My official title is “Creative Director”. I’ve been a creative professional in one capacity or another for almost two decades. But 85% of my job today doesn’t involve any “creativity” at all. It’s mostly all that other nonsense that I don’t really care for. Project management, stakeholder communications, team management, finance, invoicing, admin, legal…

The list is as long as it is boring. And so far, it all felt like the price I had to pay to do the thing I loved.

Blink twice if you can relate.

No one on the internet has stronger feelings about AI than creative professionals and artists. But is the AI (or anti-AI) debate in the community possibly a little bit misguided? Outsourcing all of our thinking to AI and going full YOLO (hello, fellow kids) is obviously not the best idea, but neither is sticking our fingers in our ears and going “Lalalala!”

As machines become increasingly capable of doing everything for us, it’s important to be discerning about what exactly we should let them do.

The thing that excites me the most about AI is that I can use it to do all the parts of my job that I don’t want to do.

To that end, I ask myself 6 simple questions when deciding whether a task should involve AI. Maybe they’d be useful for you to consider, too.

1. Do I enjoy this task?

This one sounds obvious. It will become increasingly important in the coming years. If it fills you with joy when you do it, don’t use AI. Even when it feels like you’re staring down the barrel — maybe it’s long, difficult, daunting, whatever — don’t use AI.

It doesn’t have to be a creative task either. If you love making spreadsheets, don’t let AI take that joy away from you. Even if AI gets capable enough to create the Mona Lisa of Excel sheets, don’t let it do that for you. Make your spreadsheet and enjoy writing every goddamn formula in every goddamn cell.

2. Do I want to learn how to do this?

I’m a curious person by nature, and learn best by doing. AI is great for handling tasks I don’t care for or don’t want to learn to do myself, but there are things I’ve always wished I could do that felt too out of reach or too technical or too academic, until recently. AI has been the best teacher I’ve ever had, because it is patient, honest, and lets me make mistakes without judgment.

I’ve always wanted to build my own apps and tools, and my ultimate nerd fantasy has been to build my own AI assistant, aka Jarvis. Claude helped me realise a lot of those childhood nerd-inventor dreams, and I’ve enjoyed every second of the process.

It’s how I’ve managed to build Mimir, my personal AI operating system, because I’ve had to learn so much about development and code along the way. My system prompt in Claude Code instructs it not only to write the code I ask for but also to always explain what that code does. That way, when something isn’t going the way I want it to, I can now understand and articulate what I need to change. Instead of casino slot-machining my way through the builds, hoping something sticks.

3. Is this task getting in the way of the work I want to do?

My partner, Karanjeet, is a writer, journalist, editor and Head of Content at our agency. A couple of weeks ago, she used her phone’s audio recorder app to record an interview for her weekly column. She needed those audio notes transcribed, cleaned up, and summarised — immediately.

What she would’ve had to do without AI would be to find a reliable online transcription tool, sign up for it (for a monthly subscription), upload her audio files, get individual transcripts for each audio file, feed those transcripts into an LLM for consolidation, cleanup, and analysis, and only then get to the writing she loves.

Instead, I created a new folder for her, put all the audio files in there, pointed Claude Cowork at the folder and gave it a simple prompt — transcribe all these audio files, create a new text document in the same folder with the transcriptions consolidated and cleaned, and add a detailed summary of the interview to the top of that document.

Cowork took maybe 4 minutes to get the job done, and Karanjeet lost no time or brainpower in the process.

4. Will I need to do this task repeatedly?

I love my daily routine, and I’m generally a bit OCD about order and systems in my work. So it was fairly easy for me to identify repetitive tasks I was doing every day.

The easiest, lowest-hanging fruit was my morning update. Every morning, I would check my email, calendar, AI news, investments, and then find a playlist to fit the mood. All very manual and, in hindsight, quite tedious.

Now, Mimir sends me a text every morning with my “Daily Brief”, after scanning my emails, calendar, notes, even my investments and the latest news. Then creates a custom Spotify playlist of my favourite songs perfectly suited to the day ahead.

The coolest thing about this is that Mimir isn’t looking at all these sources independently of one another — it’s finding the connections between them, understanding how they affect one another, and then deciding how those connections might be relevant to me.

And you don’t even need to build your own OS to set up something similar for yourself — I’ve helped Karanjeet with a daily brief tailored to her needs, and that one’s delivered straight to her email inbox every morning. NBD.

5. Would a human do this better, and can I afford one?

No matter how clever the machines get, some tasks just work better when human ingenuity, taste, collaboration, and strategic judgment are thrown into the mix. Not to mention common sense, which still often eludes AI models.

Our firm has been remote-first since 2010, which means that quality time meeting with team members has always been a rare and precious experience for us. I wouldn’t give that up for all the AI agents in the world.

But we’re not always in the privileged position of hiring humans for every task we want done, so AI’s cost-effectiveness could be a valid advantage for some folks and tasks.

For example, could a qualified human assistant do everything I’ve set up Mimir to do? 100%. Can I honestly justify hiring a human assistant for myself? Unfortunately, not, I’m afraid. I’d much rather spend our limited agency budget on paying our creative team salaries, because they provide the company and me so much more value and ROI.

6. Am I outsourcing judgment instead of labour?

Just because machines can think now doesn’t mean they should be thinking for us. It’s hard to be objective about your ideas when you’re constantly being told every idea is “amazing” or that you’re a genius.

I hope this one makes my therapist proud, because it requires me to be brutally honest and self-aware about my motivations.

It’s tempting, of course, to skip the thinking sometimes. There have been plenty of instances when I’d be in a hurry to reach an outcome and jump into a conversation with Claude too early in my process. It might seem easier, and the output will sound great, smart even.

But one follow-up conversation about it with Karanjeet will instantly burst that bubble and bring me back to earth, because she will immediately point out all the holes in the machine’s thinking. She has some distance and can be critical and objective in a way that I — struck by the model’s agreeableness — might no longer be able to.

Objectivity, honesty, and rational thought are more critical now than ever. Sure, AI makes everything easy, but for the things that really matter, don’t let it do the thinking for you. Apply your mind, think through challenges on your own first. Discuss your ideas with your (human) team, friends, or partners and listen when they tell you your ideas are stupid.

If you do your thinking beforehand, and only then discuss your ideas with AI — and stay objective throughout that dialogue — you’ll find the output that much richer and more useful.

I have no interest in picking a side in the AI debate because I know it’s pointless. My goal is simply to know when to use AI and when to go it alone. That’s where I believe I can find an edge.

Use AI to remove friction, never to remove ourselves.

Read on Substack ↗