The challenge
Work information is spread across several systems. Answering a simple question about a project’s status or finding a relevant document requires switching between them and assembling information manually.
The solution
A voice AI assistant in Telegram brings access to work sources into one interface. Users ask questions by voice or text and receive answers in their preferred format.
The assistant supports follow-up questions, searching connected materials, and saving information on explicit request. Access is limited to permitted actions.
Scenarios
- Planning: “What tasks are planned for today?”
- Project status: “What is done and what remains?”
- Capturing ideas: “Save a note.”
- Knowledge search: “Find materials on this topic.”
- Follow-up: “Show the source of the answer.”
How it works
The solution combines speech processing, an agent module, integrations, and background processes. The agent module handles requests and consults permitted sources. A text repository and knowledge base help locate relevant materials.
Business application
The assistant can provide a unified interface to work systems, helping employees find information, check progress, and capture ideas.
My role
I defined scenarios, designed the architecture, and built the solution with AI agents. I connected the voice interface, source search, and request processing.
Result
- One interface for several work scenarios.
- Voice and text queries and replies.
- Search across materials with source references.
- Validation through daily use.