YT YouTube2Text

YouTube2Text → Transcripts → How to Use ChatGPT Work Better Than 99% of People (Full Guide)

How to Use ChatGPT Work Better Than 99% of People (Full Guide) — Transcript

by 9x · 10,390 words · 1,429 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Chat GPT work might be the biggest shift
0:01in how you get work done since Chat GPT
0:04itself. It is an agent that runs on your
0:06computer and does a whole job for you.
0:08While regular Chat GPT gives you an
0:10answer and then you need to go and do
0:11the work, this one gives you the actual
0:13finished job. But the reality is most of
0:15the billion people using Chat GPT every
0:18month are never going to make the most
0:19of it. So in this video, I'll show you
0:21exactly how to set it up, how it's
0:22different from regular chat and Codex,
0:25and the eight things that take you from
0:26a complete beginner to a legit power
0:28user. I've been using this pretty
0:30heavily for the past month and I've
0:31already taught this live to over 250
0:33professionals. So I've seen firsthand
0:35exactly where people get stuck. Now
0:37before jumping into the tool, what
0:39actually is Chat GPT work and how does
0:41it compare to regular chat and Codex? So
0:44let's start with the one that you're
0:45probably most familiar with and that's
0:46regular Chat GPT and that's what we'd
0:48classify as an AI chat. Its core role is
0:52really being your thinking partner. It's
0:53great when you have questions, you need
0:55an explanation on something,
0:56brainstorming new ideas, or when you're
0:58settling an argument with about a random
1:01topic like whether there's more sheep or
1:02kangaroos in Australia. You provide a
1:04prompt and get an answer back, but then
1:06you still need to in most cases go and
1:08do the work or use that information. Now
1:10Chat GPT work and Codex are drastically
1:13different and that's because they are
1:14both AI agents. When you work with an AI
1:17agent instead of an AI chat, you really
1:19do need to change the way you prompt.
1:21Instead of just asking a question and
1:22getting an answer back, you delegate a
1:24task to be done. The agent then takes
1:26that task and makes a plan for itself
1:28and then works through any required
1:30steps until the task is completed or it
1:32meets the goal. Now what's interesting
1:34here is that under the hood both Chat
1:36GPT work and Codex are running the same
1:38agent. If you see here, even when I ask
1:40Chat GPT work about what system it's
1:42running, you can see that both are
1:43actually running the same Codex agent
1:46just living in two different product
1:47interfaces focused on two different
1:50areas. Chat GPT work, unsurprisingly, is
1:52focused on knowledge work. We call it
1:55your work partner. Basically any
1:56multi-step work that needs to be done
1:58across tools and involve things like
2:00pulling in data for in-depth analysis or
2:02updating systems in your CRM or task
2:04management tool or creating spreadsheets
2:06and documents. While CodeX on the other
2:08hand is focused on writing code. It's
2:11really designed to be used by developers
2:13who are building software. So, if you're
2:14someone working in any non-technical
2:16business function, whether that's
2:17finance, legal, HR, marketing, sales,
2:20whatever, ChatGPT work is the one you
2:22should be reaching for to get actual
2:24work done. All right. So, now that you
2:25know what ChatGPT work is, let's get
2:27into how to actually use it. And this is
2:29what we'll be covering in the video,
2:31these eight steps that I think are
2:33necessary to take you from a beginner to
2:35an actually confident user with this
2:37tool. First up, we will be getting into
2:39the setup, how to actually set up
2:41ChatGPT work. And this is important
2:43because there's actually multiple ways
2:44where you can access ChatGPT work, and
2:47they are not all the same. And this can
2:48definitely trip up some beginner users.
2:51Then we'll move into projects. And
2:53again, projects here in ChatGPT work are
2:55drastically different than the projects
2:57you're used to in regular ChatGPT. And
2:59here we'll really focus on how to work
3:01with the files on your computer to
3:03unlock some workflows that you just
3:04wouldn't be able to do with regular
3:06ChatGPT. Then we get into the agents MD,
3:09so you can stop actually repeating
3:11yourself and save instructions that
3:13ChatGPT should always follow whenever
3:15you're working in ChatGPT work. Then
3:17we'll get into plugins, apps, and
3:19skills. And this is where we'll probably
3:20spend the most time because I think
3:22understanding these three concepts will
3:24allow you to get the most out of this
3:26tool and be able to hand over more and
3:27more of your day-to-day tasks to ChatGPT
3:30work so that you'll just be able to get
3:31a lot more done in a lot less time. And
3:34then we'll finish off with two power
3:36features of ChatGPT work being sites and
3:39browser use. First of all, sites allow
3:40you to create interactive websites and
3:43dashboards that are actually published
3:45and that you can share. And then
3:46finally, browser use allows you to
3:48automate tasks in the browser a whatever
3:50tool you're using maybe doesn't have a
3:52connected app to ChatGPT. So, let's get
3:54straight into part one on the setup.
3:57Now, before doing anything, what I want
3:58you to do is go ahead and download the
4:00ChatGPT desktop app. And there's a
4:02couple of reasons for this. First of
4:04all, it is the only place, so here I am
4:05on the ChatGPT work product page, and
4:08you can see that ChatGPT work is
4:09available to all plans when you're using
4:11the desktop app. But if you are on one
4:13of the paid plans, you can also use it
4:15on web and mobile. So, by downloading
4:18the desktop app, even if you are a free
4:19user, you can use this amazing tool. So,
4:22to find it, simply head over to Google
4:24and search for ChatGPT desktop app. You
4:27can see here the first link, download
4:29ChatGPT, and then download it for
4:30whatever operating system you're using.
4:32Now, if you happen to have one of those
4:33paid plans that allows you to also use
4:36ChatGPT work in the web version, I would
4:38still highly urge you to go and download
4:40the desktop app. You can see here here I
4:43am on the ChatGPT web version, and we do
4:45have access to these two tabs, chat and
4:47work. But this version of ChatGPT work
4:50here in the web version does not have
4:52the full feature set. And you can see
4:54here they're even prompting us to
4:55actually go ahead and open the desktop
4:57app. So, here I've got the ChatGPT
4:58desktop app open. In the desktop app,
5:01regardless of which plan you are on, you
5:03get access to all versions of ChatGPT.
5:06So, here in the top left we can see we
5:08have ChatGPT, but then you can also
5:10switch to Codex, which is that coding
5:11agent if you are doing any sort of file
5:13coding or software building. We're going
5:15to be sticking in the ChatGPT tab here,
5:18and then we have those two tabs, chat
5:20and work. So, let's toggle over to work.
5:23Now, the reason why it's so important to
5:24be running ChatGPT work in the desktop
5:26app is there's actually two modes in
5:29ChatGPT work, and only in the desktop
5:31app do we get access to both of the
5:33modes. If we have a look here in the
5:34bottom right, there's this little
5:35computer symbol. The first one is local
5:37mode, where you're actually running
5:38ChatGPT work on your computer, and then
5:41there's also cloud mode where you're
5:43running it in the cloud. Now, if you're
5:44using the web version of ChatGPT work,
5:47you only get access to this cloud mode
5:49and not the one on your computer. Now,
5:51just quickly on the differences between
5:53these two modes, on your computer or in
5:56the cloud, the one advantage that in the
5:58cloud has is it doesn't need your
5:59computer to be on for it to be running a
6:02task. But, that's sort of where the
6:04advantages stop. And for me personally,
6:06I'm only using this on the computer
6:08mode, and that's because it is able to
6:10actually access the files on your
6:12computer. And hopefully, by the end of
6:14this video, you'll see why that is so
6:17important. And on top of being able to
6:18actually access and work on the files on
6:20your computer, you also get access to
6:22the inbuilt browser in the Chat GPT
6:24desktop app, which you do not get in
6:26cloud mode. All right, so we're making
6:28sure that we are working in the local
6:30mode on your computer. And now, let's
6:33get into projects and how you go about
6:35working with local files in Chat GPT
6:38work. And the one habit that I want you
6:39to get into is every single task that
6:42you give to Chat GPT work should belong
6:44in a dedicated project. This is a little
6:46bit different than the way we're used to
6:48in the regular Chat GPT. You see here at
6:51the bottom, I have the ability, when I
6:52want to create a new task, a new chat in
6:55Chat GPT work, to choose a project. So,
6:57let me click on that one. I'm going to
6:58create a new project. And we can give
7:00our project a name and then also select
7:03any source folder. So, add folders that
7:05Chat GPT can read and edit. So, to start
7:08with, I'm just going to grab the folder
7:09that I want to use for this example. And
7:11here, what I have, I have a folder of
7:13basically images of different receipts
7:16that I need to log as business expenses.
7:18So, I'm just using that for this
7:19example, and I will open up this folder.
7:21Now, I quickly want to show you, before
7:23I actually create this project, there
7:24are two different ways that we can
7:25provide files to Chat GPT work. This is
7:28the recommended way, but I just want to
7:30show you what not to do so that you can
7:31see the difference. Here, if we click on
7:33the plus button, we can also attach
7:34files and folders this way. So, I can
7:37definitely click on attach files and
7:38folders, grab that receipts folder, also
7:41attach it this way to Chat GPT work, but
7:44these are two different things. They're
7:46not the same. They're not the same way
7:48to provide files to Chat GPT work. What
7:50I want to show you here, if I open up
7:52this right-hand side, what you'll notice
7:54is I only see the inbuilt browser. Now,
7:57of course, you won't know what else to
7:58expect, but let me show you how this is
8:00different when we actually set this up
8:02as a project. So, here I'm going to
8:03create a new project. I'll call this one
8:05accounting, and then I can go ahead and
8:08attach that receipts folder as the
8:11source folder, and then create the
8:12project. Now, with our accounting
8:14project set up as the project, if I open
8:17that right-hand panel, we now get two
8:19options, files and browser. When I click
8:22on files, this is really, really useful.
8:24Whatever project folders you basically
8:27set up and add here, you can see all of
8:29the files in that folder. And this makes
8:32it incredibly useful because what Chat
8:34GPT work is going to do in most cases is
8:36also create files for us. We'll be
8:38creating documents, spreadsheets,
8:40whatever it is, and it's very, very
8:43useful all having that contained in a
8:45project, and you can see the files it's
8:47working on and processing in the
8:49right-hand panel here. Now, a little
8:51quick extra bit of info, any project
8:53that you set up can have more than one
8:55of these source folders. So, here I
8:56could add an extra folder, and let's say
8:59I also attach our folder of invoices
9:01where we have all the invoice PDFs. So,
9:03it can have more than one source folder,
9:05but importantly, it can only have one
9:06primary folder, and this is going to be
9:08very important for the topics that we
9:10cover later in the video, like the
9:12agents.md and the different skills that
9:15we create. So, for now, I'm just going
9:16to work with one folder. I'll remove the
9:18invoices and hit save. All right, so
9:20let's give Chat GPT work its task in
9:22this accounting project. We have all of
9:24our files here in the project folder,
9:26and now I'm giving it a quite an
9:28in-depth brief. So, first of all, we
9:29need to let it know what is actually in
9:32its project folder. So, here I'm saying,
9:33"In your project folder, you'll find
9:35images of receipts that I need to
9:37process as business expenses. I then lay
9:39out the exact process that I wanted to
9:41follow. First of all, I wanted to
9:42extract the following detail from all of
9:45these images and then create an Excel
9:47file with all of those details saved.
9:50And now we want to also organize these a
9:53little bit smarter so that when it comes
9:55time to submitting these to our
9:56accountant, everything is nicely
9:57organized. So, I've asked it to create a
9:59subfolder called processed receipts and
10:01then also basically sort each of these
10:04images into subfolders based on the
10:07month of when the actual spend happens.
10:09So, let's send this one off to ChatGPT
10:11work. And now ChatGPT work has started
10:14on this task. Now, something that is
10:15drastically different when you're using
10:17ChatGPT work as opposed to regular
10:19ChatGPT's, these tasks do run a little
10:22bit longer. So, what we have at the
10:23bottom here is as I said, we're working
10:25with an agent. The agent makes its own
10:27plan and we can see that plan here. So,
10:29it said it's going to inventory the
10:31receipt images, visually inspect every
10:33receipt, build and verify the Excel,
10:36move the validated receipts, and then
10:37report logged or review needed and
10:40archived receipts. And so, one tip that
10:42is important here because these tasks
10:44will run a lot longer. If you ever see a
10:46plan that you do not agree with, you can
10:48either just simply click on this stop
10:49button and then correct it, or you can
10:52type a follow-up prompt and then what is
10:54known as steer the conversation. All
10:56right, we can see now after a little
10:57over 6 minutes, ChatGPT work has
11:00finished with this task. We have a
11:02summary output here of what got done.
11:04But before we do that, what I want to
11:05show is you can always click on this
11:07little drop down here to see basically
11:09all of its working through the task. So,
11:11you can see here when it viewed exactly
11:13all of these images. If it ever runs
11:15into any errors, you can really nicely
11:17debug here by basically clicking into
11:20this little working area and seeing all
11:21the steps that were done in the task.
11:23But now let's actually have a look at
11:25the result. So, we can see our main
11:26deliverable here is this receipt Excel
11:29file. Let's open this one up and we can
11:31preview it here directly inside the Chat
11:33GPT desktop app. We have all of our
11:36expenses with one row per expense in a
11:39nice Excel file. If we have a look on
11:41the right-hand side here, we can see we
11:43asked it to actually provide a
11:44confidence score. So, we have right now
11:47out of all of them, it's basically very
11:48confident on majority, but one of these
11:51does require our review. And we can see
11:53our review notes here saying that the
11:54VAT amount is clear, but the printed VAT
11:57rate is too faint to verify confidently.
12:00Now, if we go back and check out our
12:02project folder, if we open up the files
12:04again, we can see now this also looks
12:05very different. Instead of all of the
12:07receipts being in the root folder, we
12:09just have that Excel file. And now we
12:11have a new processed receipts subfolder
12:13with basically new subfolders per month.
12:16So, we can see here just one receipt in
12:18January, one in February, and three in
12:21April. And then we can obviously verify
12:23that here. Yeah, we can see the date
12:2423rd of April. Now, obviously this was
12:27just one example of the power of working
12:29with local files, but you can really
12:30apply this to whatever file types and
12:33whatever documents you are working with.
12:35Maybe you're working in the legal team
12:36and you have hundreds of PDF contracts
12:39where you need to compare terminology
12:41across contracts. Or if you're working
12:42in accounting like this, extracting
12:44financial data. Or you're working in
12:46sales and you want to have Chat GPT Work
12:48analyze your sales transcripts and then
12:50take that data and make finalized
12:52proposals in a clean document.
12:54Basically, whatever task you want it to
12:55do on local files, create a project
12:58folder, put the files in there, and then
13:00brief Chat GPT Work on the task it
13:02should do. Now, one final thing on
13:03projects, obviously when you start using
13:05projects a lot, you'll end up with quite
13:08a few projects for the different work
13:09that you are doing. Now, unfortunately,
13:11we don't really have the option to sort
13:13the projects by the most recent. So, if
13:16we click on this little sidebar here,
13:17you can see organize sidebar. I can
13:19either choose to have these organized by
13:20project or just in one long list. If we
13:23obviously want this by project, we can
13:25just then manually move around the
13:27different projects. The options that we
13:29get here in terms of sorting the chats
13:31by priority, last updated, or manual
13:33order, this refers to any chats inside a
13:36given project, not the projects
13:38themselves. All right, now let's talk
13:39about the agents.md file, which is a way
13:41that you can have persistent
13:43instructions for the projects that
13:44you're working on. To put simply, this
13:46is a way that you can work a lot more
13:47efficiently with ChatGPT Work and saves
13:50it doing the same work over and over.
13:53So, if you ever find yourself repeating
13:55the same instruction, this is a great
13:57candidate for an agents.md file. So, how
14:00this works is whenever you start a new
14:02session in a project, the very first
14:04thing that ChatGPT is going to do is
14:07read this agents.md file if it exists,
14:10meaning that anytime you have certain
14:13rules and preferences that you want it
14:14to follow always in that project, we can
14:17put them in the agents.md file. And it's
14:19also a great way to store any project
14:22context that you want ChatGPT to
14:24remember. Let's take a look at a
14:26real-life example. So, here in ChatGPT
14:28Work, I was working in a project that
14:30did not have an agents.md file set up,
14:33and I asked it a simple question, "What
14:35is the Notion database where we manage
14:37our YouTube videos?" Now, this one it
14:38was able to find pretty quickly. It only
14:40worked for 43 seconds, not bad at all,
14:43you might think. But, what you see here
14:44in the working is that it had to do
14:46quite a few different calls to Notion,
14:48search a few different pages before it
14:50actually correctly found the right
14:52database and gave me the link. Now,
14:54let's ask the exact same question inside
14:56our YouTube project where we have an
14:58agents.md file set up. And we can see in
15:02like literally a matter of seconds.
15:04Obviously, the other one was 43 seconds,
15:05this was literally two or three seconds.
15:07It didn't need to go and search through
15:09any tools, it immediately comes and
15:11gives us the answer with the link to
15:12that database in Notion. So, to show you
15:15what's going on here, if I open up the
15:16right-hand side and go into the files.
15:19So, this is our YouTube folder where I
15:20manage all of our YouTube videos and
15:23have AI help me in preparing them. We
15:25can see this is a very well-structured
15:27folder. Let me actually make this one a
15:29little bit bigger if I can. Here we go.
15:31So, the way that we structure this one
15:32is every time we have a published video
15:35that appears here in its own folder with
15:37the date it was published on, the video
15:39title, and the ID. Everything is nicely
15:41organized, and then we have preparation
15:43folders for the next videos that we're
15:45working on. But, what I want to
15:46highlight here is this agents.md file.
15:49So, anytime I work in this folder,
15:52ChatGPT work is going to read this
15:54first. And this one basically lays out
15:56how the folder is structured, which is a
15:58great way to make sure that the AI
16:00agents that you're working with keep
16:02this structure the way you want it. So,
16:03for instance, here it says, "Published
16:04videos live in folders with the
16:06following naming convention." Whenever
16:08we're working on a new video, create a
16:10prep folder. And so, the ChatGPT work is
16:12able to coordinate between this folder
16:14and our Notion database. We have the
16:16link to that one saved here, and that
16:18was how it was able to get this answer.
16:20So, whenever in this project I say,
16:21"Save it to Notion," it knows exactly
16:24where to go, and I don't need to tell it
16:25each time, or it doesn't need to go and
16:27search each time. And this is going to
16:28not only save you time, but it's also
16:30going to save you usage, because those
16:3243 seconds of ChatGPT looking through
16:35our Notion and finding the database is
16:37usage that it doesn't need to use if
16:39you've just set up an agents.md file
16:42with the key information. Now, to create
16:44an agents.md file in a project is very,
16:47very straightforward. Let me actually go
16:48back to this receipts processing
16:50example. Right now, we're working in a
16:52project where there is no agents.md
16:54file. We have a couple of options. So,
16:55there is an inbuilt if you type
16:57{forward-slash} init, for I guess
16:59initiate, there is an inbuilt
17:00functionality to create this file with
17:03the instructions. I do find this one is
17:04a little bit of an overkill. It is
17:06optimized towards Codex, so you'll
17:08probably end up with way more
17:09information than you actually need. In
17:11this case, let's say there's a certain
17:13instruction that I want to save so that
17:14whenever we work inside this receipts
17:17folder, it remembers it. I'm just going
17:19to basically ask ChatGPT to save this.
17:22So, an example thing that I might save
17:24to this agents.md, I'm asking it, "Can
17:26you please create an agents.md file and
17:28make a note that it should never create
17:29a new Excel file, but rather add new
17:31receipts to the existing one, of course,
17:33checking for duplicates." So, let's send
17:34this one off.
17:36All right. Now, ChatGPT Work has created
17:38that agents.md file for us. We see it
17:40here inside our project folder. And if
17:43we open it up, we can take a look. And
17:45here it is. So, this one These are .md
17:47file, which means markdown. It is just a
17:49text file. So, here are the instructions
17:51that ChatGPT has interpreted for what I
17:54gave it. So, it's got a few headings
17:55here. Use the existing Excel register,
17:57check every receipt for duplicates,
17:59archive, and verify. Now, for me, the
18:01workflow that I use most often is just
18:03telling ChatGPT Work what I want it to
18:05save in its agents.md file and having it
18:08either create or edit it directly. But,
18:10of course, this is just a text file, and
18:12you can also manually type this out or
18:15edit it yourself. So, for instance, here
18:16I could remove a line if I needed and
18:18save it or add any new instructions.
18:21Now, if you remember earlier when I
18:22showed that if you have more than one
18:23folder on your computer belonging to a
18:26project, the reason the primary folder
18:28is important is that's where ChatGPT is
18:30going to look for this agents.md file.
18:32Quick thing, keeping up with AI is
18:34basically a full-time job these days,
18:36and most people simply don't have the
18:38time to stay on top of the new
18:39developments. But, it literally is our
18:41full-time job. We spend all week
18:43training companies on AI, and everything
18:45we learn along the way lands in our
18:47community. Now, we found that the best
18:49way to stay ahead is to learn live with
18:52others. So, every week, we run a free
18:54live workshop, real use cases for
18:56marketing, sales, finance, whatever you
18:59do, and the tools behind them, from
19:01ChatGPT to Claude Code and everything in
19:03between. There's a link in the
19:05description below. Just drop your email
19:06in there, and you'll get a link to the
19:08community. All right. Now, let's talk
19:10about plugins. If we have a look on the
19:12left-hand menu inside the ChatGPT
19:14desktop app, underneath new chat we have
19:16a few different options and here we have
19:19plugins. Now, as I said at the start,
19:21mastering this concept is probably what
19:23will take you the furthest in terms of
19:25becoming, let's say, a legitimate power
19:27user of ChatGPT works. If we click into
19:30plugins. Now, when we land on this
19:32plugins tab, what you'll notice if you
19:34scroll down, you see a bunch of tools,
19:36most of which maybe or some of which you
19:38use on a regular basis. And if we take a
19:40look at this sub-line here, "Plugins
19:42work with ChatGPT across your favorite
19:44tool." So, reading this, a lot of people
19:46make the assumption that plugins are
19:48simply just a way to connect ChatGPT
19:50with the tools that they use, but that
19:52is not entirely correct and this is also
19:54what causes a lot of confusion in this
19:57terminology of plugins. You see, a
19:59plugin is not simply about connecting
20:00ChatGPT to a tool. A plugin is actually
20:03an installable package that contains two
20:06major elements. First of all, apps. So,
20:08this indeed is where you connect ChatGPT
20:11to a tool that you want to use, but also
20:13just as importantly, a plugin also
20:15contains skills. And skills are how we
20:17teach ChatGPT repeatable workflows,
20:19repeatable ways of working. So, if you
20:21want to think about it, in a plugin you
20:23can connect to an app, ChatGPT knows
20:25where to work, where to create things,
20:26where to edit things or where to pull
20:28information in and a skill teaches
20:30ChatGPT how to work. And because of
20:32this, when you're on this plugins space
20:34and browsing through the different
20:35plugins, you will notice tools like
20:37Gmail, like Google Drive, but you'll
20:39also notice plugins like Presentations.
20:41Presentations is not a tool. This plugin
20:44does not connect you with any apps
20:45whatsoever. It just contains one skill,
20:48Presentations create polished PowerPoint
20:50and Google Slide decks. If we compare
20:53that to the Gmail plugin, this one
20:54contains both an app and a skill. So,
20:57the Gmail app allows us to connect to
20:59our Gmail account and then the Gmail
21:00skill teaches ChatGPT how to do things
21:03inside our email. So, for instance, it
21:05teaches it how to search our emails, how
21:07to paste Gmail links, how to write
21:09emails. All of this is contained inside
21:11this Gmail skill. So now that we know
21:13exactly what plugins are, let's take a
21:14look at how to install them and then
21:15we'll go through a couple of examples.
21:18So for instance, this Gmail one, I want
21:20to give ChatGPT Work the ability to read
21:23and manage my Gmail account to automate
21:24some of my tasks. I can just click on
21:26install. Now because this Gmail plugin
21:28only contains one app, what it's
21:30immediately going to ask me to do is
21:31actually connect my Gmail account. Now
21:33I'm not going to connect my Gmail right
21:35away because I do want to show you
21:36another type of plugin. Another great
21:38one that I recommend you testing out is
21:40this data analytics plugin. If we take a
21:43look at this one, or first actually let
21:44me install it and you'll notice when I
21:46click install, there's no pop-up telling
21:48me to connect to any of my apps. It just
21:50go ahead and installs it. Now very
21:52importantly, it's just because I've
21:54installed it, you should not end there.
21:55What you need to do is click inside the
21:57plugin and what you'll notice here is
21:59this plugin contains more than 23
22:01different apps that it's suggesting you
22:03connect because it realizes when you
22:06want to do any sort of data analysis,
22:08you typically need to pull in
22:09information from your tools. So they
22:11realize, "Hey, most companies
22:12communicate on Slack or Teams. So you
22:14should be connecting one of these so
22:16that when you ask a question to analyze
22:18some data, how we going on a certain
22:19project?" It has the ability to look
22:21into Slack and find information on that
22:24topic. And so what you want to do now is
22:26after you install a plugin like this one
22:28that contains potentially more than one
22:30app, is go through the list and see,
22:32"Hey, what tools am I using and which
22:34ones will be useful for me to connect?"
22:36So in this case now, it also has Gmail.
22:38Again, a lot of useful business context
22:40lives inside your emails. So here I also
22:42want to make sure I'm going to connect
22:44and again I get that same pop-up. So
22:45this time I am actually going to connect
22:48to my Gmail account. I just need to
22:49select my email, verify that I'm giving
22:51OpenAI the ability to both view but also
22:53read, compose, and send emails. I'll
22:55allow this one and then we can just open
22:57back up the ChatGPT app. And now we can
22:59see that that Gmail account is
23:01connected. Now this data analytics
23:03plugin does not only contain apps. It
23:06also, as I mentioned earlier, contains
23:08skills. So, here we can see several
23:09skills and we can choose within a plugin
23:11which skills we want to have activated
23:13and which ones not. So, you can see we
23:15can analyze data quality, build
23:17dashboards, build reports, all the way
23:19up to visualizing some data. Now, once
23:21you've installed a plugin and connected
23:23the relevant apps and made sure any
23:25skills that you want to have enabled are
23:27enabled, we can use them in a chat. Now,
23:29what's very important is
23:31the PT will automatically decide when it
23:33should use these plugins. So, for
23:34instance, here if I would ask it what
23:36emails did I receive today, it would see
23:38that my Gmail account is connected and
23:40make the connection, "Hey, I should
23:42search in Gmail." But, we can also, if
23:45we want to give it direct guidance about
23:47using a certain plugin, we can do that.
23:49So, with the @ symbol, this pulls up the
23:51list of different plugins. So, for
23:53instance, I can do @ and select data
23:55analytics. That means I'm telling Chat
23:57GPT, "Hey, for this task, I probably
23:59want you to use the data analytics
24:01plugin." Similarly, if I want it to
24:03search for my emails, I can search
24:05@Gmail and tag it this way. So, if you
24:09do know that you want it to go into a
24:10certain tool, use a certain plugin, or
24:12if you want it to use a certain skill,
24:14for skills instead of the @ symbol we
24:16can use a forward slash symbol. And for
24:18instance, I know that there was a
24:19visualize data skill inside that data
24:22analytics plugin. So, this is obviously
24:23a bit of a random example here, but I
24:25could say visualize data, pull in the
24:26emails that I received today from Gmail
24:28and categorize them into appropriate
24:30categories. And we can see here it's
24:31used its data visualization skill and
24:33the Gmail app to create this little
24:35visualization of my emails today. As I
24:37said, this is just an example, but you
24:39can see here newsletters and updates are
24:41the first category. I can basically
24:42click between these categories and see
24:45the emails of today. So, that is how you
24:47can use these pre-existing plugins to
24:49basically do more of your work. So, what
24:51I would recommend is the very first time
24:53you get started with Chat GPT work in
24:55the desktop app, go to this plugin
24:57section, have a good look around, see
25:00which of your tools actually do have a
25:02connector. So, let's say if you're using
25:04HubSpot as a CRM, just search for it,
25:06you see it here, install, and then go
25:08and get yourself connected. As well as
25:11connecting your tools, then also make
25:12sure you check out any of the pre-built
25:14plugins from OpenAI, specifically around
25:17skills, like the spreadsheets, like
25:19presentations, or data analytics that I
25:21showed. But now the question comes up,
25:23what do you do if the tool that you are
25:25using does not have an existing plugin?
25:27Are you then stuck and you cannot use
25:29it? No, absolutely not. Let me show you
25:31what you should do if the tool that you
25:32want to use or connect to ChatGPT does
25:34not have an existing inbuilt plugin. So,
25:37the first thing you need to do is
25:38obviously search that it doesn't have
25:40one. So, in my case, I want to connect
25:41our email marketing tool, customer.io.
25:44So, I do a search here and you can see
25:46customer.io does not appear here. Now,
25:48in the plugin settings, we do get this
25:50little option at the top right that says
25:52add, but I've not had much success with
25:55this. What we can do is we can create a
25:56plugin. What that does is it opens up a
25:58new chat with this plugin creator. I've
26:01tried to use this, but honestly, I don't
26:03think it is the best method for actually
26:05connecting your app to ChatGPT. So, for
26:08this, what I actually recommend is going
26:09out of the desktop app and going over to
26:12the web version. So, now over in the web
26:14version of ChatGPT, I'm going to click
26:15in the bottom left here, and then click
26:18on settings. Here, we can now go to
26:20plugins, and we'll scroll all the way to
26:22the bottom and click browse plugins, and
26:24that gives us this similar screen. Now,
26:26if you do not see on this screen a
26:28little plus button here, that is because
26:30you need to first enable an additional
26:32setting. So, back in the settings, I'm
26:34going to go back to plugins, and this
26:35time scroll all the way down and click
26:37on developer mode, and then I'm going to
26:39turn this one on. Now, it could be,
26:41depending on if you're like enterprise
26:42or business user, that you may not have
26:45the ability to turn this on, then you
26:46need to get in touch with your ChatGPT
26:48admin. I think only enterprise have the
26:51option of enabling this for all users.
26:53Otherwise, if you want to connect a
26:54custom app, you'll just need to ask your
26:56ChatGPT admin to do it. But, if you're
26:58on a personal account, so whether that's
27:00free, pro, plus, whatever it is, or
27:02you're in a company plan and you're an
27:03admin, just make sure developer mode is
27:05toggled on. And what you'll see now,
27:06this plus button appears next to
27:09plugins. This plus button doesn't exist
27:11on the desktop app yet. Maybe they'll
27:13add it soon. Now, when I click on plus,
27:15it's a little bit different. I don't get
27:16brought to a chat with the plugin
27:18creator. I can just go and connect my
27:20tool. Now, this whole technology of
27:22connecting tools to AI models is
27:24something known as MCP, model context
27:27protocol. So, what you simply need to do
27:29is for whatever tool you want to
27:30connect, so in my case it is
27:31customer.io, and then just search for
27:34MCP. And here, if we have a look at the
27:36results, they've got a couple of links.
27:37I'm going to click on the documentation
27:39link, that's always the one I would
27:40recommend. And here, you can see get
27:42started with customer.io MCP. And what
27:45you're looking for is the MCP server
27:47URL. In my case, we're based in Europe,
27:49so I'm just going to copy this one over,
27:51go back over to my ChatGPT settings in
27:54the web version. I can paste in that
27:55URL. The authentication is OAuth in most
27:58cases. You can always check the
27:59documentation of the tool. I'm going to
28:01name this one customer.io.
28:04I understand the risks and create. All
28:06right, now that we've created that
28:08custom app, now we just need to sign in.
28:10So, we can click sign in with
28:11customer.io. So, on this browser, I'm
28:13already connected to my customer
28:15account, so I can just select what
28:16permissions I want to give ChatGPT and
28:18click allow access. All right, now we
28:20can see we have our connection to
28:21customer.io. But, in order for this to
28:23show up in the desktop app, now I'm not
28:25sure if this is something only for
28:27business plans or in general, what I
28:29found we need to do is click on these
28:30three dots and then hit publish. And
28:32they do make this a little bit hidden.
28:33It took me quite a while to find this
28:35one. Then, we land on this other window
28:37where we can publish this new plugin or
28:39this new app. We just have to accept the
28:41potential risk. Obviously, we use
28:43customer.io, it's a software that we pay
28:45for, so we can say that we trust this
28:47app and then hit publish.
28:50And now back inside the Chat GPT desktop
28:52app here in the work mode, we can verify
28:55that that new up app or plugin has been
28:58enabled. If I search for at, and I can
29:00see customer.io, so I can just ask it,
29:03what can you do with this tool?
29:06And we can see now, fantastic, the
29:07customer.io app is connected and lists
29:10out all the things we can do, like
29:12building and editing email templates,
29:13managing our transactional messages,
29:16broadcast, running AB tests in our email
29:18campaigns, all the way to actually
29:20creating and modifying segments and
29:22automation. All of this stuff that we
29:23were previously doing manually, now we
29:25can have Chat GPT work do this for us.
29:28So, now that we've connected Chat GPT
29:29work both to the inbuilt plugins and any
29:32custom apps, now we need to teach it how
29:34we want it to work, and that is where
29:36skills come in. Skills allow us to
29:38create reusable workflows that Chat GPT
29:40work can use on demand. Let me first
29:42show you an example of just how powerful
29:44skills are, and now teach you exactly
29:46how to create your own using my very,
29:48very simple method. So, continuing on in
29:50this session where we asked Chat GPT
29:52work what it can do with this new
29:53customer.io app, I'm going to ask it to
29:55create an email in customer.io that we
29:57can send to employees from a company
29:59that we've been training after they've
30:01completed our AI co-work program so that
30:03they can claim their certificate.
30:06All right, it's finished working and
30:07said it's created the email in
30:09customer.io design studio. Now, very
30:10important, I ran this without any skill
30:13attached. This is the before version,
30:14and in a second I'm going to run this
30:16with a dedicated skill for creating
30:18emails with our very own style. Let's
30:20open this one up and take a look. And
30:22here we can see in our marketing tool
30:24customer.io, Chat GPT was able to create
30:26this full email for people to claim
30:28their certificate, complete with all of
30:30the dynamic tags. So, if we have their
30:32first name, it's going to put their
30:33first name, otherwise it's just going to
30:35say, "Hi there." All of that work which
30:37would take a bunch of fiddly work.
30:38However, this really doesn't look like
30:40our template, and this is quite generic.
30:43Now, let me copy this prompt, and before
30:44I send it off, what I'm going to do is
30:46I'll head to plugins, and at the top
30:47here, we also have this little toggle
30:49for skills. So, these are any skills
30:51that you create individually that don't
30:52belong to a certain plugin, and we have
30:55this what CIO, which stands for
30:56customer.io, create marketing skill that
30:59we have added. And as I said, in a
31:00second, I'll show you how to create your
31:02own custom skills. Let me enable this
31:04one now, and we'll try the exact same
31:06prompt, but this time calling our create
31:09marketing email skill and send that one
31:11off. All right, and that one has now
31:12finished. Now, before we check out the
31:14email template, the first thing you may
31:16notice is this one took 1 minute and 51
31:19seconds to create the email. If we
31:21compare that to the one that ran without
31:24any skill, that one took over 8 minutes.
31:26So, in the skill itself, we tell ChatGPT
31:28all the things it needs to do to create
31:30emails in customer.io according to our
31:33template. Now, let's actually go and see
31:35the difference in the design. I'm going
31:37to open this one up, and here we have
31:39our email template according to the
31:41exact 9 x style. We have our 9 x logo on
31:44the top left, also our, like let's say,
31:46light cream background color that we
31:48use. We're using our direct button, and
31:51we have our email footer with the links
31:53to myself and my co-founder's LinkedIn
31:55profiles, our YouTube link, everything
31:57here according to our best practices.
32:00And that is really the differences that
32:02skills make. Now, if we take a look at
32:04how that skill is actually structured
32:06back over in our plugins area in the
32:08skills tab, we can click here on the
32:10create marketing email. There is a short
32:12description. This is the description
32:13that's going to help ChatGPT work
32:15identify when it should run the skill.
32:17And then, all the skill is is a text
32:19file, step-by-step instructions. It can
32:22also use different reference files. So,
32:24here we have a template markdown that it
32:26references, but this is just laying out
32:28all the information it needs, all the
32:30styles, all the calls it should make to
32:32customer.io.
32:34The great thing is you You not have to
32:35write any of these yourself. There's a
32:37very easy process to create your own
32:40skills. And that process is pretty
32:42straightforward. Before you even think
32:44about creating the skill, the first
32:45thing I want you to do is actually do
32:47the task step-by-step together with Chat
32:50GPT work. Whether that's teaching it how
32:52to create emails according to your
32:53template or how to create sales
32:55proposals or spreadsheets, whatever task
32:58you need doing that you want to do on a
32:59repeating basis, do the task with Chat
33:02GPT work together first. Once it's done
33:05the task, make sure you refine it and
33:07check the result until you are happy
33:09with it. So really, let's say Chat GPT
33:11work does something for you, don't go
33:12off and fix any mistakes yourself.
33:14Really tell it what it did wrong and
33:16nail it all the way down till you're
33:18happy with the result. Once you're happy
33:20with the result, what we're going to do
33:21is we're going to get Chat GPT work to
33:23save it as a skill and I'll show you in
33:25a second exactly how to do this. But
33:27then most importantly, after you've
33:28actually created the skill is the fourth
33:30and final step is to carefully test and
33:33always improve your skills as you go. So
33:35let me show you an example of that. In
33:37this task that I had here with Chat GPT
33:39work, we went back and forth and I
33:40basically taught it how to create images
33:43for our blog. As you can see here on our
33:46blog, we have a very specific style for
33:48the feature images that we use for our
33:51blog articles. And basically what I had
33:53Chat GPT work do was go and download 20
33:56examples of our existing blog article
33:59images and then asked it to create new
34:00ones for three new blog articles
34:02according to their title. Now that I'm
34:04happy with the results, with the images
34:06that it is creating, what I can do at
34:08the bottom here is simply do a forward
34:09slash and type skill creator. And Chat
34:12GPT work comes with an inbuilt skill
34:14creator skill. So it's a skill that
34:16teaches Chat GPT how to create other
34:18skills. It's pretty cool. So it says
34:20create or update a skill. And alongside
34:23the skill creator skill, what I can do
34:24is give some additional information. So
34:26I'm asking it to create a skill called
34:27generate blog feature image so that next
34:30time I can just give you a blog article
34:31title and then you'll immediately
34:32propose three images according to the
34:35style that you've learned here in this
34:36conversation. So, let's send this one
34:38off. All right, so that one's now
34:39finished. We can see it not only created
34:41the skill, but also directly installed
34:43it for us. If you ever want to check
34:44this again, head over to the plugins on
34:47the left-hand menu, then to the skills
34:49tab. Now, you might not see it straight
34:51away. You can also click on it's a
34:52little bit hidden, but you can list out
34:55all of the skills and there we have it.
34:56This is our generate blog feature image.
34:59By all means, feel free to go and read
35:01this. I don't really do this. What I
35:04would rather do is actually test it and
35:06see that it's working and then correct
35:07the behavior. But, for all means, you
35:09can have all the information here to
35:11check exactly how it created this skill.
35:14So, let's test this out now in a new
35:16chat. I can either type forward slash
35:18and then here we go, start typing
35:19generate blog feature image or also the
35:22dollar symbol. I think that only lists
35:24out the skills and nothing else. And we
35:26again can type out and here we have
35:28generate blog feature image. And now,
35:30all I need to do is give this skill a
35:32title. And so, here I want to create
35:34three images for the blog article The
35:36Eight Steps to Become a ChatGPT Work
35:39Power User.
35:41And we can see straight away it's
35:42confirmed it's going to use our feature
35:44image workflow. Interesting enough, it
35:46also is going to use the image
35:48generation skill that it relies on. So,
35:49what has happened here inside ChatGPT
35:52Work, there is a default skill for image
35:55generation. And what our blog feature
35:57image skill has done is basically
35:58referenced this one. So, one skill is
36:01able to reference another one. Because
36:03in that task that I did when I created
36:04these blog article images, we were using
36:07the skill. When I told it to save it, it
36:09mentioned that, "Hey, whenever you
36:11create one of these blog article images,
36:13run the image generation skill." And
36:15here we can see it has come up with
36:17three concepts for our blog article
36:19image. I didn't need to explain it
36:21anything about what style it should use.
36:22It just went ahead and ran its skill.
36:24So, I have this first one, which is
36:25quite cool, little power toggle with the
36:27Chat GPT logo. We have the eight-step
36:30path, and we have the capability hub.
36:33Three pretty solid ones to choose from
36:36just from running a simple skill. So, to
36:38recap that process, do the task once
36:40together with Chat GPT work exactly like
36:43I had done with those blog feature
36:45images. First, I made sure that in a
36:48chat or in a session, I was able to
36:50teach Chat GPT work how to create our
36:52blog article images, had a back and
36:54forth until we got the result right, and
36:56then I used the inbuilt skill creator to
36:59turn it into a skill. And now,
37:01obviously, as I start using it, if I
37:02ever see it go in the wrong direction,
37:04maybe I want to give it some new
37:06instructions, I can continually test and
37:08improve. And then, whenever I want to
37:10update it, I can just use that same
37:12inbuilt skill creator skill and just
37:14tell it what I want to change. So, now
37:16let's move on to those two power
37:18features that I mentioned, starting
37:19first of all with sites. This is an
37:21inbuilt feature that allows you to
37:23create interactive websites and
37:25dashboards directly in Chat GPT work.
37:27So, you'll notice here in Chat GPT work
37:29that there is also this sites menu in
37:32the left-hand side. That is not how we
37:34actually go about creating sites for the
37:35first time. This is where all the sites
37:37that we create live after we create it.
37:39So, let's actually first create one. The
37:41way we do this, again, we can just type
37:43that at symbol and type sites. You can
37:45see here build and deploy websites with
37:48sites. Now, for this example, what I'm
37:49going to do is I'm going to go back into
37:51that session that we had earlier where
37:53we asked Chat GPT work to look at the
37:55different receipt images and basically
37:57sort them into categories and process
38:00the receipts. Now, rather than doing it
38:02directly here inside this chat, maybe I
38:04want to use this again when I want to do
38:06a next batch or to create a skill or
38:08whatever else. Another really useful
38:09feature is underneath every answer we
38:12get back from Chat GPT work, we have
38:14this little symbol to what is known as
38:16fork the conversation or spin off a new
38:18task. So, what I can do is click on that
38:20one, and that means that I'm not
38:22anymore, like let's say, working on that
38:24original session, but I can start a
38:25brand new session with all of the
38:27context previously. And this is very,
38:29very useful. Sometimes you start working
38:31on a task and you think, "Hey, this is
38:33really, really good context. There's
38:34three additional things I want to do."
38:36You can basically fork these
38:37conversations into separate chats, and
38:40then give those as separate tasks. So,
38:42here what I'm going to do is I'm going
38:43to use the @sites inbuilt feature. And
38:46what I've asked it to do is create me a
38:47site which I can access from my phone so
38:49that I can directly upload or attach
38:51photos of receipts from my phone, and
38:53then you can process them for me. So,
38:55let's send that one off, and now we can
38:57see this one has run for over 44
38:59minutes. It did take a little bit of
39:00time to create our app, but while this
39:02is working, as always, or hopefully as
39:04you know by now, the ChatGPT work
39:06sessions, they do take longer. So, the
39:07goal really be, we'll set a few of these
39:09up, delegate the work while you work on
39:11something else, and then check back in
39:13on the progress. So, this one's worked
39:15for 44 minutes, and we now have our
39:17page. And even more incredibly, it even
39:19went ahead and created a social sharing
39:21image. So, if I share the link to this
39:23new app or site that I've created, this
39:25is what people will see. It's given me a
39:27few instructions. So, what I do need to
39:29do is add my OpenAI API key, because
39:31this will be live on the web, and we
39:33still need an AI model from OpenAI to
39:35actually process the invoices. Before I
39:37actually go and check out this site, let
39:40me show you exactly where we can
39:41configure all this. I've already gone
39:43ahead and done this, but if we click on
39:44the sites on the left-hand side, we see
39:46our receipt inbox site that has been
39:48created. If I now click into the
39:50settings, what's really great here,
39:52there's a very secure way of us to share
39:54any API keys here. So, we don't need to
39:57One big downfall of like vibe coding
39:59apps is often times API keys get shared
40:01in just source code or things like that.
40:03Here we can easily in this site setting
40:05add any variables that we want. So, we
40:07have one variable which isn't a secret,
40:09which is just what model are we
40:10currently using for this app. And then
40:12here I've added my OpenAI API key. So,
40:15So in that conversation, let me actually
40:16now open up this app that was created.
40:18I'm going to open it up in my browser.
40:20Now I am already logged in with ChatGPT,
40:23and we can see here that this is hosted
40:24at chat.openai.com. So whenever you
40:26create these sites, you get basically a
40:28custom domain. Let me show you what
40:30happens if I open this one up in an
40:32incognito window instead and just paste
40:35this one in. You can see what we need to
40:37do to access that site. We actually need
40:39to log in with ChatGPT. So there is an
40:42extra level of security. So you can
40:44basically create sites that contain your
40:46business data, the way that you work,
40:48and know that only people in your
40:49company that you share it with who
40:51obviously need a ChatGPT login, or
40:53yourself who is logged in with ChatGPT,
40:55can actually access the site. So here we
40:57can see this one, and what I'm actually
40:58going to do, let me open this one up on
41:00my phone. And so now here we see it on
41:02my phone, this app that I have created
41:05to manage my receipts. This was
41:06literally done in one shot. I've already
41:09tested this with a few receipts. So
41:11let's actually test it out. Now I'm
41:12going to take a photo, and I have a
41:15receipt here that I can just take a
41:16photo of. Use photo. We can see now it's
41:20working through, and perfect receipt
41:23added and checked. Let's see that. Yeah,
41:24it perfectly nailed it. So 320, I can
41:26take a look at this one. So here we have
41:28that image, all of the relevant fields
41:31saved, the gross amount, the currency,
41:33the category, what it's a description
41:35of. This is pretty cool. And literally,
41:36this was 44 minutes of work, and then it
41:39built an entire custom app. Now Sites
41:41isn't only incredibly useful for
41:43creating websites and web apps where you
41:45want to quickly prototype, but also for
41:47creating dashboards. Because ChatGPT
41:49work makes it so easy and is so good at
41:51analyzing data. Here I had it come up
41:53with a full dashboard for our YouTube
41:56numbers using the BigQuery connector and
41:58the YouTube API to pull in live data.
42:02And now we have here a full breakdown of
42:04our YouTube channel, which video is
42:07performing, how many subscribers we're
42:08getting, are we performing, are we
42:10having any sort of dips. And this is
42:11something that I can then share with the
42:13team. When it comes to managing who has
42:15access to the site, so here on the
42:17left-hand panel, click on sites, and
42:19then we can just click on share. And
42:21then here I have a few different
42:23options. I can choose anyone in this
42:25workspace. So, if you're using ChatGPT
42:28in a company, you'll have a ChatGPT
42:30business or enterprise plan, you can say
42:32anyone in this workspace, and
42:33immediately any everyone at your company
42:36has access to this, or you can also say
42:38anyone on the internet, but obviously be
42:41careful with what you choose. And you
42:43also have the ability to add individual
42:45people. So, for instance, here if I
42:46wanted to add my co-founder Alex to our
42:49YouTube dashboard, I just need to enter
42:50his email and then can invite him. All
42:53right, now onto our final power feature,
42:55and that is the built-in browser inside
42:57this ChatGPT desktop app, and it unlocks
43:00a ton of powerful use cases, especially
43:02for ChatGPT work. So, if we toggle over
43:05this right-hand toggle side panel, we
43:07see that there is a inbuilt browser.
43:10open this one up, and then you can just
43:11visit any website just like you would in
43:13Chrome, in Safari, in Internet Explorer.
43:15I don't even know if that's a thing
43:16anymore, but whichever browser you are
43:17using, you can just use this the exact
43:20same way. And now obviously here the big
43:21unlock is that we don't need to manually
43:23go and visit a link here. What we can do
43:25is actually prompt ChatGPT work to do
43:28some browser work for us. So, what I'm
43:30going to do is type the at symbol and
43:32then grab the inbuilt browser. You can
43:34see here control the in-app browser with
43:36ChatGPT. And in this use case, what I
43:38want to set up for myself is some AI
43:40news monitoring. So, I'm going to paste
43:41in the following prompt after this. I'm
43:43going to say, "Use your inbuilt browser
43:45to open up this website." And this is
43:47actually the release notes from
43:50Anthropic, from Claude. And so, that I
43:52want to keep up-to-date whenever new
43:53things are coming out about this tool.
43:55Imagine I'm someone from the team of
43:57OpenAI, a bit of competitor monitoring
43:59here. I say, "Read the page directly. Do
44:00not rely on the search results or
44:02snippets. Maintain a file named Claude
44:04release notes CSV inside the current
44:07project." And now here I've basically
44:08told it what information I want it to
44:11basically capture from this page. And
44:13then I said only look back until the
44:15beginning of June because I don't need
44:16to worry about stuff too much further in
44:18the past. Let's send this one off.
44:21And we can see now a new tab has opened
44:23up here. If we actually click on it, we
44:25can see ChatGPT has now opened up this
44:28page. I guess it's going to be scrolling
44:29through now. So the live page is
44:31accessible, includes releases through
44:33July 24. I found 15 release items from
44:35June 2nd through to June 24th. This has
44:37read this way quicker than what I would
44:40have been able to do. A few dates
44:41contain multiple releases, so I'm
44:42treating each title item as its own CSV
44:45row. Very, very intelligent. While it's
44:48doing that, I can just show you here.
44:49This is basically the release page that
44:52Claude have. You could do this the same
44:54if you want to keep an eye on a
44:55competitor's blog, or if you want to
44:56keep an eye on a competitor's pricing
44:58page, you could just basically have
44:59ChatGPT visit it through its inbuilt
45:01browser. Now while this one's working,
45:03what's also amazing is this inbuilt
45:05browser. If I head over to my settings,
45:08and we can see down here in the browser,
45:11there's a ton of settings that we do
45:13get. And the one very important one is
45:14that it does come with an inbuilt
45:16password manager. So if for instance,
45:18there are some tools that do not have a
45:21connector, and we do still want to
45:23automate some of the work inside that
45:25tool, we can safely add the password
45:26here in the password manager. Then
45:28ChatGPT can log on to it. Maybe in some
45:30cases it might even make sense that
45:32you're creating a dedicated account for
45:33ChatGPT to use that tool. But this comes
45:36in very, very handy because a lot of the
45:38work needs to be happening inside a
45:40logged in experience. And especially if
45:42that tool doesn't have an API, or
45:44doesn't have any way for ChatGPT to
45:46automate the work, you can now do that
45:48using this inbuilt browser. All right.
45:50And just like that, within a few minutes
45:52we have our created CSV file. Let's take
45:54a look. So obviously this one's a little
45:55bit um hard to view here because it is
45:57just the raw CSV. But we can see for
45:59every single article, or for every
46:01single release note, we have the date,
46:02the release note, and that description,
46:04and any links that were included. Now,
46:07to make the inbuilt browser even more
46:09powerful, what we can do is actually set
46:11this to run on a schedule. So, you see
46:12here in the top left, we also have the
46:14ability to schedule some tasks. So, what
46:16I'm going to do is add a simple prompt.
46:18So, I've now asked it to create a
46:19scheduled task so that you're checking
46:20this website four times a day. When you
46:22open the site, check if there's any new
46:24releases, compare it to what we have in
46:26that CSV file. If there have been new
46:28releases, make sure you add them to the
46:29CSV file, but more importantly, use your
46:31Gmail connector to send an email to me
46:34and my co-founder. So, let's send this
46:35one off.
46:37All right, now that has confirmed it has
46:38set up this scheduled task. We can also
46:40verify that. If I click on the left-hand
46:41side here to scheduled, we can see this
46:44one is also saved here. It is set to
46:46active monitor Claude release notes
46:48every 6 hours, next run in 7 hours. Now,
46:51obviously, what's very important is you
46:53will need to have the ChatGPT app open
46:55at the time that this scheduled task
46:56kicks in, but you could just have this
46:58running at a quicker frequency, for
46:59instance. And even if your laptop is off
47:02for one of the scheduled tasks, the next
47:03time you open it and another schedule
47:05kicks in, it will then run. So, that
47:07really is 97% of what you need to know
47:10about ChatGPT work. But, knowing how the
47:13tool works and seeing what actually gets
47:15done in a normal week, those are two
47:17different things. So, click this video
47:18right here. In it, I walk through eight
47:21agents that we've built with ChatGPT
47:23work, from rebuilding our reporting
47:24dashboard off a single screenshot to a
47:26lead research run that you kick off at
47:28your desk and approve from your phone.
47:30Every single one is built on our real
47:32work at 9 x, the real thing, not just
47:35theory, and all of the prompts and
47:36skills behind each one are over there to
47:39grab. So, I'll see you over there.

About this transcript

This page contains the full transcript of How to Use ChatGPT Work Better Than 99% of People (Full Guide) by 9x, generated from the public captions YouTube serves with the video. The transcript has 10,390 words across 1,429 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.