15 Ways to Use AI at Home (and Fill Your AI Product Toolbox)

Audio Version ($)

AI overwhelm is real.

Whether you are a complete novice who isn't quite sure where to get started or deep into building an AI product, it's easy to feel like everyone else is light years ahead.

AI is a disruptive technology. People are adopting it at a record rate. The hype in the news and on all the socials is endless.

But here's the thing. You have more time than the hype suggests. So take a deep breath and let me be your guide.

I'll show you how to start with some very simple use cases—things everyone can do today. As we move through the list, each use case will get a little more complex. With each step, you'll add a new skill to your AI toolbox. And before you know it, you'll be seeking out your own use cases and designing your own workflows.

And for all of you being asked to deliver AI products and features yesterday, these same tools will help you when it comes time to build your first AI product. At least that's how it played out for me.

My Journey from AI Consumer to AI Builder

When I was new to ChatGPT, I started by asking it some of the same questions that I would ask Google. I liked that I got faster answers without the ads. But with time, my questions got more complex. And eventually, I started asking for help with day-to-day problem-solving and task execution.

Through my experimenting, I learned how to give the AI the right context to do a good job. I learned what worked and what didn’t. I learned how to create persistent memory. I learned how to conduct deep research.

And then a surprising thing happened. All this experimenting started to impact my roadmap. I started to get curious. How could I use this same technology to help me with my courses? And that was the seed that led to Product Talk’s Interview Coach.

You Can Follow My Path

And here’s the thing: You can follow this exact same path. You don’t have to know how to build AI products overnight. But you should start experimenting with how AI can support you day to day (if you haven’t already).

Today, we'll explore how you can do this in your personal life. This is often the lowest stakes place to start. Next week, we'll get into how to do this in your work life.

In both contexts, start small. Find a simple thing that is annoying to do. Ask ChatGPT or Claude or Gemini for help. Once you have a prompt that is working pretty well, try to automate it. If you are new to automation tools, check out Zapier or Make or n8n. Your company might already use one of these tools.

Along the way, you’ll make your day-to-day life a little easier, and you’ll start to learn some of the new skills that are going to be required to build AI products. This includes skills like prompt engineering—which is just a fancy term for giving the AI the right context. You’ll learn how to deconstruct complex tasks into smaller tasks that the AI can do well. As you push to automate more and more, you’ll learn about workflows (sequences of AI calls).

As you do more, you’ll gradually add new skills to your toolbox. When it comes time to build your first AI product, your toolbox will already be half full.

15 Ways to Use Large Language Models at Home (And Build Important Product Skills)

To help you get started, I'm sharing some of the use cases that I use LLMs for at home. They are organized by level of complexity. Start at the top. Once you have a handle on the first one, move to the next. Each use case will help you fill your toolbox.

My personal use cases are organized into three buckets:

1. Curiosity and Information Gathering
2. Everyday Life
3. Deep Research

This is the first article where I'll be reserving some content for our paid subscribers. Everyone will get access to the first two categories, but you'll need to subscribe to unlock the Deep Research section.

Let's dive in.

Curiosity and Information Gathering

Information gathering is where large language models really shine. They've been trained on large portions of the internet as well as thousands of books and other resources.

1. A Better Search Engine

I very rarely Google things anymore. Instead, I ask ChatGPT. I get a faster answer. I don't get inundated with ads. I started and still use ChatGPT for simple mundane queries like:

• Can my dog eat this?
• Can I slow peaches from ripening if I put them in the fridge?
• Does oatmeal go bad?
• Can my dog be off-leash at Todd Lake?
• What's a good coleslaw recipe that isn't sweet or too mayonnaise-y?

You get the idea. This is probably the most common use case for ChatGPT. If you are brand-new to AI, this is the place to start.

Skills you'll build:

• You'll get comfortable with chatting with LLMs.
• It's an easy way to overcome the "What do I use this for?" conundrum.

2. More Complex Search Queries

The previous list of queries can all be handled by Google. So why do we need LLMs? The more you talk to LLMs, the more you'll realize that they are capable of far more than simple queries.

I was recently having a conversation where we wondered how many US Senators were over 75. Try Googling that. Actually, I just did. I got a few results that listed the age of every US Senator. Not bad. But I'd still have to scroll through the list of 100 US Senators and count how many were over 75.

Instead, I asked ChatGPT and got a fast answer. There are 10 US Senators over the age of 75. I also got a list of all ten. ChatGPT cited its source—in this case Axios—and even provided a second way to get at the same data.

This was more than good enough for my purpose.

Skills you'll build:

• Start to understand what LLMs can do better than search engines.

3. Learn About Current Events

When Hamas attacked Israel on October 7, 2023, I had a lot of questions about the Middle East. Some were embarrassing. Some were questions that I thought I should already know the answers to. But I didn't.

So I turned to ChatGPT. I had a long conversation where I asked everything and anything under the sun. I learned about the history of the region. I had side tangents about the etymology of the word anti-Semitism. I learned about Hamas and Hezbollah. I even wandered into the history of Jordan.

This was one of my first truly mind-blowing experiences with a large language model. It felt so empowering to be able to ask anything I could think of and get an immediate answer.

It was also the first time that I started to truly worry about bias and hallucinations. How did I know that what ChatGPT was telling me was true? OpenAI is an American company. Were the responses influenced by American politics? How could I know?

And so I started to ask for sources. I spent some time on Wikipedia. I explored other sources that I trusted.

Even though I had to rely on additional sources to trust what I was getting, I learned that ChatGPT was a great starting point. It helped me get past the overwhelm of not knowing. It helped me frame my questions so that I could seek out better sources.

I now use large language models to explore current events on a weekly basis. I strongly recommend trying it out.

Skills you'll build:

• Learn how to use large language models to explore new topics.
• Start to understand the risk of bias and hallucinations.

4. Interpret Medical Results

There's a lot of information asymmetry in medicine. Doctors have a lot of medical expertise. Patients have little. Combine this with the fact that most doctors only have a few minutes to spend with each patient and you get an environment rife with miscommunication.

I'm a big advocate for being an active decision-maker in my own health. That means I need to make the most out of the limited time I get with my health care providers. I like to use large language models to help me prepare.

When I had an ankle break that needed surgery, I made sure I read the surgery notes before my follow-up appointment. That's when I discovered that one of my ligament repairs was a "secondary" repair. I didn't know what that meant. So I pasted the whole report into ChatGPT and asked it to explain the notes to me.

It turns out a secondary repair indicates the repair of an old ligament tear. In other words, it wasn't the result of my current injury. Interesting.

I read about the different types of surgical repairs for that particular injury and the pros and cons of each. You can bet I had much better questions for my surgeon when I finally saw him.

Whenever I get blood work done (think a lipid panel or a metabolic panel), I ask ChatGPT to walk me through any scores that are outside the normal range. I explore the different implications and potential causes.

I recently tested high for bilirubin. Bilirubin is a waste product produced by your liver as it breaks down red blood cells. This was the second time it's been high in the past few years. But all of my other numbers indicated a healthy liver.

Both ChatGPT and my doctor explained that I likely have Gilbert's Syndrome—an innocuous genetic variant that leads to high bilirubin. It explains why I bruise easily. But most importantly, it's nothing to worry about. I was glad that I could explore this on my own time to the depth that I wanted before talking to my doctor about it.

I don't recommend using large language models in place of going to see a qualified medical practitioner. But I do like to use them to prepare for my medical appointments. It helps me be a better advocate for my own health. If this is important to you as well, this is a great use case to try.

Skills you'll build:

• Using large language models to learn about a new topic.
• How to be mindful about bias, hallucinations, and irrelevant information.

5. Scratch Your Curiosity Itch

Once I got comfortable using LLMs to explore current events, I started to think of ChatGPT as a curiosity engine. I turned to it whenever I wondered about something.

My husband has long pined after a tractor. He wants to build a mountain biking trials course in our yard and he needs a tractor to move the rocks around. (This is never going to happen, but it's fun to dream.)

One day, on a long drive, we drove past a Caterpillar or John Deere yard full of tractors, and my husband said, "I wonder how big of a tractor I would need."

I responded, "I suppose it depends on how big the rocks are." My husband guessed, "Maybe 4' x 2' x 2'."

And so I typed into ChatGPT, "How big of a tractor would I need to move a 4' x 2' x 2' rock?" ChatGPT immediately replied, "It depends on the type of rock." I didn't know what type of rocks we had here, so I asked, "I live in Central Oregon. What type of rocks do we have here?"

ChatGPT then started to spit out the following: "The most common rocks in Central Oregon are basalt. Basalt has a density of... The weight of a 4' x 2' x 2' basalt rock is... Therefore you would need a tractor that can lift x pounds. Here are some tractors that meet your specifications..."

Okay, this was awesome. Are we going to buy a tractor? No. Was this a really fun way to scratch a curiosity itch? Absolutely.

Think about how many times you wonder about something that is more complex than a quick Google search. Anytime you have a curiosity itch that requires a little bit of reasoning mixed with some information gathering, this is a good time to turn to a large language model.

Skills you'll build:

• Learn how large language models problem solve.

Everyday Life

LLMs aren't just good for information gathering. I rely on them to help me solve problems in everyday life on a regular basis.

6. Fixing Cooking Disasters

One night I was cooking dinner. I had a pot of rice on the stove and the timer had just gone off. I looked at the pot and the rice was still swimming in water.

I've made rice hundreds of times. But this time something clearly went wrong. I thought about it for a minute and realized I got the rice to water ratio wrong. I made 1/2 cup of rice, but I used enough water for 1 cup of rice. Now what?

ChatGPT gave me three different ways I could salvage the rice. I tried the first one. It mostly worked. The rice was edible and I didn't ruin dinner.

This is a really simple use case, but I share it to highlight how often I turn to ChatGPT to help me get out of jams. It really is my all-around go-to problem solver.

I've used it to suggest viable substitutes when I'm in the middle of cooking something and realize I'm missing a key ingredient. I use it for recipes. I use it to suggest what to make for dinner.

These tools can do a lot. I recommend throwing anything and everything at it.

Skills you'll build:

• Learning what LLMs are good at and where they stumble.
• Build the habit of using these tools for everything.

7. Meal Planning

Keeping on the same theme, I also use ChatGPT to help me meal plan. I've tried several approaches:

• Sometimes I start with what's in the fridge and I ask, "What can I make with this stuff?"
• Other times, I give it dietary preferences and ask it to plan a week's worth of meals for me.
• Most often, I just ask for creative ideas. My husband and I get tired of our usual rotation and this is an easy way to get out of a rut.

This is a fun use case. If you don't like the suggestions that you get back, consider what context would help ChatGPT do a better job. Does it need to know your allergies, likes and dislikes, or what you have on hand? If its suggestions are too repetitive, do you need to tell it what you ate recently? Do you have a diet that you adhere to that it needs to know about?

This is a use case where context is everything and that's why it's a great one to experiment with.

Skills you'll build:

• How to give an LLM the right context to give you good suggestions.

8. Movie Recommendations

Second only to deciding what to eat for dinner, the hardest daily decision in my household is what to watch on TV.

Until we realized that LLMs can help. We started with a ChatGPT conversation where I described what we liked and disliked. I even gave it some examples for each.

It recommended a short list of movies with a synopsis of each. I then followed up with some clarifying questions. We selected a movie, watched it, and were pleasantly surprised when we enjoyed it.

For months, we used that same conversation to tell ChatGPT what we liked and disliked. It continued to suggest good things to watch. Until it didn't.

At some point, it started to suggest things it had suggested in the past that I had already indicated we didn't like. It started losing track of our preferences. This was my first introduction to an LLM hitting its context window limit.

At this point, we moved to a Claude Project where we shared a few documents with the project:

• a list of our preferences
• a list of movies we liked
• a list of movies that we didn't like

This dramatically improved our movie recommendations and is what we still use today to decide what to watch. It's not perfect. But the hit rate is much higher than the miss rate.

You can follow this same strategy for movies, TV shows, music, books, or anything else where you want some strong recommendations.

Remember to update your preferences over time so that the tool keeps getting better.

Skills you'll learn:

• Understand context window limits.
• Learn how examples influence quality (often referred to as few-shot or n-shot examples).
• Start using persistent state/memory (in my case, Projects).
• Learn how to refine over time.

9. Shopping Guide

There have been a few times where I've outright outsourced my shopping decision to an LLM. Other times I use it to help me make a more informed decision about what to buy.

I'll share two quick stories.

First, I needed a new web camera. The one I was using had significant auto-focus issues. I know this is the type of decision that I could nerd out on forever, but I also knew I just needed something that was good enough. So I outsourced this decision to ChatGPT entirely.

I told it I need a webcam that doesn't have auto-focus issues. I explained how I would use the camera (e.g. for calls, webinars, talks, recording videos) and explained that my primary criteria is picture quality.

It suggested three options. I asked some follow-up questions and then I just picked one. This entire shopping scenario took less than ten minutes. And my new camera is great. I got this one (that's an Amazon affiliate link).

Next, my husband and I rescued a border collie/pit bull mix. She's a picky eater and we want to make her happy.

We started to research better quality dog food and quickly got overwhelmed. There's better kibble. But there's also fresh food. There's grain-free food. And there's many permutations.

I asked ChatGPT to help me define some good criteria. We learned about different vet ratings that represented everything from nutritionally balanced to sustainably sourced—both things we care about.

We then asked for a comparison grid of the best options. I believe this was the first ChatGPT Deep Research report that I ever ran. After a few minutes, I got back one detailed grid comparing the best kibble and second detailed grid comparing the best fresh food. It turned a very overwhelming decision into a doable task.

I strongly recommend experimenting with using LLMs as your shopping guide. You can decide how much autonomy to give the LLM. You can use it to just pick something for you or you can use it to gather information to guide your choices. It's up to you. They each add value.

Skills you'll learn:

• How you want to partner with LLMs in different situations.

10. Travel Planner

In the Spring of 2023, I traveled to Germany for the inaugural Product at Heart conference. My husband and I decided to make a trip of it and we planned to spend three weeks exploring the best Germany had to offer.

We had a few things on our short list. We wanted to:

• bike through wine country
• spend some time on Lake Constance
• visit friends in Munich
• and of course, spend some time in Hamburg for the conference

But, other than that, our itinerary was wide open. I spent a few weeks researching and we added some new possibilities to our list: the Black Forest, exploring Berlin, visiting the castle that inspired the Disney castle, etc.

And then it occurred to me that I could ask ChatGPT. I was a little embarrassed when ChatGPT quickly created a list of everything I had spent days (maybe even weeks) uncovering.

This was an easy trip to plan for. There is a lot to see in Germany and we had never been. But I've since learned to use ChatGPT for more complex travel planning.

We recently went to San Diego. We needed to be near Del Mar and San Marcos where we both had family to visit. I wanted to be at a high-end resort on the beach and my husband wanted active surf waves.

I spent an entire afternoon looking at way too many dated, worn-out hotels in Pacific Beach (not the vibe we were looking for), when we were ready to give up. But ChatGPT came to the rescue. It suggested the Alila Marea Beach Resort in Encinitas. The location was perfect. The resort was ridiculously high end. Surfline said the beach was a good surf beach. And best of all, we could book with Chase points.

This was a weekend getaway where we had very specific needs. It was hard to find a place that met all of our criteria. But ChatGPT nailed it.

I strongly recommend asking an LLM to help you with travel planning. However, be warned, if you don't give much context, you'll get generic suggestions. So this is a great use case for practicing how to give the LLM the right context to do a good job.

Skills you'll learn:

• Defining the right context.
• Having the LLM interview you to help you surface your implicit preferences.

11. Research Service Providers

This might surprise you, but I'm really bad at mundane chores. I'll procrastinate for weeks if I need to find a service provider for anything. I hate talking to people on the phone. I don't know how to choose the right people. It's just all-around unpleasant.

We are in the process of selling our Portland townhouse. To get it ready to put on the market, I needed to find movers and someone to stretch and re-tack the carpet. And I was on a tight timeline.

So naturally, I tried to outsource as much as I could to ChatGPT.

I started with the carpet stretching. Who do I even call? Carpet installers? I had no idea. I just told ChatGPT what I needed. I told it I wanted service providers who had good reviews, could communicate well, and would show up when they were supposed to. ChatGPT gave me a short list.

Then it asked if I wanted it to draft an email that I could send. Yes, please do that. It drafted an email that included questions I would have never thought to include (e.g. "Do you use a power stretcher?" "Do you guarantee your work?") and it gave the contact info for each company on the short list. All I had to do was send the emails.

Next I moved on to movers. This was a little harder. I needed movers who could handle a long-distance move (a three-hour drive over a mountain pass) and could move a hot tub. This turned out to be a tough combination. Again, I asked ChatGPT for a short list of well-reviewed vendors. But this time I asked for an email draft and contact info as well. I emailed several and struck out repeatedly. None of them could meet both of our requirements.

But I didn't give up. I went back to ChatGPT (same conversation thread) and I told it my problem. It then refined its search and found a handful of service providers that specifically mentioned that they could move heavy items. This time it worked. I found two moving companies, got quotes, and scheduled my move.

I love this use case because it's like I have a personal coach for a task that I absolutely hate. And more than that, ChatGPT did the bulk of the work. It even suggested doing more of the work (drafting the email, finding the contact info) than I would have known to ask for. I'll definitely be doing this much more often. I recommend you give it a try.

Skills you'll learn:

• To not give up. If an LLM strikes out, tell it what went wrong and ask it to try again.
• How to let LLMs do the complex searching for you.

Deep Research

This is a category where I wouldn't know how to get the job done without the help of an LLM. I'll cover how I used LLMs to help me:

• be a more informed voter—including how I used an LLM to help me build a detailed model of my school district's expenses so I could better evaluate a bond measure
• file both an S-corp return and a fairly complex personal tax return and why I chose this route instead of continuing to work with my tax accountant
• evaluate PEX vs. copper for a plumbing repipe when we have two well-respected plumbers arguing for each side and arguing the other choice would be a big mistake
• price an empty lot next door and evaluate if it was a good purchase for us (this work was recently validated when the lot went on the market at the high end of ChatGPT's range)

As a reminder, these next few use cases are for Supporting Members and CDH Members. At the end of the article, we'll also use the comments section to discuss your use cases that you are putting into practice.