Life Unfiltered

AI agents are dominating headlines right now. But just because the word is hot doesn’t mean it’s clear what an agent is…or how to use one.

I recently joined Justin Jones and Kristi Fickert on the AI Xchange to break this down without the buzzwords. What even is an AI agent? How do you build a digital teammate? Where do you start? And what happens if you skip the important stuff?

Here’s a recap of what we shared.

What Even Is an AI Agent?

Ask five AI nerds what an agent is, and you’ll probably get five different answers. That’s fine. You don’t need a perfect definition. You need a useful one.

To me, an AI agent is a tool that can autonomously take action and complete a task on your behalf. Years ago, you would need to know how to code, and a computer science or engineering degree would be nice to have. Today, you can do it with an idea and some clear written (or verbal) instructions.

There are varying levels of complexity in creating AI agents, but in this post, let’s explore how to create one with a tool most people have heard of by now. . . ChatGPT.

Follow along, and by the time we are done, you’ll be ready to create a Custom GPT/Google Gem/Claude Project (choose your flavor) to complete a task on your to-do list today.

Two Core Types of Custom GPTs

At MAICON 2025 (the Marketing AI Conference), Andy Crestodina brilliantly pointed out that we can boil Custom GPTs down into two categories. As I’ve pondered all my use cases, I couldn’t find any that don’t fit into these two neat boxes:

Process GPTs: These follow a defined, consistent workflow like data cleanup, report generation, or marketing analysis.

Advisory GPTs: These act more like consultants. Think brand strategist, creative coach, personal stylist.

Know what you’re building before you begin. It changes everything from tone to task list.

Why I’m All-In on Custom GPTs

What excites me most about custom GPTs isn’t just the time savings (though that’s a big one). It’s the ability to scale your thinking. To take what lives in your brain, all your experience – wins, losses, and learnings along the way – plus your expertise, and turn it into something your team or your future self can actually use. 

You get to bottle your best processes, your instincts, your standards, and make them repeatable AND scalable across anyone you want to give your superpowers to. That kind of leverage is a game-changer, especially in a role where the to-do list never stops growing.

Where to Start

Start with something you do often. Something you know inside and out. Something you’d love to hand off.

That makes you the best possible trainer.

Think of it like onboarding an intern. You wouldn’t show them where their desk is on the first day and then simply say, “Compile our weekly client report.” You’d give them steps, examples, preferences, and context. Your GPT needs the same.

I’ll create a full separate post walking through every step, but here’s a brief overview on where to go to create your custom GPT.

• Go to explore in the ChatGPT dashboard.

• Choose create.

• Choose configure then follow the framework below to create your instructions. Be sure to add any documents that should be part of the GPT’s knowledge to the knowledge section. It will use that information every time you interact with the gpt. For example, brand guidelines or a relevant white paper.

The Framework I Use to Build Custom GPTs

I came across this structure somewhere along my AI journey, possibly through the RISE or Women Love AI Marketing communities. If you’re building your first GPT, save this:

Custom GPT Setup Framework

1. Introduction
   
   • Define the GPT’s purpose and audience.
     
2. Personality + Role
   
   • Name your GPT and clarify what role it plays.
     
3. Main Instructions
   
   • List what it should do, how it should do it, and what success looks like.
     
4. Interaction Guidelines
   
   • Decide tone, communication style, and how it should guide users.
     
5. Security + Confidentiality
   
   • Add guardrails to protect instructions and data.
     
6. Feedback + Iteration
   
   • Build in pause points and ask for user input to improve results.
     

This one structure has saved me countless hours of rework.

What To Build First

Here are some ideas for Custom GPT’s for work and daily life:

Life

• Weekly Schedule Strategist GPT: Helps you batch your tasks based on the type of thought process and effort required to optimize your time.

• Meal Planner GPT: Knows all the challenges your picky eaters bring to the table 😉 (see what I did there), and can overcome with dishes that delight the whole family.

• Fashion Consultant GPT: Worry no more about the dress that looked amazing on the model, but makes you look like a Midwest farmer’s wife from the 1800’s. Give it your body type and some inspo images, and voilà, you can level up your look.

• Financial Advisor GPT: Should you pay off the car early or throw more at the mortgage? What’s the smarter play long term… and how the heck do you calculate an amortization schedule without spiraling? Build this GPT and get your answer in seconds. And just maybe a few extra dollars for your “I worked hard today” sushi roll.

Work

• Audacious Marketing GPT: Ignite your creativity and color outside the lines with your next marketing campaign. I did an entire blog about this before. You can check it out AND try it here.
  
• Brand Bot GPT: Knows your brand, voice, style, guidelines, customers, and messaging. You can talk to it as a strategist or use it to create on-brand content and messaging.

• Resident Radar GPT: I walked through how to create this and the custom instructions on the recent AI Xchange. This can be for your customers or residents for multifamily folks. Provides a comprehensive analysis of your resident/customers. Who they are, what they like, and how to reach them.
  
• Weekly Performance Analysis GPT: Can be for any type of weekly report or analysis. I use mine for weekly marketing performance analysis.
  

The use cases are everywhere. Start with one. 

Lessons I’ve Learned Building Custom GPTs

• The more specific you are, the better the outputs
  
• You need to refine and iterate. This isn’t set-it-and-forget-it. Create, test, refine, and update the instructions, test again. Keep going until your digital coworker does exactly what you want.

• If you don’t want emojis or em dashes, say so. After YEARS of fighting through this, and ChatGPT still not quite following this directive consistently, this is finally addressed with ChatGPT version 5.1. (Cue celebration music)
  
• Tell your GPT exactly how you want the info structured and styled. You don’t want random words bolded throughout? You prefer a written narrative over bullet points? Only want the GPT to ask the user one question at a time instead of eight? Write it in the instructions.
  
• Pause points are your friend (and reduce hallucinations). Build in moments throughout the workflow where GPT will pause, review its work, and ask the user whether they are ready to continue or want to make adjustments.

• Only select the modes you need. If there is no reason you need your GPT to create images, don’t select that mode when you create the GPT. Keep it as clean and simple as possible.

Your GPT is only as good as your instructions.

It’s Go Time

If you’ve been thinking about building your first AI agent, just start.

It won’t be perfect. But it will be useful. And hands-on doing will teach you more in one hour than any AI webinar ever could.

A huge, enormous thanks to Justin and Kristi for sparking this conversation and continuing to make space for real talk in AI. If you’re in the multifamily space and you haven’t checked out The AI Xchange you are missing out!

Your turn: What’s one task you do every week that you could turn into a digital teammate? Drop it in the comments for inspiration.

Then go build it. And let me know how it goes.

reneemcintyre.com