AI Personal Assistants

This is for you if

• You have a large amount of information, tasks, decisions and personal or professional knowledge, and want a unified system for storing, retrieving and using them.
• You want an assistant that does more than answer questions and can execute workflows, send emails or perform defined actions on your behalf when needed.
• Context preservation matters to you, and you want the assistant to maintain connections between conversations, information, projects and stored knowledge over time.

When not to hire me for this

If you only need a general-purpose chatbot for asking questions and receiving answers, without personal memory, access to your knowledge or the ability to execute real actions, you probably do not need this level of system. The goal is not to build another chatbot; it is to build a personal intelligent layer that brings your knowledge, context and required tools together into a usable system.

How it runs

1. 1

   Understanding your workflow and information structure

   First, we define what information the assistant needs access to, what tasks it should manage, and where you currently store your information, tasks and knowledge. The way the assistant will support personal, professional and management-related work is also defined at this stage.

2. 2

   Building memory and the second brain

   Your personal knowledge and information are organised locally so the assistant can reach the context it needs without relying on a centralised cloud memory. Obsidian can serve as one of the main sources of knowledge and information structure.

3. 3

   Connecting the assistant to your tools

   The assistant is connected to the tools and environments used in your workflow. Telegram, Obsidian and Claude Code can each play a different role in the structure, with information and outputs moving between them.

4. 4

   Defining execution capabilities

   At this stage we define which actions the assistant should only suggest and which it should actually execute. Workflow execution, email sending and other required actions are defined within specific boundaries, so the assistant is more than a response system.

5. 5

   Building the interaction layer

   The assistant can be reached through the Claude Code environment, while its outputs are available in different places. Depending on the task, the result can arrive in Telegram, in Claude Code or by email.

6. 6

   Testing, deployment and iteration

   Real usage scenarios, context retrieval, response accuracy, workflow execution and information access are tested. After deployment the structure is developed further based on actual usage, so the assistant becomes increasingly aligned with the way you work.

What you are left with

A personal intelligent assistant with access to your knowledge and context that can search and manage information, generate reports, manage tasks and execute defined workflows and actions, while your primary data and personal memory stay on your own machine.