Big Idea

The argument in one line.

ChatGPT's Work mode turns chat into a standing employee: connect it to your real apps once, and eight scheduled automations quietly handle commitments tracking, hiring, social growth, and a monthly business review with no manual input.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

A solo founder or small-business owner running multiple tools (Notion, Gmail, Slack, Twitter) who wants recurring admin work off their plate.
Someone already curious about AI agents who wants concrete, working automation ideas rather than theory.
A creator running a personal brand account who wants research and drafting handled without losing their own voice in the final posts.

SKIP IF…

You don't already use a paid ChatGPT tier or aren't willing to connect Gmail, Slack, Notion, or Twitter to a third-party AI agent.
You want a no-code automation platform comparison. This is one creator's personal ChatGPT Work setup, not a Zapier/Make review.

TL;DR

The full version, fast.

Riley Brown treats ChatGPT's new Work mode as a standing employee: once it's connected to Notion, Gmail, Slack, Twitter, and Wispr Flow via plugins, he hands it a plain-English prompt and a schedule instead of doing the task himself. He walks through eight always-on automations built this way: a daily commitments-and-deadlines tracker that emails a teammate, a hiring dashboard that scans inbound emails, Twitter research and draft-tweet generation through Scrape Creators and Typefully, a podcast-guest finder, sponsorship-email forwarding, a nightly import of text messages via Codex, and a monthly business-review report emailed automatically. The lesson: plugins give an agent both context and the ability to act, and scheduling turns a one-off chat into a recurring system.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 04:07

01 · Intro: from chatbot to standing agent

Promise stated: eight cloud automations built directly from the desktop, web, or iOS app. Explains ChatGPT Work and plugins as context + actions.

04:07 – 05:39

02 · #1 Commitments and deadlines tracker

One prompt connects Notion, Gmail, Slack, and Wispr Flow meeting notes; agent builds a commitments database in 4 minutes, then schedules a daily 8am run plus an accountability email to a teammate.

05:39 – 08:26

03 · #2 Hiring dashboard

A public hiring email gets scanned daily; every applicant is logged and researched automatically, with a nighttime research pass added to the same automation.

08:26 – 10:36

04 · #3 Twitter research scraping

Agent uses Scrape Creators to pull relevant, high-engagement tweets into a Notion database every few days, purely as raw material for the next automation.

10:36 – 13:15

05 · #4 Drafting tweets

Agent turns the scraped tweet database into 10 Typefully drafts a day in the creator's voice, referencing past high-performers, but never auto-publishes.

13:15 – 14:28

06 · #5 Finding podcast guests

YouTube-researcher skill scans for guest candidates and adds them to an existing manual database; a keyboard shortcut grabs an open Notion page as prompt context.

14:28 – 15:39

07 · #6 Email forwarding

Simple rule: reroute sponsorship emails to a teammate's address twice a day. Set up entirely from the phone.

15:39 – 17:47

08 · #7 Importing texts (via Codex)

Codex (chosen for its local permissions) reads and summarizes iMessage threads with key contacts daily and stores them in Notion for later context.

17:47 – 19:23

09 · #8 Monthly mega summary

The highest-effort automation: once a month, the top model reviews every connected tool, ranks findings by business importance, and emails a polished report.

19:23 – 20:07

10 · Summary: all eight automations recap

Recaps all eight workflows and where they run: desktop, web, or iOS app, plus sponsor thank-you to OpenAI.

Atomic Insights

Lines worth screenshotting.

ChatGPT's Work mode turns a chatbot into an agent with its own computer: files, a browser, and app connections it can act on directly.
Plugins give an AI agent two things at once: context (it can read Notion, Gmail, Slack) and actions (it can write back to all of them).
A single prompt connecting Notion, Gmail, Slack, and meeting notes produced a fully organized commitments database in about four minutes.
Any recurring prompt can become a standing automation just by clicking schedule once the first run succeeds.
Cc'ing a real human on an automated daily email is a low-effort way to build accountability into an AI workflow.
Any public email address becomes a lead-gen funnel once an agent is scheduled to scan it daily and log senders into a dashboard.
Twitter research and draft-writing are safe to fully automate. Publishing the final post in your own voice is deliberately kept manual.
A keyboard shortcut can grab the open app's content as context for a prompt without retyping or re-explaining anything.
Text messages are a rich, underused data source for AI agents, since so much real business communication happens outside email.
The one automation that should run least often is the one that touches every connected tool at once, using the most capable model available.

Takeaway

Turn one prompt into a standing employee.

WHAT TO LEARN

The unlock isn't a smarter prompt. It's connecting an agent to your real tools once, then scheduling that same prompt to run itself forever.

02#1 Commitments and deadlines tracker

Connecting Notion, Gmail, Slack, and meeting notes to one prompt let the agent build and populate a full commitments database in about four minutes, work that would otherwise be manual admin.
Emailing a second human a daily summary from the automation turns a private AI process into an actual accountability system.

03#2 Hiring dashboard

A single public email address, once connected to an agent, becomes a self-updating hiring dashboard: anyone who emails it gets logged automatically.
The same automation can be split into a daytime capture pass and a nighttime research pass on each new lead, without extra manual work.

04#3 Twitter research scraping

Growing a brand-new account starts with research, not posting: the agent pulls competitor and topic tweets over 100 likes into a searchable database before anything gets written.
The research database itself is never meant to be read directly. It exists purely as raw material for the next automation in the chain.

05#4 Drafting tweets

Drafting is automated end-to-end, ten tweets in three minutes, but publishing is deliberately kept manual so a human still edits and approves every post.
Feeding the agent your own best-performing past posts as style examples keeps AI-drafted content in your voice instead of generic.

06#5 Finding podcast guests

A researcher-style automation can scan for guest candidates and add them straight into an existing manually-built database, no new tool required.
Highlighting an open app and using a context-grab shortcut lets the agent read a page's content without retyping or re-explaining it.

07#6 Email forwarding

The simplest automations are often the most useful: reroute any sponsorship email to another address twice a day takes one sentence to set up.
Automations can be created from a phone, not just a desktop, which matters when the idea occurs to you away from your computer.

08#7 Importing texts (via Codex)

Text messages carry business context that email misses entirely, so an agent that can read and summarize message threads has a fuller picture of active deals.
Some automations need a coding-focused agent specifically because it runs locally with broader default permissions than a chat app.

09#8 Monthly mega summary

The highest-effort automation should run least often: one deep monthly review across every connected tool, using the most capable model available, beats a dozen shallow daily checks.
Scheduling something to run on the last day of the month for the next 12 months in one sentence removes any need to remember to trigger it manually.

Glossary

Terms worth knowing.

ChatGPT Work: A mode inside ChatGPT that gives the AI its own computer with files, a browser, and connected apps, so it can complete multi-step tasks instead of just answering prompts.
Plugins (ChatGPT): Connectors that let a ChatGPT agent read data from third-party apps like Notion, Gmail, and Slack, and take actions inside them, rather than just chatting about them.
Scrape Creators: A tool for pulling structured social media data such as tweets, engagement stats, and media at scale, used here to feed a Twitter research automation.
Typefully: A social media scheduling tool that manages drafts and posting across multiple X/Twitter accounts through an API key.
Codex: OpenAI's coding-agent product, used in this video for a text-message-summarizing automation because it runs locally with broader default permissions than the ChatGPT app.

Resources

Things they pointed at.

00:00toolChatGPT (desktop, web, iOS) ↗

01:30toolNotion ↗

01:00toolWispr Flow ↗

08:26toolScrape Creators ↗

10:36toolTypefully ↗

15:39toolCodex ↗

00:00productAgent Native (Riley's company) ↗

Quotables

Lines you could clip.

03:01

“Notice here that after four minutes, GPT work is done. And I want you to think for a second, how long would this task take you?”

Direct provocation plus a clean ROI framing, no setup needed.→ TikTok hook↗ Tweet quote

02:36

“There's two main reasons why plugins are so useful. The first one is context. And the second one is actions.”

Names the core mental model for the entire video in one line.→ IG reel cold open↗ Tweet quote

11:57

“I believe that if I just put in 10 minutes a day, I can make them 10 times better because I have a lot of experience on Twitter.”

Honest limit on automation. Keeps a human editing pass instead of full autopilot.→ newsletter pull-quote↗ Tweet quote

19:13

“You can just deploy intelligence and this will happen in the cloud.”

Closing thesis statement, quotable on its own.→ IG reel cold open↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphorstory

Today, we're going to turn ChatGPT from something we prompt into an agent that does our work in the background 24 -7. This video is going to be a complete guide to setting up agents to do your work in the cloud autonomously directly from the ChatGPT desktop app, web app, or iOS app.

And so in part one, I'm going to explain the different parts of ChatGPT that are relevant for agentic tasks. Then in part two, I'm going to show you my eight favorite agent workflows that I've set up that have helped me and my business the most so that you can create similar workflows that help you. All right, guys, today we're creating agentic workflows that work 24 -7 in the cloud.

Whether you're using the desktop app like I am here. Or if you're using the ChatGPT web app on ChatGPT .com, or if you're using the ChatGPT iOS app, we are going to be able to create these cloud agents that do our work. Of course, we're going to be talking about the eight agentic workflows that really drive revenue in my business.

But first, I do want to talk about some basics that you should understand before we get here. So the first thing that you need to understand is that ChatGPT has evolved. And there's a new setting in ChatGPT called ChatGPT.

This is where a chatbot becomes an agent, right? This AI agent has access to a computer. with files, with a browser, and it is becoming much more like an employee.

And you can use this directly in the web app. You can use it in the desktop app, chat and work. Also, you can use it in the iOS app.

You can go from chat to work. And in my opinion, the most useful part of GPT work are something called plugins. And so I could use the Notion plugin, for example, and I'm going to say, please create a database called commitments and deadlines.

Please look at my Gmail and look at my Slack and find any commitments or deadlines and put them in a database from the past week. Oh, also look at my meeting notes at WhisperFlow because they now have meeting notes built in and we can. Run this prompt.

And so there's two main reasons why plugins are so useful. The first one is context. And the second one is actions.

Plugins allow your agent to get full context of all of the data that's inside Notion, that's inside Gmail, that's inside Slack, and that's inside my meeting notes on WhisperFlow. But not only... Does it have context and able to pull the data from these four platforms that I use for my business?

It can also take actions in all of these. It's actually going to create a database inside Notion. You'll see in just a second.

Okay, so notice here that after four minutes, GPT work is done. And I want you to think for a second, how long would this task take you? Especially if the result is this.

So if we open this up inside the ChatGPT browser, inside the ChatGPT desktop app, we can see the database that it created. It went through Notion, Gmail, Slack, and all of my meeting notes and found all of this information. It added the commitment.

It added the counterparty. It added the evidence. It added the owner, the source, analyzed four different platforms, and created a new Notion database, and it did that in four minutes.

And that is how powerful it is. Okay, so for me, it was about five minutes of total work, one of which was typing out this prompt, and then four minutes of waiting for its response. However, there's something inside the ChatGPT app, and that is this scheduled tab.

And so on the schedule tab, we can do something like this. So you'll notice here that it said, if helpful, I can set up update commitments database daily so new commitments are captured automatically. So I can either click on this directly or I can type into this.

And so this brings me to special agentic workflow number one that I use, which is this commitments and deadlines workflow. And so what I can do is I can just click on this right here. And so now that we can see this right here, all I need to do is go up to schedule and we'll see it right here, which is update commitments database.

This will run daily at 8 a .m. OK, let's actually add one additional thing so that I'm held accountable by another human on these things. Hey, I actually want you to email.

I'm going to put Emily at. agentnative .inc, please send her an email with a link to this database so that she can view them. And then whatever ones you add, please add and then make sure to mention that this is a full list of the commitments and deadlines Riley has agreed to.

And this should be emailed every single day as well. And then you can always mention like this should be part of this automation. And so basically, every morning, It'll analyze all of these things.

It'll put them in a Notion database, and then it will send an email to Emily from me so that she can see it and then hold me accountable to those items. And so there you go. We've created this first automation that just happens every morning.

AI will go through all of these thanks to the plugins and get that work done every single morning, 8 a .m. That's workflow automation number one. Okay, so the second automation that I'm going to do with ChatGPT work involves hiring.

I tweeted this the other day. I said, if you want to use or test agent platforms and write about what you learn and get paid for it, please send me an email. I'm hiring at Agent Native and I put my email right here.

And so what I can do now. I can very easily have AI analyze my email and add all of the people who've sent me an email into a database. In fact, I have done this already.

This is my hiring dashboard. I have a lot more fields here. I just can't show them because I don't want to give away any sensitive information.

And what I want to do is I want this to be updated every single day. And so what you can do is you can just go to wherever your Notion database is and I can say, hey, I've been hiring. for this position.

And then what I'm gonna do is I'm just going to give a screenshot of this tweet. That's all I'm gonna do. And they are sending emails to the email listed there.

And all of the entries are here. in this database. And then I can paste that Notion link right here.

Please, can you create an automation that adds all of these people to that database, create another automation for nighttime, which researches all the people and adds research about these people in the notes every single day in the morning, add them at night, research them. And so I can fire this prompt in. And again, it works directly from chatgpt .com.

It'll create this automation. Okay, 43 seconds later, we now have these right here. Both run daily.

If you put your email out there or any email that you want to connect to GPT work out there on the internet to gather information, you can just use GPT work to monitor that email and turn it into a dashboard. Okay, so the next two AI agent workflows that I want to create involve social media, specifically tweets on my new brand account.

You know, I've done a pretty good job growing my Twitter. I'm at around 233 ,000 followers. But I do want to grow my new account that I created a few months ago called Agent Native.

And that's the name of my company that I'm building. I want to do two things with this. The first thing that I want to do is research for tweets.

And then I want to actually schedule. the tweets. And so the first automation number one involves scraping and like it's just doing research on Twitter.

And then two is scheduling. And what we're going to be doing is we're going to be using typefully for scheduling and we're going to be using scrape creators for scraping. OK, so this brings me to automation number three, which is Twitter scraping.

So here's the prompt that I'm going to use is I want to grow agent native. my Twitter account for the brand. I comment on the agent space and make high quality content and around OpenAI Codex and similar tools.

Look up the similar tools. And you can also look at the... Handle which is agent native and then I want you to search Twitter and find relevant popular tweets Over a hundred likes that show important updates to agents and becoming agent native new product updates new model releases And I want and we're gonna be using scrape creators in the description below I have a little guide that you can get for scraping social media And I want you to create a new database in at notion.

We're using another plug -in called scraped tweets. Please do this now.

Scrape the latest. So now I'm going to have it scrape. It's going to create a Notion database, store all the data for the tweets.

Then we'll create an automation to do this every few days. Okay. So this is going to take a little bit and then we'll create the automation once it's done.

Okay, so after eight minutes it's done and I can actually open this up directly in the browser. We should see all of the tweets are pulled. Here we have the link URL.

Any media URL is automatically pulled. This is really cool. We see all of the replies, reposts, etc.

Okay, this is just step one. And then we're going to create tweets from the Agent Native account which provide commentary about these posts. That's going to be number four.

We need to finish this one and say, hey, please create an automation that automatically gets this done at... 9 a .m. every morning.

I want that to go off and update this database. Okay, and there you go. This is going to happen every single morning at 9 a .m.

Okay, so that brings me to automation number four, which is creating commentary posts from my agent native account. So for this, I'm actually going to be using Typefully. Typefully allows me to sign into a bunch of Twitter accounts.

And if I go to settings here, and click settings and go to API, I can get an API key. I've already set this up, but I gave this API key to GPT work and now it can use a skill.

I believe it's just going to be called typefully drafts. And so now what I'm going to do is I'm going to say, please. Can you create five draft posts that are actually let's do 10 draft posts on the best posts from the ones you've just gathered.

So these should either be quote tweets or it should post the video. Look up how to do that. Please do that on 10 of them and put it on the agent native account.

Do not actually post them. Just draft them on typefully. And so.

That is going to be the workflow. We gather all of the tweets here in a giant Notion database. We're not ever going to look at it.

AI just stores all the data there. And then it's going to take the most recent ones or the best ones from the past day and add them to Typefully. And this is where I can come and I can just approve them and schedule them.

And so that's how we can post more on Twitter, except I will actually do some manual editing because I don't want AI to tweet for me. I believe that if I just put in 10 minutes a day, I can make them 10 times better because I have a lot of experience on Twitter. All right.

So after three minutes, it is done. And if I right click on this, open an external browser. There we go.

Look at this. It created. I'm going to zoom out a little bit.

It created all of these tweets. All 10 of these tweets are. ready to be posted.

Voice is becoming real input layer for coding agents. Gemini can now ground what you say against the code, file names and variables, and what's on the screen. You won't just type prompts into agents, you'll direct them while you work.

This is very cool. And it quote tweeted them, it did the post the video thing pretty well, which is really cool. This is awesome.

Okay, so now what I want you to do, I want you to do this every day. And I want you to add five posts every day. Always pay attention to the previous tweets that I post on Agent Native and take from my voice, especially the ones that perform well.

Take the ones that perform well, and those should be the ones that you use as examples going forward. So please create that automation now. All right, 9 a .m.

every day. We now have that one set up as well. Okay, so the next four are a little bit quicker.

Here is automation number five, which is finding podcast guests. So I'm going to say, please do deep analysis using the YouTube researcher skill. This uses scrape creators.

It just does it in a specific way to analyze the YouTube transcripts to look for good guests for my podcast. And then I said, please add them to this database. Now, here's a fun thing that you can do when you're using the desktop app.

Check this out. So if you're typing a prompt here, I can just open up Notion and let's go to the pod. guests database and I created this manually a while ago so what I can do here is while on this I can click both command keys so check this out it automatically just takes the page and adds it as context this should also be an automation for every morning at 11 a .m.

and so now let's just wait And it's going to do some research, add some more people to this database, and then it's going to create the automation. And there you go.

It created the automation. And remember, you can always click directly on the automation to see that it did it correctly. If you want to change any of the details, you can edit it directly in here.

Okay, so for number six, I want to do one from the phone. you can create these scheduled tasks directly from your phone. So I'm going to type this in right here.

This one is actually very common and I highly recommend it if you have multiple emails. So... The email I have connected is my main agent -native email.

But many people reach out to me about brand sponsorships. And so I want to be able to do email forwarding. So I'm going to say, when someone emails me about a brand sponsorship, please reroute it to Riley at RakugoMedia .com.

Forward it to that email so my team sees it. Do this. 2x per day.

Set up that. Automation and I can fire that prompt out And there you go. This one is very simple But I find that a lot of people would find this one useful and again, you can click directly on it right here You can see daily forwarded sponsorship leads We can click on this right here and this will forward anything to my email to this new email Which is really cool.

And again, you can edit it directly if you find some mistakes in the automation. Okay.

So now I want to talk about the seventh automation that I create within the ChatGPT app. This is the only one where I do it inside Codex because Codex has more permissions by default. And for this one, we're going to be using the new messages.

integration, which allows me to pull directly from messages. And what I can do here is I can say, please summarize my messages from Adam D from today in text. This allows Codex to go straight into the database where my iMessages are stored, and it can get back to me.

with exactly whatever question I'm asking. It can even summarize every text you've ever sent. This will use a lot of tokens and might be somewhat expensive, but you can do it.

And the point is you have full access to your text directly in Codex. And so this is a very clean summary. And I changed it to the last five days because there wasn't actually much from the last one day.

But this summarized our conversation very concisely and perfectly. And so the point is I can take any. text message chain with anyone and I can ingest it with Codex and I can actually store it in a database.

And so that's exactly what I've done for sponsorship texts. I have a manager for my content and they actually have a full team. And so I have chats with all of them.

And because I want my agents to have context over everything, every single day at 8 a .m., it reads my texts with four different contacts. If they're in a group chat, With other people, it'll also bring those in and it stores that data inside Notion, just like the previous ones.

So that if I ever ask any of my agents about any deal, it has all of that context from text messages. Because so much of my communication is done via text, that's really important to have your agents do things effectively because it needs all of your context. And so that is automation number seven.

I've had this one running for a long time. All right, guys. So for the final automation that I highly recommend is one that is once a month.

And so I've actually been workshopping this. I've only ran it once last month, and it was incredibly good. And for this one, we are actually going to be using the highest model, 5 .6 sole extra high.

And this model, and really do deep analysis here. Write the most detailed summary ever. And I'm just going to say, don't do this now.

Schedule this for the last day of the month for the next 12 months. So this is basically review the past month across Notion, Gmail, Slack, Calendar, Whisperflow, Google Drive, Scrape Creators, and everything that it has access to. I have my entire company.

All of its context is within one of these plugins. And it is able to analyze all of it with the smartest model ever. And it can do this once a month and really get you to focus.

And I can tell it exactly where to put it. Create this report in Notion and then email it to Emily as well. And so this will happen automatically at the end of every month.

You can just deploy intelligence and this will happen in the cloud. And you can even create these automations directly from your phone. I could have done that whole thing from my phone.

You can just create these automations by talking to chat GPT work. All right, guys. So those are the eight automations or agent workflows that I've created that run automatically using GBT work, right?

We analyze emails and Slack messages for commitments and deadlines. We automated our hiring database. We're scraping Twitter.

We're drafting tweets. We're finding guests for our podcasts. We're doing email forwarding.

We're importing all of our texts. This is the only one where we use Codex so it can run locally on our computer. And then we're doing this giant.

mega monthly summary with all of our context. And here they are. You can see them in the desktop app, the web app, or the iOS app.

Thank you guys so much for watching. And OpenAI, thank you so much for sponsoring this video. I'll see you here for the next one.

The Hook

The bait, then the rug-pull.

ChatGPT's new Work mode turns the tool from something you prompt into something that runs your business while you're not looking, and Riley Brown spends this video wiring up eight of those automations live, one plugin connection at a time.

Frameworks

Named ideas worth stealing.

02:36concept

Plugins = Context + Actions

Context (read data from connected apps)
Actions (write back into connected apps)

The two reasons plugins make an agent useful: they let it pull real data from Notion, Gmail, Slack, and meeting notes, and they let it actually create and update records inside those same apps.

Steal forAny AI feature Joe builds that needs to reason over more than one connected data source.

00:00list

The Eight Agent Workflows

Commitments and deadlines
Hiring dashboard
Twitter scraping
Drafting tweets
Finding podcast guests
Email forwarding
Importing text messages
Monthly mega summary

Riley's full personal automation stack, in the order built in the video.

Steal forA starter menu of recurring admin tasks worth automating with an AI agent before inventing new ones from scratch.

CTA Breakdown

How they asked for the click.

VERBAL ASK

08:26link

“We're gonna be using Scrape Creators, in the description below I have a little guide that you can get for scraping social media.”

Soft resource-link CTA folded into the Twitter scraping segment itself, not a hard pitch at the end. The video closes instead with a sponsor thank-you to OpenAI rather than a product ask.

MENTIONED ON CAMERA