VITALIY
VOLOSHIN

From business knowledge to content and AI sales

Agentix is a platform for content creation and sales automation with AI. It brings company materials, publication planning and customer conversations into one workspace. The team collects product information in a knowledge base and uses it to prepare text, images and video. Materials can be adapted for different team members and published to social channels on a schedule. AI sellers use the same knowledge base to answer customer questions and help turn enquiries into sales leads or purchases. This connects promotion and sales in one workflow, from product information to a conversation with a potential customer.

Try Agentix

The challenge — Rework the platform and establish a foundation for growth

I joined an existing platform built by an external team. It already had projects, subprojects, file uploads and content generation, but key workflows were unreliable: social connections failed, the AI seller barely worked and campaign planning needed substantial rework. The client needed a developer who could understand the architecture, refactor the code and continue building the product. We began by working through the platform’s goals, business processes, roles and feature relationships together.

My contribution — From investigating failures to doing the core product development

I joined Agentix when it needed substantial technical rework and took on the main share of refactoring and subsequent development. I first had to understand the inherited code and the causes of unreliable behavior, then work with the team to define and implement a coherent product model. My work spanned business processes, architecture, data models, APIs, interfaces, AI workflows and integrations. Alongside fixing existing problems, I developed new functionality and connections across the platform, from company knowledge and content production to publishing and customer conversations.

Understanding the inherited platform
Investigated the existing projects, subprojects, files, generation tools, social connections, campaign planning and AI seller. Traced how these parts worked together, where relationships broke down and why user workflows failed. Understanding the code, data and product logic was essential to addressing the causes of the problems.
Product design with the team
Worked with the team to define how the product should operate, from maintaining company knowledge to preparing publications and handling enquiries. Designed user workflows, responsibilities and access rules. Used that shared model to connect features and translate agreed decisions into data models, backend logic and interfaces.
Architecture and substantial refactoring
Carried out the main share of the existing code’s rework. Investigated problematic areas, changed relationships between modules and aligned behavior across the data schema, backend logic and interface. Changes to campaign plans, permissions or materials had to fit the rest of the user journey.
Business knowledge and project structure
Developed connections between a project representing the overall business and subprojects representing its divisions. Worked on how company knowledge, files, presentations and generated materials feed content workflows and AI sales. Shared business context had to work with the materials of a specific division.
Content and network templates
Developed and connected the stages of content preparation: source materials, templates, storyboards, generation and saved results. Designed shared template workflows so the team could define a common starting point and participants could adapt materials for their own channels. Connected this process to project data and access rules.
Campaigns and publishing schedules
Reworked the connections between campaign planning and material preparation. Campaigns need goals, business context, content formats and volume, while individual entries connect to storyboards and completed files. Worked on the path from a planned topic to content creation, attachment and a chosen publishing time.
Social integrations and messaging
Worked on integrations and their connections to the platform’s internal workflows, including linked accounts, publishing and customer communication. Refined seller binding, webhooks and message sending. External events needed to align with application data and states so publications and conversations remained part of the overall process.
The AI seller as a distinct system
Developed the barely functioning seller workflow across the data model, API and interface. Worked on project and subproject knowledge, conversation goals, scenarios, funnels, channel binding and message handling. The seller needs business context and configured rules to guide conversations toward the next action.
Roles, permissions and implementation across layers
Implemented account modes and access rules for features, projects and materials according to participant roles. Connected these rules to campaign plans and other workflows. Performed the main share of product development across layers, keeping entities, relationships, backend behavior, APIs and interfaces aligned.

From product knowledge to customer conversations

  1. Set up a workspace

    Work in Agentix starts with registration and choosing a subscription. Users then enter their workspace and create a project for their business. A project brings together company information, source materials and the tools needed to prepare content and handle customer inquiries. Multiple businesses can be managed in separate projects. Within a project, subprojects separate business areas or product lines, each with its own files and materials. This helps organise work before the first campaign: deciding where shared company information belongs and where to keep materials for a particular area. The next step is to fill in the knowledge base that will provide context for content generation and AI sellers.

    Account registration
    Subscription selection
    Workspace and projects
  2. Separate the business from its divisions

    The knowledge base holds the information the platform uses to create content and communicate with customers. Users describe the company, products, audience, offer and communication style. The information is organised into sections, making it possible to complete the business profile progressively and see which areas are ready and which still need attention. Documents, presentations, images and other source materials are stored in the library for the relevant business area. Shared company knowledge is complemented by materials from a specific subproject, providing the context needed for a publication or an AI seller. This structure supports multiple business areas, each with its own materials and a shared foundation of company knowledge.

    Business knowledge, organized by topic
    The business, offer and positioning
    Uploaded documents and visual materials
  3. Create videos, images and avatars

    The production section brings together tools for creating videos, images and avatars. Each task has a dedicated interface: users choose a mode and prepare source materials in the video studio, while image generation starts with a description and output settings. A separate avatar tool lets users prepare a character for video content. Production connects to the next stages of work: results can be saved, used in a template or included in a campaign. Users can first prepare individual elements, such as an image or video, and then use them in publications. The workflow takes an initial idea and generation settings through to materials that can be used elsewhere in the platform.

    Video production studio
    Image generation
    Avatar creation
  4. Create a shared starting point for the network

    A template provides a starting point for content used by different team members. Its creator prepares the format, source materials, text and generation settings. The team defines which elements are shared and what a participant can adapt for their own publication. The prepared version can then be reused while retaining the idea and structure of the material. Distributors find available templates in a dedicated library. They select an option, open it and prepare their own content from that foundation. This connects the team creating templates with participants using them in their channels. Shared materials do not need to be assembled again for every publication: each participant starts from a prepared template.

    Template creation
    Member template library
    Creating content from a template
  5. Turn an objective into a content schedule

    A campaign plan connects a promotion goal to specific publications. Users describe the task, choose a period, select formats and decide how much content is needed. Plans can be prepared manually or with AI using business knowledge. This produces topics and campaign items for which users then create content or select existing materials. The resulting plan appears in a calendar, showing planned publications and their dates. Content is attached to campaign items, with publishing times set for the selected channels. Distributors have their own calendar for working with the plan available to them. This connects content preparation to a schedule and makes the campaign a sequence of concrete actions.

    Campaign creation
    Campaign calendar
    Member publishing calendar
  6. Deliver content to the selected channels

    To publish content, users connect the social accounts where it will appear. Connections are managed in a dedicated section, while completed content is available in the member’s material library. From there, users select a publication and proceed to publishing with the text, images or video already prepared. Content creation and distribution become stages of the same process. Before publishing, users select a channel and publication settings. Content can be scheduled for a specific date and time, with its sending status tracked in the platform. This section shows the full path: connecting accounts, working with ready materials and examples of published content on Instagram and TikTok. It connects the work inside Agentix to the resulting social posts.

    Social account connections
    Ready materials and publishing
    Published content on Instagram and TikTok
  7. Respond with business context

    AI sellers use business knowledge to handle customer inquiries. The platform supports multiple sellers, each configured with product information, a conversation goal and a scenario. These settings determine which materials the seller uses and which next step it should guide the conversation towards, such as an inquiry, meeting or purchase. Conversations are collected in an inbox where the team can review inquiries and each customer’s message history. A team member can inspect replies and join a conversation when needed. The workflow covers seller creation, configuration and monitoring of customer conversations. The knowledge base used to prepare content also provides context for answering questions about the company and its offer.

    AI seller list
    Seller creation and settings
    Customer conversations

Architecture connecting the whole process

Projects, subprojects, knowledge, materials, templates, campaigns and sellers form the product’s core. The rework needed consistent relationships and permissions across data, APIs and interfaces. React connects to NestJS through GraphQL; PostgreSQL stores entities and relationships. Redis and BullMQ handle background tasks separately from the API, including long-running content operations. This structure supports a connected workflow while keeping responsibilities separate.

  • TypeScript
  • React
  • NestJS
  • GraphQL
  • PostgreSQL
  • Redis
  • BullMQ
  • Drizzle ORM
  • AI APIs
  • Docker

Let’s discuss
your project

Tell me about your project

Leave your contact details and a few words about the project. I’ll read your enquiry and get in touch to discuss it.