Budget: 3000 UAH Deadline: 3 days
Welcome! The Business Atlas team is ready to implement your project for developing an AI assistant for account managers. Your specifications clearly describe the RAG (Retrieval-Augmented Generation) architecture, which is our area of expertise.
We propose to avoid expensive custom backend development from scratch and build the system based on a flexible low-code architecture and AI agents. In my experience, there are over 50 successful projects in the UA/EU markets where this approach allowed for the implementation of complex solutions significantly faster, cheaper, and without loss of stability.
Our vision for technical implementation:
• Request processing and RAG (n8n / Make): The n8n platform will act as the main conductor of the system. When the manager writes in Telegram, a webhook sends the message to n8n, where the request is transformed into a vector (embedding). The system searches for relevant chunks of data in the vector database, passes this context to the LLM (OpenAI/Claude), and returns an accurate response to the manager without hallucinations.
• Data sources (Knowledge Base): We will set up automatic reading and updating of the knowledge base from your Google Docs and prepare modules for uploading and parsing exported Telegram chats.
• Security and access: We will implement user rights verification. We will create a simple registry database (for example, in Airtable, Supabase, or PostgreSQL — we have extensive successful experience deploying n8n + Airtable connections). Before each response, the bot will verify the user's Telegram ID, and if the employee's status changes to "Inactive," the system will immediately block access.
Our experience in similar projects:
My portfolio already includes ready-made RAG solutions that we can quickly adapt: autonomous AI agents for onboarding and training staff, which are perfectly oriented in internal corporate documentation, as well as AI agents for e-commerce that analyze the context of dialogues with clients.
Estimated timelines and costs for MVP:
• Parsing the knowledge base (Google Docs, chats) and setting up the vector database: 1 week.
• Creating the RAG workflow in n8n, integration with LLM and Telegram: 1 week.
• Setting up the access system, logging, and final tests: 1 week.
Total implementation time: 2.5 – 3.5 weeks.
Estimated cost: $2,500 – $3,500 turnkey (depending on the volume of initial data and choice of vector storage).
I personally lead the architecture and AI logic design, while our technical specialist Lavr will implement the technical assembly of the workflow in n8n/Make. We finalize the exact budget after conducting a brief expert diagnosis.
Are you ready to discuss the details in person?