Working with AI agents: dots, Muse and Grok Bot
Learn the skills of working with an always-on AI agent: delegating a task with a clear brief, setting approval rules, reviewing the work, protecting data and talking about it at interview. Compares OpenAI dots, Meta Muse and Grok Bot, and works without access to any of them.
25 min · 11 pages · practice task · quiz · part of Work with AI
Sponsor this courseWhat you will be able to do
Skills you add to your Herizon profile when you finish: Task Delegation, AI Risk Management, AI for Productivity.
- What an always-on AI agent is, and how it differs from a chat assistant
- How OpenAI dots, Meta Muse and Grok Bot compare, and which of them you can use in Finland
- How to delegate a task with a five-part brief: outcome, sources, limits, deliverable and review point
- How to set approval rules, so an agent drafts freely but asks before it acts
- How to review an agent's work with a five-point checklist
- What to keep out of an agent's reach, and who is responsible when it makes a mistake
- How to talk about agent experience in a job interview without overclaiming
Who it is for
Working professionals who may soon be asked to supervise an AI agent, and job seekers who want to talk about it honestly. No technical background required. The course is designed for the Herizon community of international professionals in Finland, and the skills apply everywhere.
The situation you work through
Noora Salmi, a made-up operations coordinator at Kaari Foods, a made-up Helsinki food wholesaler, is asked to hand the weekly supplier report to an AI agent after the company gives staff ChatGPT Business Premium with dots.
Your practice task
Pick one recurring task from your work or job search and write an agent plan on one page, on paper, without any agent account. Part one is the brief with outcome, sources, limits, deliverable and review point, plus a schedule request and whether the agent may send anything. Part two is the approval rules, one rule for each action the agent might take, with anything that leaves your organisation, spends money or changes a password marked. Part three is a five-point review checklist with the evidence you would open for each check, tested on four lines of made-up agent output with one planted error that your checklist catches. Submit the page as evidence.
What is inside
- What an always-on agent is
- Three agents side by side
- The five-part brief
- Make it recurring
- Approval rules
- Drafts, sending and stopping
and 5 more pages
Finish with a short quiz to earn your certificate.
More in Work with AI
- Understanding AI49 min · 15 pages
- Using AI responsibly at work24 min · 10 pages
- Getting started with Claude31 min · 16 pages
- No-code automation with n8n35 min · 11 pages
- Using Grok for work and the job search22 min · 10 pages
- Grok Bot: an AI teammate that works while you don't26 min · 11 pages
- Notion for teams and your job search29 min · 11 pages
- ChatGPT at work25 min · 12 pages
- Gemini Notebook: research from your own sources20 min · 10 pages
- Microsoft Copilot at work21 min · 10 pages
- Granola: AI meeting notes without a bot29 min · 12 pages
- Mistral Vibe (formerly Le Chat): an EU-based AI assistant24 min · 11 pages