I Stopped Trying to Be Productive With AI. Here's What I Started Doing Instead.

By Sadie St. Lawrence | Human Machine Collaboration Institute

For the past few months, I have been running an experiment on myself.

I stopped trying to be productive with AI.

Not stopped using AI. I use AI every day, more than most people I know. But I stopped using AI to make myself faster, and started paying attention to what happened.

The results have surprised me enough that I want to write about them, because I think we are collectively telling ourselves the wrong story about what AI is for. And the story we are telling is making us worse at the work that actually matters.

Let me explain what I mean.

The dominant narrative right now is that AI is a productivity tool. It writes faster than you can. It codes faster than you can. It researches faster than you can. It summarizes, drafts, categorizes, plans, and generates at speeds no human can match.

The implicit conclusion of that narrative is that you should be using AI to do more, faster, in less time.

I bought that narrative for a long time. I built it into my own workflows. I structured entire weeks around what AI could accelerate. I felt the surge of getting through my inbox in twenty minutes instead of ninety, drafting a strategy document before lunch that used to take three afternoons, generating a research summary in seconds that would have taken hours.

And then I noticed something uncomfortable.

The faster I got at the work AI could accelerate, the less time I was spending on the work AI could not touch. The deep thinking. The relationship-building conversations. The genuinely creative work that requires you to sit with a problem long enough for a real answer to emerge. The strategic decisions that only get made when you give yourself permission to not be immediately productive.

AI had made me faster at a specific category of work. And in doing so, it had quietly displaced the work that only I could do.

That is the productivity trap. It is the assumption that faster is always better. That efficiency is always the goal. That the value of AI is measured in tasks completed per hour rather than in the quality of what you produce over the course of a year.

I do not believe that anymore.

The reframe I have been running is simple to describe and harder to actually practice.

Instead of asking how can AI make me more productive, I started asking a different question.

What kind of work do I want to protect from AI, so that I can do it better?

Not to protect from AI in the sense of never using it. To protect in the sense of not letting AI substitute for the parts of the work that require me. The parts where the value is in the doing, not in the output.

Writing this post is a good example. I could have used AI to generate a first draft of it in about ninety seconds. It probably would have been fine. Coherent, structured, hitting all the expected beats.

But I would not have arrived at the reframe I actually want to share with you if I had done that. Because the reframe emerged in the writing itself. In the process of trying to articulate what I have been noticing, I discovered what I actually think. And no amount of AI-accelerated drafting would have gotten me there faster. It would have gotten me to a coherent but shallower version of the same idea.

There are certain kinds of work where the shortcut is the destruction. Where the friction is the point. Where slowing down is not inefficient. It is the entire mechanism by which the work produces value.

I have started guarding those kinds of work more carefully. And I have started using AI more intentionally in the places where it genuinely serves me. Not to be faster. To be freer.

Because I still open AI tools many times a day, I needed a way to check myself before I did. So I developed a simple test I run in my head before I open Claude, or any other AI tool, for a substantive task.

Three questions. Ten seconds. That is it.

Question 1: What am I actually trying to produce here?

Not the artifact. The value. If the goal is to send a coherent email to a partner about a scheduling conflict, the value is fast and functional. AI is a good fit. If the goal is to think through a hard strategic question, the value is the thinking. AI as a shortcut will produce an artifact that misses the point.

Question 2: Would using AI here make me better at this over time, or worse?

This is the question most people never ask. Every use of AI is not just an event. It is a training experience for you. Are you building a capability, or are you outsourcing one? Some things you want to outsource. Others you want to keep sharpening yourself. Know which is which before you make the choice.

Question 3: What am I giving up by moving fast here?

The hidden cost of speed is depth. If moving fast means missing the ten minutes of reflection that would have changed the direction of the work, the speed is not a gain. It is a loss disguised as a win. Ask this question before you accept the acceleration.

Three questions. That is the whole test. If AI passes all three for a given task, I use it without hesitation. If it fails any one of them, I slow down and do the work myself.

I want to be honest about what this practice has produced, because it is not what I expected.

I am not less productive. If anything, my output over the last quarter has been the highest quality it has been all year. But it has been distributed differently. Fewer emails sent quickly. Fewer summaries generated in seconds. More thinking. More original writing. More conversations where I was present enough to notice what was actually being said.

The work I am proudest of from the last three months came from moments where I chose not to accelerate. Where I wrote the first draft by hand. Where I spent the extra hour in a conversation instead of trying to compress it. Where I let the strategic question sit unresolved for a few days before I answered it.

None of that would have happened if I had continued optimizing for productivity.

If you are reading this and finding yourself nodding, I want to leave you with a question worth sitting with.

What is the highest-value work you do, and how much of it are you currently outsourcing to AI?

For most people I ask this question, the honest answer is more than I intended to.

The productivity narrative has trained us to reach for the tool by default. To assume speed is good. To measure ourselves by throughput. But the work that actually determines your career, your impact, and your growth is almost never throughput work. It is depth work. And depth work rewards a completely different practice than productivity work does.

The reframe is not to stop using AI. It is to stop using AI as a substitute for the work that requires you.

If you want a longer version of this argument, and a full framework for developing the orchestration skill of knowing what to accelerate and what to protect, that is what Becoming An AI Orchestrator is built around. The book calls it the practice of intentional non-use. It might be the most important chapter I wrote.

What is the highest-value work in your life that you are currently trying to speed up, and what would change if you slowed it back down? Hit reply. I read every response, and this is exactly the conversation I want to keep having.

Let’s be honest: most of us are still using ChatGPT like an over-qualified search bar or a draft writer. But OpenAI’s new GPT-5.6 update and ChatGPT Work completely shift what’s possible, moving the platform from a back-and-forth assistant into a true work orchestration engine. In this hands-on breakdown, I take a single messy folder of emails, meeting notes, and raw files and turn it into five executive-ready deliverables without manual heavy lifting, including an executive brief, a cleaned financial model, a board deck, an interactive project dashboard, and a recurring scheduled workflow. See how AI is moving from simple retrieval to true task coordination, and where your human judgment remains essential.

Watch the full 8-minute masterclass on YouTube to level up your workflow!

Sadie St. Lawrence is the Founder & CEO of the Human Machine Collaboration Institute and author of Becoming an AI Orchestrator. She writes weekly about the future of human-machine collaboration, AI in practice, and what it actually takes to build at the frontier.