Delegate to an AI Agent — Hand It Your First Recurring Task
AI just became a colleague, not a chatbot. With ChatGPT Work and agent tools, you can hand off a whole recurring task. Learn to write the reusable brief that gets it done right.
The idea
In July 2026 OpenAI launched ChatGPT Work (and agents landed across the big tools) — AI that connects to your files and apps, makes a plan, and completes multi-step tasks like reports, research, and spreadsheets largely on its own, working in the background. This is the shift from AI-as-answer-machine to AI-as-colleague. The skill that matters isn't "try it once" — it's learning to delegate a recurring task by writing a reusable brief, exactly like Tuesday's delegation to a person: what, why, done-looks-like, guardrails. Write it once, and the agent does that task for you again and again.
Who is this for?
Neha Iyer
Market Research Analyst · Bengaluru
Neha spent every Monday morning pulling competitor updates into a summary — three hours of copy-paste.
She wrote a reusable brief for an AI agent: sources, what to extract, the format, what to flag. The agent swept and drafted; she reviewed in 20 minutes.
Now she runs the same brief every Monday. Three hours became twenty minutes — spent on the analysis only she can do.
"I stopped doing the task and started managing it. The agent is the intern; I'm the editor."
What you'll get
One recurring task handed to an AI agent
A real multi-step task briefed and run by an AI — not a one-off question.
A reusable agent brief
The what / why / done-looks-like / guardrails format you reuse every time.
The "manage it like an intern" mindset
Direct and review the agent's work instead of doing the task yourself.
A repeatable time-saver
A workflow you've handed off for good — hours back, every recurrence.
Watch
Jeff Su · 4.6M views · ~10 min
Do the challenge
1Pick a recurring, multi-step task
Choose something you repeat that has clear steps and clear inputs — a weekly summary, competitor scan, data cleanup, first-draft report. Recurring + multi-step is exactly where agents earn their keep.
💡 Great agent tasks look like intern tasks: clear inputs, a defined output, and steps that follow rules. Fuzzy, judgment-heavy work stays with you (for now).
2Write the reusable agent brief
Same four parts as delegating to a person: WHAT (the steps), WHY (context), DONE-LOOKS-LIKE (an example of a good result), and GUARDRAILS (what to avoid, what to flag for you). Clear briefs are the whole game.
💡 You are my assistant handling a recurring task. Follow this brief every time. WHAT: [the steps, in order] WHY / CONTEXT: [why it matters, who it's for] DONE-LOOKS-LIKE: [describe or paste an example of a good output] GUARDRAILS: [what to avoid; what to flag to me instead of guessing] Today's inputs: [paste this run's inputs]. Do the task and return the result in the format above.
3Run it and review like a manager
Let the agent do the task, then review the output the way you'd review an intern's: spot-check facts, check the format, fix what's off. Your job shifts from doing to editing — that's the skill.
4Save the brief and reuse it
Keep the brief somewhere handy. Next time the task comes up, paste it with fresh inputs — no re-explaining. That reuse is what turns a one-off experiment into hours saved every week.
Your template
DELEGATE TO AN AI AGENT — Reusable Brief TASK (recurring + multi-step): ________________ WHAT (the steps, in order): 1. ___________________________________________ 2. ___________________________________________ 3. ___________________________________________ WHY / CONTEXT (why it matters, who it's for): ______________________________________________ DONE-LOOKS-LIKE (example of a good output): ______________________________________________ GUARDRAILS: - Avoid: ______________________________________ - Flag to me (don't guess): ___________________ HOW I RUN IT: □ Paste brief + today's inputs □ Let the agent complete it □ Review like a manager (spot-check + format) □ Save the brief → reuse next time
Knowledge Check
5 quick questions to make sure the main ideas landed.
Know someone who needs this?
Share the challenge — streaks are more fun together.
Tomorrow
Single-Task on Purpose — Beat the Myth of Multitasking
Multitasking feels productive and quietly wrecks your output. The brain can't do two focused things at once — it switches, and pays a tax each time. Learn to single-task on purpose.
