making AI a habit

Or: how to use AI in 10 days (ok, 30, I just couldn't resist the rom com reference)

Before we jump in:

1. If you’re a manager interested in AI, I recently appeared on Claire Vo’s podcast How I AI where I demoed how Custom GPTs can make you better at your job, plus shared some of my favorite prompts.

2. Last call: enrollment closes in one week for the June cohort of my AI for Managers Maven course. You’ll walk away with practical tools that save you 5-10 hours a week immediately, plus frameworks for for helping your team use AI to level up what they can accomplish.

---

30 Days of GPT

[tl;dr: If you’re looking for my 30 Days of GPT Learning Guide, it’s here].

I think about habit formation a lot. Like, all day, every day. Partially this is because when you build consumer products, you need to think a lot about how to get users to build habits around using your products, especially if the desired habits are, like, exercising and going to bed at a reasonable time rather than scrolling.

As a result, I see almost everything thorough the lens of habit formation, including learning. I talk to a lot of people about how best to “learn” AI, and what it means to be “fluent” in AI. The idea of being “fluent” in AI is very funny to me because any measurable skill you master today will be obsolete in like a week. So when I see CEOs saying that their new hires must all be “fluent” in AI, I’m like, what could that possibly mean? I’m not clowning on the idea, which I think is basically right. But I get asked all the time, how do I measure AI competence on my team? And I always kind of laugh at the question, because my answer is soooo vibes-based. I’m like, IDK, I observe and I judge and I adjust my approach based on what I’m seeing.

And what I am seeing is people attempting to improve their AI skills in a traditional “study and learn” way, which I think is wrong. When you’re studying a new skill, there are generally clear milestones, benchmarks, correct answers. If you’re learning Excel, you can learn the formulas and how to build pivot tables and be considered “proficient.” AI is very different for a number of reasons: it’s non-deterministic, it’s rapidly evolving, it’s highly context-dependent, it takes liberties with interpretation. Working with it regularly, you begin to develop an intuition for what it does well, the ways it tends to err or mislead, the way your behavior impacts how helpful it is. To me it is less like learning a skill and more like learning to work with another person. Which means the only way to get good at it is the same way you get good at working with any person: repeated exposure, constant small adjustments, building patterns of interaction that actually work.

This is why, if you want to level up in AI, you should think about it like building a habit, not acquiring knowledge. Rather than asking, “do I know the right prompting techniques?” ask, “am I using it every day? Am I experimenting with new tools? Am I iterating on my approach when something doesn't work?” This allows you to improve your skills by developing instincts about when to trust it, when to push back, and how to steer the conversation toward what you actually need.

I don’t know anyone who has tried, in earnest, to use ChatGPT or Claude or Gemini every single day for 30 days and hasn’t had their approach to work radically transformed by it. Thirty days is the magic number because it's long enough to push through the inevitable frustration phase: the first week when you can't think of what to use it for, the second week when your prompts feel clunky, the third week when you're still not sure if the outputs are actually good. By week four, something clicks. You stop thinking about “using AI” and start thinking with it. Not coincidentally, 30 days is (according to some people) also about how long it takes to build any new habit.

I made a guide to building a habit around AI, which is available for free here. It’s designed for people who don’t have an obvious set of use cases and aren’t particularly technical, though I’ve found that even engineers pick up useful prompting techniques from it. The whole point is to remove the friction of figuring out what to do with AI so you can focus on just doing it.

The guide includes 30 days of two-minute tasks that walk you through the prompting techniques I’ve found most effective: deliberate adjective use (“be 100x more specific,” “your suggestions are too generic, be more tactical”), providing examples of good outputs, asking AI to evaluate your work against specific criteria, adding emotional context to get better responses, and more.

Instead of trying to learn AI like it's a new language, treat it like a thinking partner. You don't learn to work with a thinking partner by studying them. You learn by working with them and gradually figuring out how to bring out each other’s best work.

xoxo,

hils