Let ChatGPT Tell You What to Automate

The most common reason people aren’t automating more of their work isn’t that they don’t know how. It’s that they don’t know where to start.

Which tasks are good candidates? Which workflows are worth the time to build? What are you doing manually that you shouldn’t be?

For most people, figuring that out requires a deliberate audit — sitting down, reviewing your week, making a list of pain points, prioritizing by frequency and effort saved. It’s a useful exercise, but it’s also another thing to do, and it rarely happens more than once.

There’s a simpler approach. You can use ChatGPT to do this for you automatically every day.

The Setup: ChatGPT Tasks

ChatGPT has a Tasks feature that lets you schedule recurring prompts — things you want it to check in about, remind you of, or surface on a regular cadence.

Set one up with a prompt like this:

“Based on what you know about my work, my daily tasks, and the things I’ve described doing manually in our conversations, suggest one AI workflow or agent I should consider building. Be specific about what the agent would do and why it would save me time.”

That’s the whole setup. One daily task. Takes about two minutes to create.

Why It Works

ChatGPT has your conversation history. If you’ve been using it regularly — for research, drafting, brainstorming, problem-solving — it has a detailed picture of how you actually work. Not the idealized version of how you work, but the real version, including all the friction.

It knows about the report you’ve mentioned regenerating every Friday. The meeting prep you described doing manually. The follow-up emails you keep drafting from scratch. The task categories that come up in your conversations week after week.

You’ve stopped noticing most of these things. They’ve become part of the invisible fabric of your job — accepted as just “what the work requires” rather than as tasks that could be handled differently. When you build up enough familiarity with something, the option to change it stops feeling available.

ChatGPT hasn’t developed that familiarity. It looks at your patterns from the outside, and from that angle, the automation opportunities are often obvious.

What the Suggestions Look Like

In practice, the suggestions tend to fall into a few categories.

The frequently-repeated task. “You’ve described writing this type of email three or four times in our conversations. An agent could draft these from a template triggered by a specific condition.” These are the easiest wins — high frequency, clear pattern, immediate time savings.

The multi-step manual process. “You mentioned that every time X happens, you manually do Y, then Z. That sequence could be automated.” These take a bit more setup but often have the highest payoff because each individual step doesn’t feel onerous, but the accumulated time across the sequence is significant.

The friction point you’ve learned to work around. “You frequently describe workarounds for this type of task. The workaround suggests there’s something that could be handled more directly.” These are the most valuable suggestions because they point to pain you’ve adapted to rather than addressed.

The “I Never Thought of That” Moment

When I started doing this, I was genuinely surprised by what came up.

There were patterns in my own work that I’d been describing to ChatGPT for months without registering as automation candidates. Things I’d stopped thinking about fixing because I’d gotten good enough at them that they didn’t feel like a problem anymore.

ChatGPT pointed them out anyway. “You’ve been talking about this. You should build an agent for that.” And the response each time was some version of: right, I never actually thought about that.

That’s the specific value here — not that ChatGPT gives you ideas you couldn’t generate yourself, but that it removes the self-filtering. When you audit your own work, you naturally skip past the things you’ve accepted. When an outside perspective looks at the same patterns, the accepted friction is still just friction.

Connecting It to Action

The daily suggestion is only useful if you do something with it.

A few ways to handle this:

Log suggestions and pick one per week. Not every suggestion needs to be acted on immediately. Keep a running list and pick the one that has the most impact given what you’re working on that week.

Use the suggestion as a starting point. When ChatGPT suggests an agent idea, follow up with: “Walk me through what this agent would need to do step by step” or “What’s the simplest version of this I could build?” The first suggestion gives you the idea; the follow-up conversation gives you the build plan.

Apply the frequency test. The easiest candidates are tasks you do more than once a week. If ChatGPT surfaces something that fits that threshold, it’s worth building immediately. The time savings start compounding quickly.

Most people build their AI agent library reactively — they discover a workflow exists after reading about it somewhere and then try to replicate it. The ChatGPT daily task flips that: it surfaces the workflows that are specifically relevant to your work, based on what you’re actually doing.

The tool you’re already using every day knows more about your work than you remember. Ask it what to build.

How to set this up: Open ChatGPT, go to Tasks (or use the prompt “Set a daily reminder for me to…”), and create a recurring daily task with the prompt above. You’ll start getting suggestions in the next 24 hours.