How to build an AI personal assistant

Most people use AI reactively: opening a new chat, re-explaining who they are and what they need, and starting from zero every time. A custom AI assistant fixes that. You set up your role, communication preferences, audience, and key documents once, and every conversation after that starts from that foundation instead of scratch.

Most enterprise AI platforms, such as ChatGPT, Gemini, and Copilot, let you create a saved configuration, sometimes called a custom GPT, a Gem, or a bot, that holds everything you'd otherwise spend 5 minutes re-explaining each session. The difference is most obvious on recurring tasks: weekly updates, recurring reports, types of emails you send often. You open the assistant, give it new information, and it already knows how you want it handled.

This video walks through a simple 4-step setup (role, communication preferences, audience, and reference documents), four ways to use it as a weekly operational layer, and the three most common mistakes people make when building one.

What this video covers:

1. Why starting from zero every AI session is the biggest productivity leak most people don't notice
2. What a custom AI assistant is, and how it differs from a regular chat
3. Four-step setup: role, communication preferences, audience description, and key reference documents
4. Four weekly operational uses: prioritizing your to-do list, gut-checking your calendar, identifying what you're missing, and planning what carries over
5. Why specificity is what separates a convenient tool from one that changes how you work
6. Three common mistakes: over-engineering before using it, loading in too much, and setting it up but not building it into your workflow
   
   Videos by Sammy Goldstein and Bridget Griswold.