measure two or three operations before launch. Without those numbers, you will have nothing to show the effect against later.

The same SberAnalytics and Sber Business Soft study gives a market benchmark. After introducing automation tools, 37% of companies reported a lower share of manual operations and 36% reported fewer errors in documents and reports (CNews, January 22, 2026). The article “Automating routine work: where to start” goes into more detail on how to calculate the cost of routine work and pick the first area to automate.

Why do internal bots fail to catch on?

In short: most often a bot is abandoned because the old route through the chat stayed open, because it doesn't write anything to the records, or because nobody owns it.

1. The old route stayed open. If the group chat still accepts requests, many people will keep posting there. Set a date after which a message in the chat no longer counts as a request.
2. The bot only forwards messages. A bot that drops requests into yet another chat just adds an extra step for people.
3. Too many questions. A twelve-step form in a cold warehouse loses to a phone call. The form should have exactly as many required fields as the records need.
4. No feedback. An employee submits a request, can't see what is happening with it, and a week later calls again.
5. No owner. Reference data goes stale, nobody adds new sites, nobody looks into errors in the exchange with 1C, and within a quarter the bot is showing wrong data.

What does it take to make a bot work for you?

In short: a documented process, access to data in your business systems, access control and an owner after launch. Everything else is engineering work.

1. A documented process. For each scenario, define who starts the action, which fields are required, who carries it out, what the deadline is, what counts as an exception and who resolves it.
2. Access to data. Products, warehouses, sites, counterparties and employees from 1C, the ERP or the CRM. The bot takes its reference data from those systems and writes the results back to them.
3. Integrations. Which systems the bot reads from, which it writes to, and what happens when one of them is unavailable. 1C has built-in tools for this: an automatic REST interface over the OData protocol, and custom HTTP services (1C:Enterprise platform website). The piece on systems integration covers the ways to connect systems.
4. Access control. Who sees which warehouses, departments, amounts and documents. Who approves, and up to what limit. How the bot recognizes an employee, and what happens to their access when they leave.
5. Channel and storage. A messenger for employees and customers, a server in Russia, and a data processing agreement with the builder or hosting provider if data passes through them.
6. An owner after launch. A person inside the company who adds sites and scenarios, resolves disputed cases and closes the old route through the chat.
7. Support. Messenger and business system updates, exchange errors, and new scenarios as employees get used to the bot.

This list shows why a bot's cost can't be quoted before the process has been reviewed: the number of scenarios, reference lists, integrations and exceptions determines it.

Builder, off-the-shelf product or a bot built around your process?

In short: an off-the-shelf solution covers a standard scenario, but a bot that writes to the records has to be built around your own reference data, permissions and exceptions.

Ready-made bots and builders work well where the process is the same for everyone: broadcasts, booking, collecting contact details. Work scenarios are harder. Every company has its own procedures, reference data, access rights and exceptions. In one, the warehouse clerk confirms deliveries. In another, the shift supervisor does it with a photo, and in a third, the buyer confirms after checking against the order. A bot built on someone else's model either forces people to work that way or piles up manual workarounds. We have written separately about choosing between an off-the-shelf solution and a system built around your process: custom system or boxed software.

At INCUBE AI, we build bots as part of a working system. The logic and database live on your side, and the messenger is connected only as a transport. Reference data and permissions come from 1C, the CRM and inventory records, and results go back into them. Our solutions catalog includes an inventory bot and scheduled auto-reports, and we tailor those to each client's process too. We work under a contract, data stays in Russia, and we support the system after handover. If you know which employee actions you want to move into a messenger, tell us about your task. We will go through the scenarios, identify the integrations and estimate the scope of work from them.

Sources and links

• Mediascope via 3DNews, Andrey Sozinov, July 21, 2026 — monthly reach for June 2026: MAX first among messengers for the first time, 86.26 million people (70% of Russia's residents over 12), Telegram 75.69 million
• Mediascope via Habr, March 5, 2026 — Telegram's monthly reach in January 2026: 95.98 million users over 12
• Mediascope via Expert, Anna Zakharova, September 18, 2026 — daily reach for August 2026: MAX 69.9 million, Telegram 33.8 million; restriction of access to Telegram in April 2026
• Kommersant, July 6, 2026 — restriction of media uploads in Telegram since February 2026
• VK, press release of August 13, 2026 — the MAX partner platform is open to all Russian entrepreneurs and self-employed individuals
• ComNews, May 15, 2026: VK WorkSpace and Hi-Tech Mail survey — 4,000 Russians: time lost switching between programs and searching for files
• Sostav, June 18, 2025: NAFI and Yandex 360 study — April 2025, 1,400 respondents: up to four hours of routine work a day and its most common types
• CNews, February 26, 2026: hh.ru survey — survey of May 5–12, 2025, 3,283 respondents: time spent on reports compiled by hand and on approving documents, requests and queries
• CNews, January 22, 2026: SberAnalytics and Sber Business Soft study — November 2025, 559 respondents: what companies automate, the share of chatbots among tools, the effect of automation
• CNews, September 11, 2025 — X5's AI assistant in the Sfera employee app
• BotHelp integration catalog, BotHelp policy, Puzzlebot policy — builder integrations, data storage and data processing terms, viewed on September 26, 2026
• HTTP services and REST interface, 1C:Enterprise 8 platform website — built-in tools for exchanging data with external systems
• 63-FZ “On Electronic Signature”, Article 6, ConsultantPlus — when a simple electronic signature is equivalent to a handwritten one
• Russian Labor Code, Article 86 and Article 88, ConsultantPlus — processing and transfer of employees' personal data
• 152-FZ, Article 6 and Article 18, ConsultantPlus — legal grounds for processing, processing on another's behalf, and data localization as amended by 23-FZ
• 152-FZ, Article 12, ConsultantPlus — notifying Roskomnadzor before a cross-border transfer begins (Part 3)
• 152-FZ, Article 22, ConsultantPlus — notification of processing and the deadline for reporting changes (Part 7)
• 156-FZ of June 24, 2025, Article 5, ConsultantPlus — consent to processing as a separate document from September 1, 2025
• Klerk, Galina Nerubashchenko, September 12, 2025 — Roskomnadzor's position on processing personal data through foreign messengers
• 149-FZ, Article 10, ConsultantPlus — the ban for banks, state-owned companies and other organizations, as amended by 584-FZ
• 41-FZ of April 1, 2025, Rossiyskaya Gazeta — ban on using foreign messengers to send information to citizens from June 1, 2025
• Russian Code of Administrative Offenses, Article 13.11 and ConsultantPlus review of fines from May 30, 2025 — fines for failure to notify and for data leaks under Federal Law 420-FZ