I recently had a breakthrough by talking to a computer while walking my dog. Creating a family meal plan had been on my to-do list for about a year and it had been gnawing at me. Writing out what everyone, including my 1-year-old, would eat at each meal for the next week and then figuring out what to get at the grocery store — it’s a real chore. So, dog leash in hand, I asked an AI assistant to do it for me.

It took roughly three seconds, and I had the meal plan and the grocery list done. Some of the meals matched what we’d been doing, which is how I knew it was on point. Others were more creative than I expected, like giving our baby a bean salad. I somehow felt both productive and humbled. I also felt a little less burned out.

That probably sounds so obvious, if you’re someone who’s been optimizing life with ChatGPT since its launch two years ago. Yes, AI can take boring tasks you don’t want to do and complete them quickly, without complaint. But I’m guessing most people still aren’t quite sure how to use generative AI tools, so they aren’t using them.

Take my advice: Tell it to do something annoying that you’ve been meaning to do but haven’t had the time.

Maybe you need to write a thoughtful birthday card to an aunt you don’t know that well, or maybe you need to come up with a new workout plan. You might be surprised at how well AI does annoying chores like these, and how relieved you are to see AI doing basic helpful things rather than destroying humankind.

The tension between AI as a problem-solver and AI as a source of new problems is something you have to reconcile if you want to understand the technology, according to Ethan Mollick, an associate professor at Wharton and author of the newsletter One Useful Thing. Mollick has written a lot about the practical uses of AI, and he says that using an AI tool for about 10 hours is the best way to figure out how it will work for you.

“I feel like you have to move on from relief to amazement and existential dread to really get this product,” Mollick told me recently.

The thing is, I’d probably spent about 10 hours trying to figure out how to use AI before it finally clicked for me in my daily life. Most of those hours were at work, as I tried to get ChatGPT or Claude to explain complex topics and struggled to know what was accurate. It was ultimately getting AI to do one annoying chore — write up a complicated meal plan complete with grocery list — that brought me to that point of amazement. And I’m currently dealing with some dread as I try to figure out how to get AI to solve more problems.

AI models are getting more powerful and sophisticated every day. Many companies are burning through billions of dollars as they compete to build better and better models. OpenAI is on track to lose $5 billion this year, while Microsoft spent a staggering $19 billion in one quarter alone as it ramps up various AI-related projects. Mark Zuckerberg even said he was surprised by how much his company, Meta, was spending on the data centers and chips needed to keep up in the AI race.

All of these big numbers tell me that AI is only going to become more prevalent in our everyday lives. This push isn’t just about populating the web with smarter chatbots, and the flood of AI slop is hardly the biggest problem we’ll face as we shift paradigms. Researchers can’t even agree on what the biggest dangers of AI will be. At the very least, we should all start questioning reality more often, as AI is getting frightfully good at deception.

We’re already starting to see some of the next-generation AI products that tech companies have been throwing money at. OpenAI is doing a holiday-themed “12 days of OpenAI” series of new product announcements that include breakthroughs like a photorealistic video generator and ChatGPT integration on iOS and macOS devices using Apple Intelligence. Microsoft, meanwhile, announced Copilot Vision, an AI that can see what you’re doing on the web and talk to you about it. In a recent interview, Microsoft AI CEO Mustafa Suleyman told me that this is the first step toward a future in which AI can do more for us.

“It should be able to take actions on your behalf, fill in forms, buy things, book things, plan, navigate, click on drop-down menus, and so on,” Suleyman said. “All of that is coming down the road and in the foreseeable future.”

The tech industry’s big ambitions are a bit of a distraction for the average user. The mind-bending things AI could do isn’t actually impacting my life on a daily basis, but if I can get the technology to hammer out a few more annoying chores, I’d be thrilled. Mollick covers this concept a lot in his writing. I’d even argue that his latest newsletter addition bluntly titled “15 Times to use AI, and 5 Not to,” got me to ask that meal plan question.

It did take me about 10 hours of failing at using AI before I started to succeed and feel amazed. It’s easy for me to say now that it was time well spent because it certainly didn’t feel like it when I was tinkering with it and failing to find it useful. Then again, one reason why I put off annoying chores is I usually feel too burned out to do them. If an all-knowing computer-based intelligence is a solution to that burnout problem, I’ll take it.

A version of this story was also published in the Vox Technology newsletter. Sign up here so you don’t miss the next one!