I've travelled to 44 different countries so far. Needless to say, I know a thing or two about booking a trip.

Before AI, my routine looked like this: searching Google Flights, setting price alerts, checking hotel sites and Airbnb for accommodations, browsing a few newsletters for deals or inspiration. In my backpacking days, it meant buying a copy of the Lonely Planet just to know how to avoid tourist traps. There were a lot of moving parts involved, like having 15 browser tabs open at the same time, notes scattered everywhere, and somehow still second-guessing the itinerary.

Does that sound like you?

That was the old way. Today, an AI agent can run all of those tasks at the same time, without you managing each one. You describe the trip. It figures out the rest.

That's what we're getting into this week.

Everyone has used a calculator (I hope, I think?). You type in your input, the calculator processes it and displays a result. Funnily enough, that's how most people are using AI right now, like a calculator. Ask one question and wait for an answer to spit out.

Meanwhile, let's say you have an assistant working with you. That interaction would be different from how you'd use a calculator, right? You would tell them what you're trying to accomplish, and they would figure out the details by asking questions or talking to the right people. Then they'd come back to you when the job was done.

That's essentially what an AI agent does.  It figures out what it needs, goes and gets it, and keeps going until the job is done. The reason it can do this is that it's connected to tools:  a search engine, a database, and another AI model. That connection is what lets it go find things instead of just knowing things. Without it, you're just using a calculator. Web search is what turns Claude from a calculator into an agent. One click in settings, and you’ve unlocked it.

Say you open Claude, web search is on, and you ask it to find the best week next year for a surfing trip in Greece. You've turned Claude into an agent. The agent doesn't know weather patterns off the top of its head, so it goes and pulls years of daily weather data for Greece. But that's still not enough. It doesn't know what makes good surfing conditions. So it goes and consults a second agent who specializes in surfing. Learns that high tides plus sunny weather plus low rain is the sweet spot. Combines everything and then finds the best week for you!

Just like if you had asked a travel agent. They might have years of experience, but they may never have booked a surfing trip to Greece. So they call someone who has. They gather the intel and come back with exactly what you need. All of that happened in the background to answer one simple question: "When's the best time to go surfing in Greece?"

That's an AI agent. A system that figures out what it needs to know, goes and gets it, and comes back with an answer you can actually use.

I simplified the Greece example to make it digestible. If you want to see exactly what's happening under the hood, IBM breaks it down here: ibm.com/think/topics/ai-agents

The good news? You don't need to be technical to use one. Try It Tonight at the bottom of this issue walks you through setting up your first agent in under five minutes

Imagine this: you have a meeting tomorrow, it’s with a new potential client, or maybe it’s an interview for a dream job.

You know the name of the person you’re meeting with and the company, so what would you do next to prepare yourself? You’ll probably go on LinkedIn or a quick Google search to find any info about the person or company, or you’ll ask your AI tool to do the same thing. This isn’t a bad approach, but let me show you an even better way to prepare yourself and a setup that you can reuse for future situations.

Create a new Claude Project (if you don’t remember how, here’s the Companion page from Issue # 2). Call it “Meeting Prep”  and paste this into the project instructions:

Then every time you have a meeting, you type one line:

What's happening in the background is exactly what we unpacked above. The agent is not answering one question. It runs multiple searches, pulls from different sources, and connects dots you didn't ask it to connect. That's the difference between a prompt and an agent.

Remember, you need web search toggled on in Claude before this works. It's one click under the chat settings, but easy to miss.

AI agents are getting dramatically better, fast A year ago, AI agents failed at real-world tasks 4 out of 5 times. Today, they succeed 3 out of 4. That jump, from 20% to 77%, happened in a single year according to Stanford's AI Index. Stanford HAI

Your next shopping agent is already being built AI shopping agents can now understand your preferences and budget, compare prices across thousands of retailers, and complete purchases on your behalf. You say, "Find me the best wireless headphones under $200" and it handles research, comparison, and checkout. Amazon, Google, and a wave of startups are racing to make their agent the default way people shop. The question isn't if this becomes normal. It's whose agent gets there first. Rep AIModern Retail

6.7 million people tore apart a Monet. It was a real Monet. An anonymous artist posted a real Monet painting on X, told people it was AI-generated, and watched the pile-on begin. People called it "busy, artificial, nature in turmoil." One person wrote a 700-word breakdown of why it was obviously fake. It wasn't. The lesson: we're so primed to distrust AI that the word alone changes what we see. Worth thinking about. FortuneOfficeChai

Something I bring up with friends more than they probably want to hear: AI is powerful, and it is also hungry, power hungry, figuratively and literally.

Data centers already consume nearly 2% of global electricity, and that number is growing at 12% per year. Training a model like GPT-4 used enough energy to power roughly 20,000 US homes for a full year. That's real, and I think it's worth saying out loud, even in a newsletter that exists to help you use more AI. All About AIWikipedia

My take is not to stop using it. My take is to use it on purpose. The meeting prep agent we just walked through saves you 20 minutes and makes you sharper in the room. I think that’s worth it. Creating a funny video for your friends. Maybe not, maybe you use something else to do that. 

Your $20/month isn't the only thing paying for this. The planet is picking up part of your AI tab. Act accordingly. 

Cinz.AI

Your unfair advantage, one week at a time.

Got a question, a topic request, or just want to say hi? Just reply to this email. I read everyone.