Transcription
Open AI just released Chat GPT work, and this has been their biggest update of the year because now Chat GPT has access to agentic powers, which means [music] it can autonomously create spreadsheets, documents, websites, connect with your applications, and a lot more. So, today I'll walk you through what it is, how it works, the different features, and the use cases. Stick around because by the end there's a feature that does everything on its own without you even touching anything.
On the 9th of July, Open AI released their GPT 5.6 model family with a new work feature. And you might be wondering, Chat GPT was already working fine for most people, so why did they need to change anything? Well, it turns out that fine wasn't enough anymore because for about 3 years using Chat GPT has looked the same. You ask it something, it gives you an answer, and then it's back on you. You're the one doing the actual work. So, this is where they decided to fix it by introducing GPT work. And Open AI has made it official that this is rolling out gradually. So, if what I show you on screen doesn't exactly match what you have, don't worry. It's just a staggered roll out, and it'll look the same for you very soon.
On the web version, it's very simple. You've got two tabs right at the top, and you can toggle between them with a single click. But on the desktop app, it's a slightly different space. On the left, you have a drop-down option to switch between work and code X. And if you want a regular chat, just click on the chat option, and you'll get a separate window to start a new conversation. I'll close that for now because GPT work is what we're using today. On the left sidebar, you've got multiple options, something in the center, and a few options alongside the prompt box as well. Don't worry about all of this right now. You'll understand it naturally as we go through the video.
So, let's actually give it something to do and see how this thing works. I'm going to ask it to build me a monthly budget spreadsheet for a $4,000 income with categories, savings goals, and the formulas already set up for easy calculation. Let's hit send. Watch this. While it's working, it actually shows you the whole process in real time. It starts by building a polished budget workbook with an editable income and expenses tab and an automatic totals tab. Then it loads a tool, edits the file, and runs actual commands. Then it verifies its own work by doing a final calculation and layout check before saving the file. After that, it views images of what it built, checks and edits multiple times until it gets to a final result. And this is what we got. Honestly, I didn't expect it to look this clean. It built three separate tabs and split everything the way an actual finance person would. In the monthly budget tab, there's a $4,000 income at the top. Then it broke expenses into categories like housing, utilities, and groceries, all filled with actual numbers. And at the bottom, it's calculating total expenses, savings contribution, and even your savings rate.
Now, if you're thinking, "How do I actually use this thing?" you need to go to the start here tab, and it literally has its own instructions on how to use the workbook. It says, "Set your income, plan your spending, track your actuals, and fund your goals." Then there's a color guide. Yellow cells are the ones you can edit. White ones are formulas. So, you just need to fill in the yellow cells. Let's go back and test it by filling in what we actually spent this month. Say on housing, we spent exactly what we planned, but we went over on groceries and utilities. So, let's put in 520 and 350 instead of the planned 450 and 250. I'll fill in the rest with some spending under and some over budget. Now, watch what happens in the summary while I enter these numbers. Planned expenses were $3,050, but actual came out to 3,160. And see that red number next to it? It shows we spent $110 more than expected. And look what that does to your remaining cash. The planned amount said you'd have $400 left at the end of the month, but you only have 290. That same $110 gap showing up on the other end. Now, you see that $550 savings contribution? Let's see where that number is actually coming from. Go to the savings tab. There are different goals, and each has its own target and monthly contribution. 300 going to an emergency fund, 150 to vacation, 100 to gifts. Add all that up and it's the same 550 we saw in the summary tab. So, these numbers aren't random. They're all connected and all of this happened from just one prompt.
And if you're thinking that's not a big deal, let me show you something. I typed the exact same prompt in a regular chat on the free plan and look at what it gave me. Just a list of budget categories typed out as bullet points and a flat spreadsheet with four columns. No savings tab, no instructions tab, no color coding, nothing. So, the real difference is right in front of you. Regular chat gives you a rough draft and you still have to build it yourself. ChatGPT work gave you something completely different with the exact same prompt. And again, this was done without touching any of the options on the left. Now, imagine what it can actually do once you use it to its full potential.
From here, I'll switch to the web version because that's where all features are already rolled out. Let's look at projects first. If you're using projects in the desktop app, it actually creates a real folder on your local computer. On web, it stores everything in the cloud. Let's make one. I'll click on projects, then create new. It's asking me to name the project, so let's say I have a lawn care and landscaping business. I'll name it Greenleaf Landscape and hit create. Now, the option I want you to focus on is sources because if I just start typing random tasks in this project, ChatGPT has zero clue what Greenleaf actually is. So, I'm going to add some sources so it has context about my business. If I click add sources, there are multiple options. Upload, add from library, text input, Google Drive, and Slack. I'll go with upload and give it my business profile, which has all the details this project needs to know.
And here's a bonus for you. If you don't know how to create a business profile, I've built a simple framework called the BASIS framework. B stands for business basics, your name, what you do, where you're located, and the size of your business. A stands for audience, the details of the people or clients you serve. S stands for sales, what you sell, your top offerings, and their prices. E stands for expression, your brand design, your voice, and your style. And the final S stands for setting goals, what exactly you're building or fixing in your business. Use this framework and you'll have a solid business profile to use as a source for your project.
Let's go to the chat and ask it to create a one-page service brochure for the upcoming fall season. Now, before sending, we have the option to select a model from the newly added 5.6 model family, Soul, Terra, and Luna. Soul is the highest level, consumes more of your usage limits, and is built for complex tasks. Terra is for daily medium-level work, and Luna is for regular fast tasks. To show you the difference, I'll run this prompt with Soul first, then run the exact same prompt on Luna. Let me put them side by side, Luna on the left, Soul on the right, exact same prompt. First thing to notice once they're both done, the time. Luna finished in flat 4 minutes. Soul took 6 and 1/2. That's already a signal that Soul did something extra. In Luna's version, the fall cleanup package is priced at 199, holiday lighting starting at 250, and it kept the green leaf tone and mentioned the service area. That's a solid job. It did everything I asked for. Now, look at Soul's version. Same fall cleanup package, priced at 195 with a note explaining that it saves $20 compared to booking separately. It kept holiday lighting at 250, same as Luna. But then it added something I didn't even ask for, a cross-sell mention of weekly mowing at 45 a visit. Because why not upsell when you've already got their attention? And this detail at the bottom actually got me. Look at this line. It says the profile didn't include a phone number or email, so the call to action says call or message Greenleaf Landscaping instead of inventing contact details. Soul caught a gap in our business profile and worked around it, while also delivering two different file formats.
Now, let's look at what they have actually built. I mean, do I even need to explain this? It's clearly a no-brainer. In these examples, Luna is good for first draft, but Soul gave you something you can directly hand to a client. So, the real takeaway is that Luna is fast and honestly good enough for a lot of quick stuff, but Soul actually thinks like it has a stake in the outcome. It caught a mistake, added value nobody asked for, and gave us a more usable output.
All right. We've seen how the model changes the quality of what comes back, but there's another confusing piece to this, and that's plugins. I got confused by this myself because plugins don't work the same everywhere. When you hear plugins, you naturally assume they connect apps for real-time data. But in the desktop app version, plugins are more like artifacts. They create and edit things like presentations, documents, and PDFs. For example, if I use the spreadsheet plugin, it shows templates to create a project tracker or a budget planner. So, where's the live data feature? Turns out it's in the web version. There, you'll find plugins across categories like productivity, creativity, communication, security, finance, and a lot more. And if you click expand, there's a list of hundreds of plugins across each category.
Let me show you how this works through a real scenario every service business deals with. I've got a list of clients in my Google Drive who've already had their service done, but some of them still haven't paid. So, I want ChatGPT to go fetch that list, figure out who hasn't paid, and draft a reminder email for each one of them. But before I can do any of that, I need to set it up with the right plugins. I'll go to plugin, search for Google Drive, click install plugin, then continue to Google Drive, provide the necessary permissions, and done. My Drive is now connected to ChatGPT. I'll follow the same process for Gmail. Once both are installed, I'll go back to my project and type, "Check my client tracker sheet in." Now, pay attention here. To access a plugin while writing a prompt, type the at symbol and the full list shows up. Just select Google Drive, find everyone marked as unpaid, and draft a short polite payment reminder email for each one and save it as a draft in Gmail. I'll select Saul for this one and send. It says it'll use the Drive and Gmail workflows as requested. Then it runs into a name issue. The exact title search didn't return a spreadsheet. So it changes approach, broadens the Drive lookup to title keywords and recent files and starts searching for spreadsheets containing the word client. It finds the tracker, identifies nine unpaid accounts and after about 2 minutes it creates and verifies nine payment reminder drafts in Gmail without actually sending anything. Total outstanding amount around $1,200. At the bottom there's an option to open the Gmail drafts. Let's click it. And here [music] we go. Nine drafts ready to send. The content inside is relevant, too. That's the power of plugins. And this was just one example. Plugins can handle almost anything you deal with day-to-day. Pulling a client list and sending out appointment reminders, grabbing your latest invoices and turning them into a monthly expense summary, pulling product or inventory data and drafting restock orders automatically. Whatever you're doing with your files and applications, there's a good chance a plugin like this can just handle it for you.
Now here's one thing about plugins. They're absolutely insane at pulling information from different places, but they still don't know how you actually do things. What if you could just teach it once and it remembers forever? That's exactly what skills are built for. Skills are a set of instructions that guide ChatGPT on how to perform a specific task. So let's say this is a business that deals with thousands of clients every week. Manually re-explaining what to do every single time is not practical. So instead I'll create a skill for it. I'll go to plugins, click on skills, then click create with chat and explain what I want it to do. I'll type the task and send. It asks whether each reminder should include the outstanding amount and due date. I'll allow both. After about 2 minutes the skill is created and installed. It's named draft client payment reminders and saved as a .md file. To use this skill, I'll open a new chat, invoke it by typing @ and the name of the skill, and hit send. After about a minute and a half, it creates and verifies nine Gmail drafts, this time with the outstanding amounts included as well. So, every time you need this task done, just invoke the skill and you are done.
But, here's the catch. I'm still the one opening the chat and manually triggering it every single time. If this is supposed to be an agent, shouldn't it just go and do this on its own without me asking? Well, it can. And that's exactly what the scheduled feature is built for. You set it once to run daily, weekly, or on a specific date. And from that point, it just happens in the background. To set this up, go to the schedule tab and type what you want. I'll invoke my skill and ask it to run every Friday at 5:20 p.m. Hit send. After about 30 seconds, the client payment reminder is scheduled for every Friday at 5:20 p.m. That's it.
All right, here's a quick recap so you don't forget anything. Chat GPT work is an autonomous AI agent powered by the GPT 5.6 model family, which includes Sol, Terra, and Luna. We clearly saw that Sol thinks harder and delivers more polished, complete work, while Luna is faster for quick tasks. Then, we built a real project around a lawn care business, gave it an actual business profile built using the Basis framework that covered pretty much everything. Then, we saw how plugins work differently depending on where you are, templates on the desktop app, and live connectors on the web. Then, we pulled live data from Drive and drafted weekly emails through the Gmail plugin. After that, we built a custom skill to stop re-explaining things every time. And then, we tied the whole thing together with the scheduled feature. So now, a quick question for you. At the beginning, I told you that by the end, there's a feature that does everything on its own without you touching anything. Have you figured out what that feature is? Drop your answer in the comments.
If you're running any kind of business, or you're just drowning in repetitive tasks every week, this is worth setting up. It may not be perfect right now since it's still rolling out unevenly, but the core idea is real and genuinely useful. Give it one small task and see what comes back. And if this video actually helped you understand something new, hit that like button, drop a comment, and subscribe to the channel for more breakdowns like this. Thanks for watching and I'll see you in the next one.