Software and AI systems for business

Majento builds business applications, internal systems, AI agents, Telegram bots and Mini Apps. We automate workflows across sales, finance, operations, HR and marketing, connecting company data and tools. We also train teams to work with AI and provide systems analysis. You can choose a single stage or work with us from problem analysis through development and launch. Services and team.

Our office is in Tashkent, and MAJENTO LLC is a resident of IT Park Uzbekistan. We work with companies in Uzbekistan and Central Asia, Russia, Belarus and with international teams. Projects run in Russian and English, remotely or with in-person meetings by agreement. We adapt the working format to the team, its systems and data requirements.

Majento was founded by Serge Shima and Batyr Latipov. Serge handles product, AI and R&D; Batyr handles strategy and business development. The team includes developers, designers and managers. They take a task from process analysis through launch and ongoing operation. Founders and team.

Co-founder Serge Shima has worked with ML and AI since 2017 and runs corporate training programs. Among Majento's projects: an agent-based system for content creation and localization for Tangem, and support and development of the Pepsi Music Chart mini-app together with Wunder Digital. The company page covers Majento's projects, Serge Shima's training work and his previous career in marketing and technology.

We build business applications and systems around company needs: internal services, workspaces, record-keeping tools, reports, knowledge bases, integrations and interfaces for employees and clients. These can support sales, finance, HR, operations and marketing teams. For Telegram we develop bots, catalogs, ordering services and Mini Apps. For brands we build games, polls and interactive promo apps. First we define what a person needs to do in the product, then we design the interface and logic. Custom software development.

Custom software makes sense when a ready-made service doesn't account for your data, workflow logic or employee roles. Sometimes it's enough to configure an existing CRM, connect two systems or update a procedure. Majento analyzes the process before development and proposes the right scope: from a small integration to a standalone application. How we choose a solution.

No. Menus, status checks, duplicate detection and data transfer between systems are often handled by plain rules. AI is useful for text, voice, documents and conversation history. We test the task on examples before choosing a technology. When rules are enough and when AI helps.

A chatbot is a way to interact with a system through messages. Behind it can be buttons, rules or a language model. An AI agent, beyond answering, can access permitted tools, look up data and execute steps of a task. In a Majento project we define its sources, permissions and actions that require employee confirmation in advance. Agent vs chatbot breakdown.

We help process inquiries and requests, route documents for approval, prepare reports and summaries, find answers in knowledge bases and move data between business systems. These tasks span sales, finance, HR, operations, management and marketing. For example, we can bring requests from several channels into one workspace or connect documents to approval stages. For marketing we connect catalog data, stock levels, campaigns and orders. Action rules and result checks are agreed with the client. Business automation and marketing.

We start with sheets, formulas, reference tables and the source of each metric. Together with the report owner we check for gaps, duplicates and units of measurement. The first version is compared against examples calculated by hand. AI can help parse a workbook and prepare explanations, but numerical correctness is verified separately from dashboard appearance. Working with Excel.

Majento connects marketing work to business data: catalog, prices, stock, orders and campaign spend. Ad copy is just one part of this work. Current product data, action approval and result verification against orders matter. For example, a suggestion to promote a product needs to be checked against its availability. Marketing automation.

Yes, these tasks are among the scenarios Majento works on: summaries from documents and correspondence, recording decisions, preparing action items and analyzing customer calls. Before development we define where information comes from and who verifies the output. A summary should show sources and discrepancies so the manager can make an informed decision. Example daily digest and call analysis.

A regular bot suits menus, forms, booking and status checks. An AI assistant is needed when people ask free-form questions and answers must be found in documents. A catalog, cart, game or multi-step order is better built as a Mini App. Majento picks the format based on real user interactions and the data the system will handle. Telegram format comparison.

We design the handoff together with the bot scenario: which interactions require a human, what history they'll see and who sends the final answer. If a source is unavailable or the bot cannot answer within the agreed workflow, the bot should report this and pass the context. Order changes and other consequential actions follow agreed permissions. How a working bot is built.

Design at Majento helps a person understand the system and complete a task. Developers and designers work together on user flow, clear copy and interface states. Graphics and animation help read data and understand what happens after an action. For branded projects we tie illustrations and visual style to the brand character. We check the interface on desktop and phone, readability and control, animation and behavior in different languages.

Training starts with employees' tasks. We find out what the team works with and what they already know, then select exercises. In practice, participants set tasks for AI, work with documents and data, check answers and note working techniques. The methodology is based on AI Masters programs, an educational project by Serge Shima, co-founder of Majento. Training details.

Yes. Programs are tailored for managers and staff in marketing, sales and operations. Employees learn on work they already know: comparing documents, preparing customer replies, analyzing reports and creating presentations. An AI Masters program can cover 10 to 500 employees. Large teams attend in groups of up to 15. Corporate training formats.

Managers work with summaries, reasoning behind conclusions and choosing tasks for implementation. Marketing and sales work with campaign materials, content and customer questions. Operations teams work with documents, reports and repetitive procedures. The common part covers task formulation for AI, result verification and data handling rules. Program tailored to participants.

We use the client's documents, spreadsheets, reports and tasks that can be safely used in a session. Permissible materials are agreed with the client. For confidential information we prepare a training example; anonymized data is used only with client consent. The AI Masters program covers text, documents, presentations, data and research, automation and security. AI Masters knowledge cards are available for review. Session content and materials are agreed per team.

In a session, a participant checks quotes against the document, numbers against source data, and checks whether the answer meets all task requirements. For example, when comparing two delivery descriptions, AI should provide quotes from both documents and flag a missing field. We check unsuccessful answers too. If the program includes skill assessment, the employee completes a real task against agreed criteria. Example first workshop.

Depending on the program, the team keeps prompt templates, instructions and data handling rules that they verified on their own examples. We separately note tasks where an agent or integration would be worth building. Sometimes training ends with a new work procedure; sometimes it becomes the start of a development project. How a session leads to a project.

Yes. Training, systems analysis and development can be ordered separately. After practice, a team sometimes needs only a prompt template or a new procedure. If a recurring task emerges with available data and verifiable results, we assess it as a separate project. Team AI training starts at $2,600, one-to-one mentoring at $1,500 and a team AI hackathon at $4,800. These are minimum prices; the final amount depends on the program we agree on. Training and next steps.

We start with one process and the person responsible for its outcome. We analyze data, working systems and constraints. Then we design a solution, build a first version and test it with users. After verification we connect agreed systems, train the team and arrange ongoing support. We agree on working scenarios and test how the system handles incomplete data and failures. AI implementation process.

Yes. We pick one process, document the current way of working and agree what the pilot must prove. We test the first version on normal and problematic examples with the process owner. The prototype gets access to working systems after the client confirms the test results. A pilot helps evaluate the solution before scaling the project. From verification to implementation.

First we check available interfaces and data requirements. We design connections to CRM, ERP, knowledge bases, documents and other sources through available APIs and integrations. We choose a cloud, on-premise or hybrid setup after assessing the specific system and infrastructure. Permissions, data storage and human-confirmed actions are agreed with the client's IT department and security team. For risk assessment we use NIST materials and OWASP.

Majento designs answers with file and version references for where information comes from. Contradictory documents, outdated versions and missing fragments are checked separately. The system should show a discrepancy or pass the question to an employee if there's no reliable basis for an answer. The user's right to read a document is verified before search and before delivery. Answers from company documents.

These situations are included in agreed checks before launch. We define where an integration error lands, who handles an incomplete request and how an employee continues manually. An answer without a reliable source requires human verification. Exceptions in automation.

Before launch we agree on acceptance criteria, access rights, instructions and support escalation. The handover materials and the rights to them are set out in the contract. Work volume depends on integrations, data quality, user roles and error behavior. These conditions are defined after task analysis. The same app name can mean different projects, so we don't promise a universal price or deadline upfront. Custom development stages.

Iva is an open-source AI assistant in Telegram built by our team. It works with text, voice messages and documents, stores knowledge and agreements, helps with tasks, search and reminders. Iva can be installed on your own server; source code is published on GitHub. It's a working product that shows our approach to agents, memory and integrations.

The team's product work includes Sezish for meeting dictation and transcription, Sotish for AI marketing of stores and Claimp, an open player for preparing DJ sets on macOS. Sotish is still in development; its current scenarios are described on the product page. Products and links.

The team publishes source code for products and tools for agent work: memory, a knowledge base from voice notes, Russian text editing and skill management. Open projects include Iva and agent-second-brain. Source and documentation let you study how solutions work and use them per the specific repository license. Open developments and GitHub.

The Majento blog covers AI agents, development and business implementation. We look at task formulation, document-based answers, Excel analysis, employee training and agents in group chats. There are materials on lost requests, customer call analysis and daily digests for managers. Articles help understand a task before the first project conversation.

Tell us what task you want to solve, who handles it and what systems you use. You can start with team training, systems analysis or development, we choose based on your situation. The first consultation is free. Project cost and timeline are defined after analyzing the task, data and integrations. Discuss your task on Telegram or write to hello@majento.ai.