Egor Maraev
AI transformation expert
Co-founder of Robot & Sons
- Training for teams
- Custom development of AI solutions
I transform how teams work by implementing AI agents,
to increase profit
- Sales
- Marketing
- HR
- Finance
When people come to me
You want to implement AI in your company, but do not understand where to start
You know which task can be automated and are looking for the optimal solution
What I offer
At the first meeting, I find out how your team works with AI agents and propose an approach. It may include one of the following elements or a combination of them.
Schedule a first free consultation
where we will figure out how you can raise your profit with AI
or email meegor.maraev@gmail.comTestimonials
“Already during the program we managed to implement some projects, and there was a feeling of magic. And it is important to keep learning this magic.”
Read the full testimonial
The result of the program was an understanding of what AI is, how you can work with it, and what functions it has. A great deal depends on where this AI can be connected, so most of the possibilities become clear if you have access. Already during the program we managed to implement some projects, and there was a feeling of magic. And it is important to keep learning this magic. Thanks to Egor for opening my eyes to AI, for his openness and sincere involvement in the projects.
“Two months after the start of the course, I had immersed myself quite deeply in the principles of automation and built two content creation systems for my current projects.”
Read the full testimonial
Before taking Egor's course, my use of neural networks was limited to research and writing short texts with ChatGPT. I had attempts to figure out AI agents, but as a result of my first experiments I simply wiped my operating system by mindlessly clicking around in Cursor. The learning materials I used back then did not give a clear foundation for getting started and were designed more for independent exploration.
In this respect, Egor's course was a perfect fit for me as a beginner – Egor gives exhaustive instructions, explains and shows everything, and the student only has to follow him and get into the processes in practice. Already in the first week, the course students sort out the main connections, try building their first automations and receive recommendations on organizing their work process.
Two months after the start of the course, I had immersed myself quite deeply in the principles of automation and built two content creation systems for my current projects. Each of them involves copywriting in a specific style, preparing graphics, choosing topics, autoposting, assembling and updating the content plan, and more. It was Egor's course that gave me the skills needed for this, and the opportunity to ask questions and consult with Egor, the curators and other students makes it possible to solve any difficulties that arise in the work process faster and more effectively.
I wholeheartedly recommend Egor's course and consultations and continue to explore the possibilities of AI agents with great enthusiasm.
“My everyday productivity has grown noticeably, and now I habitually look for more effective approaches to many of my tasks.”
Read the full testimonial
I recently completed Egor's in-depth AI course. Before the course, I was not entirely sure that AI skills could have such a strong impact on everyday work. The program changed my view.
Egor went far beyond writing prompts. In the course, I learned how to create reusable skills for specific workflows and connect AI agents to external services, applications and workspaces. I came to understand much more clearly how AI can help with both routine tasks and more complex work, including market research, data analysis, competitor monitoring and gathering information from different sources.
My everyday productivity has grown noticeably, and now I habitually look for more effective approaches to many of my tasks. It was a thorough, very practical course and an excellent investment. I confidently recommend Egor to professionals who want to apply AI in real work.
“After just two hours of working with him, I started actively using Claude Code for research, vibe coding and managing advertising accounts in my business.”
Read the full testimonial
I want to recommend Egor for his ability to explain how AI agents work clearly and practically. After just two hours of working with him, I started actively using Claude Code for research, vibe coding and managing advertising accounts in my business.
After Egor's 30-minute consultation with one of my employees, this employee was able to independently create an AI agent that prepares high-quality commercial proposals for engineering services.
I strongly recommend Egor to everyone who wants to master practical skills of working with AI agents and quickly achieve noticeable progress.
Projects
Team training, working AI solutions and personal experiments.
“Ono Samo” course
A basic course on creating AI agents based on ChatGPT Work. More than 100 people have completed the course; you can join at any time.
AI for top managers
A training program for E-promo Group top managers based on Claude Cowork. The executives independently created 12 pilot automation projects and learned to assess their economic efficiency.
AI agents inside a corporate portal
An MCP server for Gmail and Google Calendar and a Telegram integration inside the E-promo Group portal. Employees can work with their mail, chats and calendar and create agents for project tasks.
Automatic cost estimates and tender preparation
The agent calculates video production cost estimates: preparing one estimate went down from 1.5 hours to 20 minutes. After training, the production director created an agent herself for selecting tenders and filling in documentation.
Designing climate systems from a floor plan
An agent for SVS Climate creates a preliminary project calculation from a floor plan. The client receives a preliminary answer in 2 hours instead of 4 days.
Agent — financial manager
Financial analytics and management accounting through an agent in Telegram. Only I and my partners can ask questions; only I can edit the spreadsheet through the agent.
Project status control system
An agent in Cult Group's project chats collects statuses and sends summaries to producers every morning. The production director sees the situation across all projects on a dashboard.
English learning assistant
Personal project
A local application corrects the sentences I type or dictate in English. The enhance mode helps me see richer language and save new words to a local database.
Corporate information anonymizer for working with LLMs
Personal project
A project in development. I am creating a system that replaces sensitive information before working with an LLM and then allows the result to be decoded back.
At the intersection of three competencies
Entrepreneur
For 14 years I was the owner and CEO of creative video production studios — Blaster Studio and Punch Creative. Now I am a partner and AI director at Robot & Sons.
Trainer
I have been teaching for 7 years; in 2026, representatives of 27 companies completed my “Ono Samo” course on AI agents.
Techie
I am passionate about engineering problems, dive deep into technical aspects and design systems.
My principles
A person should not do
technical work
Creating something new, making judgments, reaching agreements with people. Seeing where to move and leading others. This is exactly where human experience, attention and responsibility are needed.
Transferring data, compiling reports, searching for files and repeating the same actions can be delegated to machines. I help separate this work from the decisions that remain with people and hand it over to AI agents.
Free up attention for what requires a human.
AI under control
It is important to understand what the agent does, where it gets information and how it arrives at the result. There should be no blind spots in its work that have to be taken on faith.
When I teach, we go through Claude's specific actions: which files it reads, which tools it calls and how to verify its conclusions.
When I develop a corporate agent, I restrict access to data and tools, test scenarios and leave confirmation of significant actions to a human.
Control means understandable actions, a verifiable result and limited permissions.
More than 100 people have completed this course.
A basic course on creating AI agents based on ChatGPT Work. The ideal starting point if you want to figure out this topic at your own pace and right away at the deepest possible level.
Course format: a recorded video course + weekly meetings with Egor on Tuesdays. You can sign up at any time and start studying right away.
View the course content, read testimonials and pay for participation
Top managers of the E-promo Group management company completed training based on Claude Cowork.
As a result of the training, the top managers independently created 12 pilot automation projects within their verticals and also learned to calculate the economic efficiency of automation.
Here are the automations they built — without help from the IT department:
HR departmentAutomatic sourcing and scoring of candidates from the HH.ru database and comparison of the best candidates with the Talantix database.
AccountingBased on bank statements that arrive by email daily, an automatic cash flow report is created. Revenue is broken down by department, and emails are sent there.
MarketingA system for automatic generation of banners that takes brand design into account was created.
Tender departmentAutomatic analysis of suitable tenders and their scoring. The time to process information on one tender went down from 18 to 3 minutes.
The estimated value of the time freed up by the implemented workflows is about 250,000 rubles per month.
Most mid-sized companies have internal corporate AI portals: a collection of popular LLM models accessed through a shared chat.
However, almost everyone uses their portals as a chat interface rather than as full-fledged AI agents that could complete employees' specific tasks.
E-promo Group also has such a portal. I developed an MCP server for Gmail and Google Calendar and also built a Telegram integration.
Now every employee can use the corporate AI to securely read their Telegram and mail and create meetings in the calendar.
Based on the corporate portal, employees can independently create agents that will monitor their project activities, write project reports, analyze conversations and schedule meetings.
While working closely inside my studio as CEO, I created an agent that calculates video production cost estimates.
A link to a demonstration of the agent will be added later.
As a result, preparing one estimate began to take producers not 1.5 hours to read the brief and create the estimate, but 20 minutes — to read the brief and check the estimate.
The significant time savings allowed us to reduce the number of staff producers and get through the crisis period of early 2026.
Then I trained the production director to work with agents, and she herself created an agent that checks tenders on the tender platform, selects the ones relevant to us and fills in the documentation with our data.
This also allowed us to work much more efficiently with the same small team.
I created an agent for SVS Climate, a company that designs and installs climate systems.
From the customer's floor plan, it creates a preliminary project calculation.
From a business point of view, such an agent allowed the customer to give a preliminary answer to their client much faster — in 2 hours instead of 4 days.
This greatly reduced the share of lost hot clients who value speed in a contractor.
In my company, financial analytics and all management accounting are handled by an agent that you can contact directly via Telegram and ask to answer questions about the current accounts.
Only I and my partners can ask a question via Telegram. A strict check for trusted recipients of information is in place.
At the same time, I can also edit the spreadsheet through the agent, while the partners cannot — a separation of permissions.
The data itself is stored in a Google Spreadsheet, and if desired, you can go in and look for yourself.
This is an agent that sits inside the Cult team's project chats and collects information about the status of all projects.
Every morning, the agent sends the project producer a summary: what happened in the project during the day, which materials are awaiting approval and where there are unresolved issues.
The producer can forward this summary to the client without formatting it themselves.
In addition, the production director sees the situation across all projects on a dashboard: what is happening, whether there is client negativity anywhere or everything is fine — for each project.
This agent was originally built with the n8n system and later moved into code.
I designed a system that helps me develop my English.
To develop a language, you need to constantly practice composing complex sentences and gradually expand the boundaries of your vocabulary and lexical constructions.
I built a simple local application that corrects my English.
In the first step, I dictate or type a sentence the way I see it, and immediately see a corrected version of that very sentence.
In addition, there is an enhance mode that lets me see new words and richer language.
I can save new words to a local database of words that interest me.
The project is not finished yet; I am working on it right now.
A popular question among corporate clients is: how do you work with confidential information if it cannot be given to an LLM?
To solve this, I am creating a system that replaces sensitive information with neutral values.
For example, it changes the specific names of employees inside a corporation to other, meaningless names.
After the modified document has been processed with the LLM, the result can be decoded back, producing the document in the required form. At the same time, the LLM never read the original document.
As soon as I finish the project, I will post the details here.