Reflexive Tools, Reflective People

This week Tobi Lütke, Shopify’s CEO, published an internal memo that’s been passed around everywhere. The core of it: using AI reflexively is now a baseline expectation for everyone at Shopify. It’ll be part of performance reviews. And before asking for more headcount, teams need to show why the work can’t be done with AI.

The reactions split along predictable lines. Some people called it visionary. Others called it a layoff plan with better branding. I think it’s more interesting than either, and I’ve been thinking about one word in it: reflexive.

Reflexes are learned

A reflex is something you do without deciding to. But almost every useful reflex was once a deliberate effort. You learned to type by thinking about each key, until one day you didn’t. You learned to check the logs before guessing, to write the test before the fix, to ask “what problem are we actually solving?” - deliberately, many times, until it became who you are.

So I agree with the memo, in a sense. Using AI well should become a reflex for anyone who works with information. The people I see getting the most from these tools aren’t the ones who use them occasionally and carefully. They’re the ones for whom reaching for AI has become as natural as reaching for a search engine was twenty years ago.

But a reflex is also a thing you stop examining. And that’s where I’d add a word of my own.

The reflection must stay

The danger of a reflexive tool is that it can quietly replace a reflective person.

When the answer appears instantly, it’s tempting to stop asking whether it’s the right question. When the draft is already written, it’s tempting to stop asking whether it should be written at all. When the code works, it’s tempting not to ask whether it’s good. The tool is fast. Reflection is slow. Speed has a way of crowding out slowness, unless you protect it on purpose.

I’ve noticed this in myself. On days when I use AI for everything, I produce more. I’m not sure I think more. Some of my best decisions this year came from closing everything and sitting with a problem on paper, without any help at all.

What I’d want from a team

If I were writing a memo like this - and I suppose this post is one - I’d ask for two habits, not one:

Use the tools by reflex. Don’t treat AI as a special occasion. Make it part of how you work, every day, until you know its strengths and weaknesses in your bones.

And protect the time to think. Regularly step away from the tools and ask the slow questions. What are we building, and for whom? What would happen if we didn’t? What are we not seeing because everything is moving so fast?

The first habit makes you productive. The second keeps your productivity pointed in the right direction.

On headcount

The part of the memo that worried people most was the headcount rule. I understand the worry. But I’d frame it a little differently. The question was never “can AI do this instead of a person?” It’s “what can a person do with AI that neither could do alone?”

Asked that way, the answer is often: more than you’d think. A thoughtful senior person, using these tools fluently, can carry responsibilities that once needed a team. That’s not a reason to value people less. It’s a reason to value the right people much more - the ones with judgment, care, and the habit of reflection.

The oldest question

There’s a teaching in Pirkei Avot: who is wise? The one who learns from every person. I’d extend it, carefully, to every tool. Learn from it. Use it constantly. Let it make you faster and broader.

But keep asking who you’re becoming as you use it. The tool will not ask that for you.

Let the tools become your reflexes. Don’t let them become your judgment.