Log in

Using Personas With Muse by Meta and OpenAI Dots: Getting Started

Updated: November 12, 2026 · By Brian Arrington (Founder, AI Consultant and Business Coach)*

The short answer

Muse by Meta and OpenAI Dots can both run a BotPlaybooks persona, but each holds the instructions in a different place. Muse uses Soul.md. A dot takes working instructions in chat, with guardrails turned into Custom Rules. Start with one job and one agent.

Some of the newest AI assistants are always-on agents tied to one person. Muse by Meta and OpenAI Dots both work this way. That changes how you use a persona, because you do not get a fresh Bot for every job. Here is how BotPlaybooks recommends handling each one.

The key difference: choosing one job

The BotPlaybooks guides recommend running one persona at a time on each of these assistants, because you typically have one Muse or one dot rather than a separate Bot for each job. Check the platform's own help pages for current details.

For a business owner, that means choosing. Pick the persona for the job that matters most this week, run it well, and switch when your priorities change.

Muse by Meta

According to the BotPlaybooks Muse guide, you can ask Muse in chat to change its personality, or edit its personality and identity settings yourself. For Meta's own instructions, see Meta's help page on how to customize Muse's personality and memories. Menus and file names may change, so check that page before you start.

The steps:

  1. Back up first. Copy the text of your current personality settings somewhere safe.
  2. Add the persona. Paste the persona's prompt at the end of those settings under its own heading, such as "## Persona: Brief."
  3. Name it. Set the name, and a tagline if you like, in Muse's identity settings.
  4. Test it. Open a new chat and send the persona's example first task.
  5. Switch when you need to. Remove that block or restore your backup. The BotPlaybooks guide also suggests asking Muse: "Go back to your default personality."

OpenAI Dots

The steps from the BotPlaybooks Dots guide are below. For OpenAI's own instructions, see OpenAI's Dots documentation.

  1. Give it the persona. Message your dot with the persona's prompt and ask it to follow it as its working instructions, for example: "From now on, work as Brief, my Call-Prep Briefer."
  2. Name it to match the persona if you like. Naming also updates its handle.
  3. Turn the guardrails into Custom Rules. Custom Rules in your ChatGPT settings let you choose which supported actions your dot can take on its own, which need your approval and which it should not take. "Always draft before sending" becomes "Ask before sending any email." "Never delete or merge CRM records" becomes an "Ask before taking action" or "Hand off to you" rule.
  4. Connect only the tools the persona lists, in ChatGPT's Plugins tab.
  5. Test with the persona's example first task. To switch personas, tell your dot to stop using the old instructions, give it the new persona and update your Custom Rules.

The guide also notes that Custom Rules and instructions cannot override ChatGPT's built-in safety requirements.

Why the guardrail step matters most

Step 3 for Dots is the one owners skip, and it is the one that protects you. A persona's guardrails are written as instructions. Custom Rules turn them into settings. If your assistant can act in your accounts, put the approval rules where the platform enforces them, and keep reviewing what it drafts.

Which persona first?

If you can only run one at a time, start with the task that costs you the most hours each week. The persona finder at botplaybooks.com/find can suggest good fits from 175+ personas based on a few sentences about your work.

If you could run only one persona this week, which job would it take on?

BotPlaybooks is independent and not affiliated with or endorsed by Meta, OpenAI or any AI platform. Product names belong to their owners.

BotPlaybooks is a brand of Arrington Consulting LLC.

Disclaimer: This post is general information, not professional, legal or financial advice. Results vary. AI can make mistakes, so always review outputs before acting.