You set up an AI agent. Now what? 5 first jobs that actually help

You installed an agent (Grok Bot, Hermes, OpenClaw, Muse...) and then didn't know what to ask it. Short answer: 5 steps. 1) List what you did more than twice last week; 2) Start with a read-only daily briefing; 3) Give it one scheduled watcher; 4) Hand over one draft-don't-send task; 5) Write down what worked, then expand.

Last updated: Thursday, October 15, 2026

How do I fix this?

1. List what you did more than twice last week

   Open your calendar, inbox and browser history. Write down anything repetitive: summarizing emails, checking a site for changes, drafting the same kind of reply, collecting links. Those are agent jobs.

2. Start with a read-only daily briefing

   Ask the agent for a morning summary: unread email that needs you, today's calendar, and 3 news items on a topic you care about. Read-only jobs can't break anything and show you quickly what the agent is good at.

3. Give it one scheduled watcher

   "Every day at 8am, check [page] and tell me if [price/status/post] changed." Watchers are where agents beat chatbots, because they run without you asking.

4. Hand over one draft-don't-send task

   Let it draft replies, posts or reports, but you press send. Review its drafts for a week before giving it any permission to act on its own.

5. Write down what worked, then expand

   Keep a short note of the prompts that worked and save them as the agent's standing instructions or skills. Add one new job per week instead of ten at once.

Which tools can do this for me?

All tools for “Automate busywork with AI” →