Resources:
Join the discussion: Ask Jordan questions on ChatGPT
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
Try Our Free AI Prompting Course: Register for our free Prime, Prompt, and Polish AI Course!
ChatGPT's New Tasks Mode: A Gimmick or a Leap Towards Agentic AI?
Big developments are on the horizon in the AI industry, with the introduction of the new “tasks” mode in ChatGPT. A feature that's changing the dynamics of automated tasks, it introduces possibilities beyond simple automation or reminders. However, it's worth noting that it's still in its beta phase and exhibits certain limitations, such as the inability to integrate GPTs directly within tasks effectively.
The Power of Context Tasking
An intriguing part of ChatGPT's tasks mode is the vital role of context tasking. Unlike typical AI solutions which may sometimes operate in a vacuum, context tasking allows for more personalized and precise inputs, thus enabling more nuanced outputs. The power of context tasking, however, does not absolve the need to manually input some information.
Automating Daily Routine with Task Stacking
The new feature doesn't stop at standalone tasks. It allows for what's termed as 'task stacking', turning individual tasks into "mini-agents". These mini-agents can tackle intelligent task sequences, where one task's result serves as input for the next. However, the process can be disrupted if initial tasks are delayed, which, in turn, disturbs the whole chain of task executions.
Managing Tasks: More Than Just Scheduling
This feature is not all about scheduling. It extends to more comprehensive task management, like naming and editing tasks. One unique aspect is that tasks are linked to specific chats, meaning deleting a chat automatically deletes associated tasks.
Enhancing Predictive Abilities with Future AI Models
There is buzz around potential enhancements and efficiencies that might come with the introduction of future AI models, like GPT-4. With these, the possibility of automating tasks management on the front-end becomes a reality.
The Agentic Aspect of AI
What separates ChatGPT's tasks mode from the run-of-the-mill AI solution is its semblance to agentic AI. The tasks scheduling feature, including task stacking and context tasks, loftily approximates autonomous actions. But remember, it's the concept of autonomous "actions" that might actually convey the capacity better than "tasks".
Overcoming Technical Limitations
As with any burgeoning technology, there are associated technical issues. Current limitations include the inability to upload files, and some users have reported task delay issues. Moreover, accessing this feature requires navigating to account settings, and not the ChatGPT settings. It’s also worth mentioning that, as of now, this feature only offers functionality for up to 10 tasks at once.
OpenAI's Edge in the AI Space
Keeping competitors at bay, OpenAI continues to dominate the large language model space. It remains ahead of the curve owing to its significant user base, and the addition of the tasks feature in ChatGPT validates its dominance in the AI industry.
With the incorporation of such agentic behaviors and task autonomy in AI, the discussions around the true power of AI are certainly diverting — from being mere tools towards becoming autonomous systems that are capable of functioning indepently. Undoubtedly, ChatGPT's new tasks mode seems poised to drive this shift.
Aim towards Task Automation
It's high time business, and decision-makers understand and implement ChatGPT's task automation. Repetitive tasks can be anticipated and automated, thus enhancing workflow through task stacking and reverse engineering. Businesses should act smartly when setting up these tasks, giving them autonomy to function within the parameters set by the AI.
Shaping Predictions for AI in 2025
An interesting five-part series is set to launch that focuses on AI predictions for 2025. With previous high-accuracy predictions as a testament, businesses and industry gurus should anticipate in-depth, non-random future predictions.
Conclusion
While debates around whether ChatGPT's new tasks mode is mere gimmick or a step towards more agentic AI hover, one thing is clear — the potential for advanced uses, especially in business contexts, is undeniable. Businesses that learn to navigate and utilize these features are expected to have a significant advantage over those who don't. Hence, the time is ripe to embrace the future of task management with AI.
Topics Covered in This Episode
1. Importance of Context Tasking
3. Task Management in ChatGPT
2. Utility of ChatGPT's Task Feature
3. Limitations and Challenges
4. Opportunities for Future Improvements
5. Agentic AI and Task Automation
Podcast Transcript
Jordan Wilson [00:00:17]:
OpenAI has just released a new feature called tasks. So on the surface, it might seem like something small. Okay. I can use ChatGPT to send me a reminder to go do something once. So while that's kind of true on the surface, I think that most people are missing the broader applications for this new task feature. But I am going to tackle today on if it is more of a gimmicky marketing feature or if this new ChatGPT tasks is something that has teeth, and is it maybe the first steps toward a more agentic AI? Right? Is this the baby steps to what we're going to maybe be using when and if OpenAI releases an agentic platform? So I'm excited to talk about that. I hope you are too. So welcome to everyday AI.
Jordan Wilson [00:01:18]:
If you're new here, we do this every weekday, Monday through Friday. Daily livestream podcast and free daily newsletter helping us all keep up with all this AI news and developments, but not just keep up, but how we can get ahead to grow our companies and to grow our careers. I want you to be the smartest person in AI at your company. So if that's what you're trying to do in 2025, this is your new home. So please go to your everydayai.com. Sign up for our free daily newsletter written by me, a human, every single day after our, podcast. Sometimes it's just me. Sometimes I interview guests.
Jordan Wilson [00:01:54]:
I go and recap that in our newsletter. So the best insights from, the day's conversation as well as everything else happening in the world of AI news, it is there. Also on our website, it is a free generative AI university. Go learn from 100 of the leading experts in AI. Whatever category you care about, it's all on our website. Alright. So, one other thing I gotta plug before we get started, our 2025 AI predictions and road map series starting in a couple of days, January 20th. It is a 5 part series.
Jordan Wilson [00:02:27]:
These are bangers, y'all. I I've spent a ridiculous amount of time, over 2024, writing these down, going back, redeveloping them. You do not want to miss it. I'm not kidding. Alright? It was one of our most popular shows last year, our AI prediction shows, and they were still holding true, like, a year later. You do not want to miss it. Trust me. Get your your company together.
Jordan Wilson [00:02:51]:
Get your board members, your your team. Listen to them all together. Alright. Before we get into today's topic on ChatGPT task and why I think this is actually a little more agentic than gimmicky feature, let's start as we do always by going over the AI news. Live stream audience, got a little secret question for you. Let me know in the comments. I'll probably, feature some of my favorites in our newsletter today. Alright.
Jordan Wilson [00:03:15]:
So first, Cisco Cisco Systems is launching a new AI defense to safeguard AI applications amid rising security threats. So sis Cisco has unveiled a new security solution called Cisco AI Defense designed to protect the development and deployment of artificial intelligence applications within enterprises. So Cisco AI Defense aims to provide a unified security layer across all AI applications, allowing businesses to pursue AI initiatives with enhanced confidence. The solution includes features such as AI application discovery, model validation, and runtime security, addressing the need for comprehensive security measures in the growing AI landscape. This solution is expected to be available in March and integrates with Cisco's Security Cloud using network enforcement points for AI security at the network level. Next, Apple is facing even more challenges with its rollout of, quote, unquote, Apple Intelligence. So according to some recent reports, Apple is facing a recent decline in iPhone sales, and it's underscoring a significant challenge as the company navigates a competitive landscape increasingly influenced by edge AI. Right? All these smartphone makers are putting very powerful, AI on their phones.
Jordan Wilson [00:04:37]:
Apple's trying to do the same. Apparently, not working very well because reports are showing the company's iPhone sales have dropped about 5% globally in the last quarter of the previous year, reflecting consumer hesitation amid underwhelming product upgrades. So while Apple has traditionally held a strong position in the smartphone market, it is now losing ground to competitors, particularly in China where local brands are rapidly advancing and integrating their own AI technologies. So the absence of Apple's new AI suite, Apple Intelligence in crucial markets. So, yeah, it's not rolling out everywhere. Even though where it's rolling out, I'm like, okay. Big deal. I'm not gonna use it too much.
Jordan Wilson [00:05:17]:
Right. But, yeah, it's not rolling out in places like China, and that has and that has limited, Apple's appeal in marketing effectiveness, especially as rival brands roll out their own AI solutions. Alright. A lot on those stories and more, but the biggest story in AI, at least today is this new update from OpenAI called Tasks. So, we're gonna get into it, and I'm gonna talk about if is this more of a gimmick? Right? I think a lot of people don't even understand. I always spend a lot of time, right, before I open my my big fat mouth. I spent a lot of time thinking about this, seeing what others, you know, even in at OpenAI, right? How they're marketing it, how others are reacting. So, yeah, I didn't sleep a lot because, number 1, I didn't have a show quite ready for today.
Jordan Wilson [00:06:09]:
Right? But I knew that there was supposed to be, a release coming on Wednesday or Thursday. Right? So I'm like, I'm gonna keep this slot open. So stayed up late, woke up early, pushed this thing through its paces. So I'm gonna show you why I think it's actually a little bit, more than a marketing gimmick. And I think, actually, even OpenAI isn't really positioning this very well. Right? And I'll get into that a little bit more later. And I think most people are completely missing the point of ChatGPT tasks. Right? One other thing, you know, maybe this is your first time listening.
Jordan Wilson [00:06:45]:
I don't know. Maybe you listen twice a year. I don't know. I spent a lot of time in ChatGPT. I'll just say that. I might be embarrassed to say how much, but I spend a lot of time in ChatGPT. Alright? On my phone, using all the different modes. I think I have I was counting yesterday.
Jordan Wilson [00:07:06]:
I think I have 5 different paid accounts. Right? So I have a free account. So, like, yesterday as an example, we did a ChatGPT free versus paid. So I have a free account. I have a ChatGPT plus. I have a couple of Teams accounts. I have about pro $200 a month account, enterprise account because we help, companies, you know, when they wanna onboard their, team or department or entire company onto ChatGPT teams or ChatGPT enterprise, they hire us because this is what I do. People are getting this wrong.
Jordan Wilson [00:07:37]:
Right? I'm looking. It's not hot take Tuesday, but I might come with some spicy takes anyways, if that's okay with you guys. Alright. So let's let's talk a little bit first about the details, how this works. Alright. And I'd love to know from our livestream audience. Thank you all for, joining. You know, we got Philip and Michael on big bogey face on YouTube and Manhan and Jackie and Brian here, on on LinkedIn and plenty of others.
Jordan Wilson [00:08:09]:
So thanks for joining. But, get your questions in now. I'm gonna try to tackle some of your questions at the end. Podcast audience, I always keep, my LinkedIn URL, my email in the show notes. So if you're listening to the podcast, check the show notes for this episode. It's all in there. Alright. So let's talk exactly what this is and how it works.
Jordan Wilson [00:08:27]:
So it's a little confusing at first. So number 1, you have to be on a paid plan. This is not on free plans, and it hasn't rolled out to enterprise plans yet. I don't know if it's going to. So you do have to be on a paid account. So even for me, weird. I told you, I have, 5 paid accounts. I only have this on, I think, 3 of them right now.
Jordan Wilson [00:08:51]:
So I'm guessing, you know, even if you don't have access to them, don't worry, to ChatGPT tasks, it'll probably pop up pretty soon. Normally, when you when there's something new in any model, you get like a pop up, like a literal pop up and it says, hey, this is new. I didn't get one for tasks, FYI, so you might not either. But here's the way it works. So, number 1, you have to be on a paid plan. Number 2, if you don't see it, try logging out, logging in, clearing your cache, cookies, all that good stuff, right? Or you know, if you're on, an app, try updating it, uninstalling, reinstalling if you really, really want it. But it is a new mode. Okay? So what that means, anytime you go into ChatGPT and you click, you know, new chat.
Jordan Wilson [00:09:33]:
Right? You're gonna have an even another option now. Right? So, anytime you want to schedule tasks, you go you need to go into the GPT 4 o with scheduled tasks mode. It's a mouthful. Right? And kind of the description there, it says ask ChatGPT to follow-up later. Okay? So any task that you need to schedule, you will go there. I'm gonna show you how to do this. We're gonna try it live. I don't think it's gonna work today live, FYI.
Jordan Wilson [00:10:05]:
But I I do have, an example we could go through because, I think OpenAI is getting a little bit slammed because when it first came out, like, you know, I was I was pretty early on it. It wasn't very, you know, trafficked, I guess. It it like, it wasn't getting hit very hard. Now it's getting hit pretty hard, because a lot of things aren't firing off in time, but that's that's normal. It is in beta. Keep that in mind. Okay. Then it's a little tricky to find these.
Jordan Wilson [00:10:30]:
So number 1, anytime you schedule something, you do it with natural language. You do it by quote unquote chatting with ChatGPT. But the way that you find it because you're not gonna find it in the settings. So you're gonna go over to the right hand corner if you're on desktop as an example. You're gonna go to your, not even your settings. It is the, essentially the drop down, when you click on your account name. So this isn't the ChatGPT settings. It's technically the account settings.
Jordan Wilson [00:10:59]:
So if you go into your ChatGPT settings, you're not gonna find this. Alright? But essentially, you will have your kind of task, library once you go and click that. So for our livestream audience, showing this on my screen. So first, tap left on my screen and tap left on your screen. Make sure you are in g b t four o with scheduled task. It's its own dedicated mode. After you schedule tasks, edit them. I'll go over that.
Jordan Wilson [00:11:23]:
You're going to go into, the task icon there. And then that is gonna pop up what then is on the bottom of the screen. Okay. And that is your task that you already have. So you can have one time task. You can have ongoing task, and it's gonna show you there from there. And that is kind of the task manager even though it's not labeled yet. You know, it's gonna say scheduled task, pause task, run task, etcetera.
Jordan Wilson [00:11:48]:
And like I said, you can go edit them there, change them, refine them, etcetera. Okay? And, yes, you can schedule them. You can have them run every single day. You can have the same task run multiple times a day, but you can only have 10 at once, which kind of stinks. Alright. I wish it was more. Here's a couple important things to know, before we dive in. Alright.
Jordan Wilson [00:12:17]:
So number 1, it doesn't work with every single mode. What do I mean by that? Well, I love projects. Right? So projects is an organizational structure where you can throw all of your chats in a project and at the project level, you know, you can give it custom instructions and, you know, throw some documents in there. So then all your new chats, you know, would have access to that overarching, you know, information. Can't use it. Alright. You can't use the canvas mode, which I like. So that part was a little disappointing.
Jordan Wilson [00:12:48]:
My 2 favorite, 2 of my favorite, kind of modes or, functionalities of ChatGPT in projects in canvas, you cannot use. Keep that in mind. That's important. Also, once you are using tasks, you can't be like, okay. Well, now I'm gonna use the o one, mode to logic and reason my way through these tasks. Can't do that either. Alright. It is its own dedicated mode.
Jordan Wilson [00:13:15]:
GPT 4 o. It's special variety. Alright. So also, like I said, it's only available on paid plans right now. You can edit and delete tasks, but but for now, you have a 10 task max, which I think is crazy, especially if you're paying the $200 a month for a pro plan. I would hope that you would have, like, as an example, I don't know, 10 x more the amount of, tasks. So, right now, even though I have a pro plan, I can only use 10. So I don't know how much I'm gonna be using this right away because I'm probably gonna be doing a lot of tutorials and things like this.
Jordan Wilson [00:13:52]:
And, you know, it's a lot of managing, but, I've already found some great use cases for it, which I'm gonna share. Couple other things. It's much more much more than about reminders. Alright? There's also I don't know why people are are saying, like, oh, you you know, this is you know, it syncs things up with your calendars. It's not like that. Alright? To use this to essentially remind you, like, go take the, you you know, take the trash out 9 AM. Can you do that? Yes. Right? And the other thing is, you know, there are some customizable, you know, notifications.
Jordan Wilson [00:14:27]:
So you can have a push notification if you're using the desktop app. You can have a push notification, on the browser, right, if if you were using it in the browser. You can have, email notifications as well. So I think from a marketing standpoint, I think, OpenAI maybe missed the ball a little bit. Right? Here's here's an example. So one of the first things that OpenAI put out, you know, in as an announcement, it was this. Remind me to work out every morning at 9:30 and pump me up with a motivational speech. K? I don't know.
Jordan Wilson [00:15:16]:
Not a huge fan of kind of the use cases and the marketing that OpenAI put out on this because I think it was confusing. I think people thought of it as just like, oh, okay. Well, I have reminders on my phone. Right? If you have, like, an iPhone, I have a reminders app. You know, I can do 2 thirds of, you know, kind of these, you know, use cases that Apple or or, sorry, that OpenAI seems to be sharing. Right? And people are thinking, oh, this is gonna help me, you know, sort my calendar. No. It's not.
Jordan Wilson [00:15:44]:
That's that's not what this is. It's not what this is. Alright? This isn't to remind you to do things 10 times a day. You can already put things in your calendar or on the reminders app. Right? If you have an iPhone, and I'm sure Android people, you have something like that too. But I think that's the majority of what people are thinking this is for, and that's what people are using it for. Even super smart people, that, you you know, I follow online, and they're like, okay. Well, allow me to do this.
Jordan Wilson [00:16:08]:
Right? No. Stop that. It doesn't read your calendar. It doesn't sync anything to your calendar, and don't use it to remind you to do something. Sipping my coffee. I'm not gonna say that's dumb. That is such a misuse of some of the most powerful technology that we have available. OpenAI's GPT 4 o.
Jordan Wilson [00:16:30]:
Even though the model itself is not the most powerful model in the world anymore, that crown has gone to at least for now, Gemini, their new Gemini, 2 experimental version. Right? But I think it is the most fully featured front end large language model in the world. Right? Because right now with Gemini, on the front end, you can't really use a lot. With ChatGPT, you can. So, yes, I did say right now you can't use all of the features and functionality when you are scheduling tasks, but I think people are missing the majority of it. Okay. So it's much more than reminders. Don't use it as reminders, please.
Jordan Wilson [00:17:08]:
Alright? Don't do some of the use cases that OpenAI and others are sharing because does it work? Yeah. Is it cool? Sure. Is it useful? Absolutely not. Right? I mean, maybe if you had a 1,000 tasks to work with or if tasks were unlimited, yeah, if you need a large language model to remind you to work out, sure. Right? But how about something useful? How about a business use case? Let's talk about that. Alright. And, also, I do think this is not a marketing gimmick, even though I don't think OpenAI did a good enough job controlling the narrative on how this can be properly used and what it should be used for. So I don't think OpenAI said the wrong thing.
Jordan Wilson [00:17:49]:
I just don't think they said the right thing because now the narrative has gone off in this different direction, and people are, I think, completely missing the power of this. This is extremely powerful to have ChatGPT go do task autonomously on your behalf. It is a precursor to what we will probably be seeing sometime soon from OpenAI, in their operator. Right? It's a precursor to that. These are the individual steps that I think is going to cause an agentic open AI to run. So even though it's not perfect now, even though it's in beta, even though it's a little buggy, you need to be using this. Because I do believe this is the foundational, work and the foundational knowledge that you need to get used to, to be ready for when agentic AI does come from OpenAI. Right? ChatGPT and OpenAI are the leaders in large language models.
Jordan Wilson [00:18:42]:
It's not even close. They have more users than the next 5 companies combined for front end large language models. So you need to be using this. You need to understand this. Right? This isn't some dorky thing. It's natural language scheduling actions with natural language. Right? It's so powerful, yet so simple. Alright.
Jordan Wilson [00:19:06]:
I don't know if that was actually a hot take, but I had to get that off my chest. I had to get that off my chest. Yeah. Samuel is saying, the feature feels like a 12 days of OpenAI runner-up. Yeah. It's probably better than all but, like, 3 of them, if I'm being honest. Right? Yes, Douglas. If you follow me on Twitter, which I'm extremely boring, Yeah.
Jordan Wilson [00:19:28]:
There is a new operator, subdomain for OpenAI. So we know it's coming out pretty soon. Danny, I like this question. Said, maybe Chad GPT should have called them task agents. I think something else aside from tasks would have been fine. I would have loved the word actions. Right? Because when people think of task, they think of reminders. Right? They think of, oh, I have to do something, and then when they combine that with these notifications, people think it's a simple reminder system, but it's not.
Jordan Wilson [00:19:55]:
I would have called it actions because you can schedule actions autonomously. Right? You don't have to go in and do them because you can have them work on a recurring basis. Alright? And ready, here are the 2 biggest tips. Alright. Go ahead. Write these down. I'm gonna try to show you an example. It's probably not gonna work, because it's a little slammed right now, although this was working yesterday.
Jordan Wilson [00:20:22]:
Go do these things. Go talk about them, right, at your next department wide meeting if if if your team uses ChatGPT. Right, if your company does, you're gonna be the smartest person. Steal my language. I don't care. Ready? Task stacking and context tasks. I don't know why OpenAI or no one else is talking about this. Right? It's the whole point of a large language model.
Jordan Wilson [00:20:47]:
The way that it works iteratively. Okay. So stacking tasks, I'm gonna show you an example. Schedule multiple tasks in the same chat. I'll show you what I mean by that. Because then you kind of have these autonomous actions working with each other. In context tax, context tasks, what I mean with that is people aren't talking about taking advantage of the context window. Right? People are just starting from scratch and saying go do this.
Jordan Wilson [00:21:16]:
Well, okay. Unfortunately, right now, in this new mode, you cannot upload files. Bummer. Right? Because wouldn't that be great? You could upload all your files and then go have ChatGPT run these actions on a regular basis on your behalf, keeping in mind all your files. But you can copy and paste all that information before you schedule those tasks. Why is no one talking about that? This is simple, like, large language model 101, y'all. Work with the context window. Put it to work for you.
Jordan Wilson [00:21:46]:
Right? Don't just start from blank and have it remind you to fold your laundry. Right? I have enough smart dryers beeping at me. Did did does anyone else is it just me? I have so many, like, if you've been following the show, you know, I I I hate all my appliances. And they all beep at me, and now they sing at me too. And I got some new appliances because all my other ones broke. And now they're all singing at me, and I don't know what I should do. Right? I'm like, is that the dryer? Is the dishwasher out of detergent? Did I leave the fridge open? Right? These are all things I I literally struggle with on a daily basis. Alright.
Jordan Wilson [00:22:21]:
So let's look live. Alright. Full disclosure, probably not gonna work very well. Alright. I warrant you. But here's here's what we're gonna try to do, and the reason why this was working fine yesterday, but I've been out for a couple of hours already today. Nothing's working well, but I'm gonna at least gonna show you how to schedule these and talk through, maybe a little bit of how you can use this as well. Alright.
Jordan Wilson [00:22:49]:
So let's go ahead and do this. Live stream audience, could you all do me a favor? Could you let me know when you can see my screen? That would be helpful. Alright? Alright. So, hopefully, hopefully y'all can see it now. Alright. So, this is my scheduled task. Alright. So I'm gonna go ahead I'm gonna go ahead and delete some of these.
Jordan Wilson [00:23:16]:
Alright. So again, when I go into ChatGPT, if I want to schedule a task, right now it says ChatGPT 4 point o. I need to click on that, and I need to go to, ChatGPT with scheduled tasks. And you'll see depending on how big your browser is, now it just says ChatGPT tasks. So, I cannot upload anything. Right? So this is grayed out, but you can use, this is actually wrong because you can't use the o one model right now because once you do, your task will no longer work. So once you start putting information into here, you can't use the o one model. Don't worry.
Jordan Wilson [00:23:52]:
I already tried. You can use DALL E. I don't know why you'd want to. DALL E is not good. You can search the web. That's the biggest thing that I'm telling people. Alright. And you can use standard voice mode.
Jordan Wilson [00:24:02]:
So you can talk to your text as well. Alright. Douglas says, I can see. I am in the browser right now. Okay? It's it's a little bit better, little bit better for me at least sharing my screen. So let's go ahead and try this. I have this open in a word doc, so you don't have to see me struggle to, type this. Okay? So now when you see things pasted in live stream audience, this is what's happening.
Jordan Wilson [00:24:30]:
So, right now my local time it is 7:55. So I'm gonna schedule this for 7. I'm gonna say 7:57. Alright. And then I'm gonna read this to you once the task is scheduled. So again, I'm putting in a natural language command. Okay. And one thing I'm gonna do so I just did this.
Jordan Wilson [00:24:51]:
It replied back to me right away. I'll read what this is, but I need to go in because unfortunately, it's going to truncate or shorten your instructions. And if your instructions are very important, you don't want to do that. So I'm gonna go back in, update those. Hopefully, I'm getting this in in time. Alright. And then I'm gonna go back and do my next one. Alright.
Jordan Wilson [00:25:14]:
Don't worry. I'm gonna explain this later, but what I'm essentially doing here is I'm doing some task stacking. Alright. And I don't think this is gonna work because like a busy airport on the holidays, I keep getting delayed messages. Right? So I'm gonna try my next one here. So I'm gonna go for, I'm gonna go for 6:58 AM. Alright. I'm gonna go ahead, paste this in.
Jordan Wilson [00:25:39]:
Don't worry y'all, I wanna do this live, but I'm going to go in edit this next task, right. And on the scheduling, I'll show you how all this works. I wanted to get this going first to see if we could get this to actually work. Alright. And then on my last task here, let me paste this in. I'm gonna do 7:59 AM. Alright. So I'm talking with ChatGPT.
Jordan Wilson [00:26:14]:
It's creating my task. I don't like that it shortens it because I'm being specific for a reason, ChatGPT. Alright. So now let me go ahead and read these. And if it works, which I, again, this is working fine yesterday, not working today because there's a lot of delays. This is in beta. Right? And there's that their servers are getting slammed because everyone's like, oh, you know, I need to be reminded to take out the laundry. Alright.
Jordan Wilson [00:26:41]:
So here's what I'm doing. So I'm gonna read my 3 different prompts that I just put in, and ChatGPT tasks turns those into tasks. So there's no, like, you don't start with a task manager. So it's a little confusing because you just have to put your prompt in in natural language and tell, tell this mode you have to be in that mode, okay, to make a task. So I'm gonna read them. There's 3 of them. We'll see if it actually works. Probably not.
Jordan Wilson [00:27:13]:
But I'll show you an example of when it does work. Okay. So I'm saying find the latest marketing news for today. Do not tell me trends or speak in generalities. Only focus on pinpoint specific updates on news, industry updates, and actual news happening regarding the marketing industry. Call these top 10 marketing news when replying back. Reply back with 10 bullet points of the biggest marketing related news stories. Be extremely specific yet do not waste words.
Jordan Wilson [00:27:45]:
Please research at least 10 different websites with ChatGPTcsearch and or browser Bing. Please do this every day at 7:57 AM. Alright. So now I can go in here, and you can right click on any task. Because when you do that, ChatGPT doesn't go out and go complete that action, right? It just makes a task. So then I can click on the little three buttons there and I can edit the task, I can pause the task, or I can view all tasks. So let's go to edit, and this is what, the editing view. So I wish that there was a new task button, and would just bring this up because I think people are gonna get confused, right, but that's fine.
Jordan Wilson [00:28:27]:
So it gives your task a name by default. It gives it instructions. So as you kind of saw or heard, it made my instructions much shorter, which yo, Chad GPT. I know what I'm doing, homie. Alright. Don't don't make my task, instructions shorter, because it's gonna leave out some important details, which you might really need. Alright. So essentially, it gives your task a name, instructions, and then depending on what you tell it to do, it's going to schedule it already, which is really cool.
Jordan Wilson [00:28:57]:
So at any time you can go in and rename your task. You can update the instructions if something's not working correctly, and then the schedule. There's a drop down schedule. So you can do doesn't repeat, daily, weekly, monthly, annually, or custom. Alright. You can change the time. You can pause, delete it, etcetera. Also at that same time, you can click any task within your chat and go to view all tasks.
Jordan Wilson [00:29:22]:
Alright. So there's all of my tasks that I have. I have to keep deleting them, because you can only have 10 at a time. Alright. So now let's go back. And you'll see already, it says delayed. It says they're all delayed. So that's a bummer.
Jordan Wilson [00:29:38]:
I figured this might happen. But let me talk about the concept of what we're doing here. So this is an example of text stacking. What do I mean by that? You have, like I said, baby steps of agentic AI out there for you right now. Okay? Don't think of these as one off calendar reminders. Stop doing that. Think of them, each task, as baby agents that can go out, perform actions on your behalf. Right? It can search the web.
Jordan Wilson [00:30:19]:
It can work with the context that you give it. And then it can come back and report to itself. That's the thing. You do not have to be the one orchestrating all of these agents to work together. One other important thing I should mention about tasks, they live in the chat in which they were created. That's important. Let me repeat that. Any task that you make, it lives in the chat in which it was created.
Jordan Wilson [00:30:49]:
Alright. So as an example, if I delete this chat, the tasks are gone. So you might think of that as a bug. Sure. I think of it as a feature. And hopefully, tasks in the future will be more like GPTs, and you can pull on them and use them anywhere. But if you are smart with how you set up your chats, which if you've ever taken our, you know, free prime prompt polish course, you are. I'm sorry y'all.
Jordan Wilson [00:31:15]:
I haven't done those in a couple of weeks. I know a lot of you are reaching out. I'll explain why at some point here I haven't been able to do them recently. Okay? But we will have it back on. Don't worry. But you have to be smart with how you set up your chats and your tasks there within. Because if you do this correctly, and if in this case, right, I just set up 3 different tasks all in the same chat. I didn't go into a new chat each time because I want these little mini agents to go out, perform an action, come back, and report.
Jordan Wilson [00:31:48]:
So, you know, in this example, task 1 is doing something, and it's putting it in the context window. And if it was working correctly, task 2 would pull that context for its own task. Right? So at that point, instead of it going human task 1, human task 2, human, task 3. Instead, it's human, task 1, task 2, task 3, human. So human is still in the loop, but instead of me inserting myself in between each of these tasks, I'm essentially telling through some I won't say impressive prompt engineering, but through some smart prompting for the chat to use the context window. And I know you're probably saying, oh, Jordan. It doesn't work. It works.
Jordan Wilson [00:32:40]:
It's just delayed right now. So I went in and refreshed it. Some of these some of these aren't working. Some of them are. So, yeah, this this one's really not gonna, do do do very well here. So let me tell you what the task 2 and task 3 were. So the first one I'm essentially saying, yo, at 7:57, go find all the marketing news. And then I should have spaced this out a little bit, but I'm trying to jam it all into a live stream, but I'll show you a working example.
Jordan Wilson [00:33:10]:
And then I say every day at 7:58 AM, please look at the context of this chat and choose 3 of the above stories from top 10 marketing news. Right? Because I told task 1, I said label it as such. So then I can tell task 2 to go look at that. So I'm saying go look at the top 10 marketing news that are most relevant to the everyday AI show by Jordan Wilson. So it's also supposed to be using ChatGPTsearch, but in other chats, I've already given it all this context. Then I'm saying use ChatGPT search to find information from 2025. Don't leave ChatGPT search up to its own devices. It's gonna pull in old stuff that's not relevant.
Jordan Wilson [00:33:52]:
Always always be pinpoint specific. So then I'm saying use ChatGPD search to find information from 2025 for each of those 3 stories above. So I'm saying out of the 10, that task 1, go pick 3 that are most relevant to me. So I'm giving it autonomy. I'm giving it agency, and I'm saying, hey. Go see what this other task did, this other mini agent. I want you to use what it brought you, and I want you to build off of it, and then I'm giving it a more task. I'm saying, you know, give me a very simple yet actionable bullet point tip or trick on how to put those stories into action to grow my business.
Jordan Wilson [00:34:29]:
Be extremely specific. Avoid generalities. Then I'm saying, call these three actionable steps when replying back. Obviously, our first chat or task was delayed, so this chain reaction didn't work very well. And then the last task I'm saying, every day at 7:59, please look at the actionable steps and see what other companies in 2025 have implemented something similar or find additional resources resources that can help me implement these strategies for everyday AI. Again, be pinpoint specific. So do you see what I'm doing here y'all? This one is an example of task stacking. Alright? I am starting to build within ChatGPT, which, like I said, I think GPT 4 0 is the most powerful model in the world right now, because of its other functions and capabilities on the front end.
Jordan Wilson [00:35:25]:
Right? Not on the back end. If you wanna go into Google's AI studio, yeah, Google Gemini, their new 2 point o experimental technically better. But on the front end chat, this is better. Alright. So this one didn't work. This one didn't work. Right? Because the, the first task was delayed, so therefore the second and the third task couldn't pull the context. But I said before I did this, I'm like, I know it's gonna be delayed, but it was working for me last night, and it probably will start working for me probably in, like, 2 to 3 days.
Jordan Wilson [00:35:54]:
So think of what you can do already. Think of all those repetitive tasks that you do. Right? I need you to break down and reverse engineer your buckets, your buckets of work. What are those things you do every single day? I know this is weird. I recommend people use like a task manager, Right? That literally just looks at everything you do. I know it's creepy big brother. Right? But then if you do that for a week or 2, you can export all that data. You can throw it into something like ChatGPT o 1, and then have it start building these reverse flows for you where you can then go set out, you know, all of these little task agents.
Jordan Wilson [00:36:29]:
And, again, this is the worst it's ever gonna be. And, again, I think OpenAI really missed the ball in terms of talking about the capabilities. Alright. Let me quickly show you, hopefully, an example of when this worked. Alright. Here we go. So in this example, I'm doing some context. Alright.
Jordan Wilson [00:36:52]:
Context tasks. So before I gave ChatGPT its tasks, here's what I did. I literally said, do not create a task yet. Instead, just find these trends in the data. Alright. So I gave it all of my let me switch this here. I gave it all of my episodes. All you know, we have how many episodes do we have? 300? 400? My gosh.
Jordan Wilson [00:37:24]:
437 episodes. Alright. I give it all my episodes, downloads, dates, all of that stuff for my podcast. I said do not create a task yet. Instead, just find trends in this data. So you can still use this mode of ChatGPT to chat with. Just say, like, don't create a task, homie. So then after this, you know, ChatGPT went.
Jordan Wilson [00:37:46]:
It found some, you know, patterns and, you know, in my data blah blah blah. Then I said, here's some more information about me essentially. Right? Generally, I I don't use custom instructions. Not smart. I don't use ChatGPT memory. Not smart. I use projects. Smart.
Jordan Wilson [00:38:03]:
Right? Because then you can have custom instructions for certain chats that aren't applied unilaterally to everything that you do inside of ChatGPT. But because I can't use projects with this new ChatGPT task, I do have to unfortunately manually feed some of this context in there so it knows what I'm doing. But you need to be context tasking. Right? So I'm giving it. I'm saying now I gave you all this information, all this all these data and trends about everyday AI and what I'm working on, and then I gave it some more, information about the ChatGPT show or or sorry, the, the, everyday AI podcast. And I actually just to be quick, I just use deep research. I'm like, I don't wanna type this all or copy and paste it from my website. Alright.
Jordan Wilson [00:38:42]:
So I pasted it in there. Then after all of that, you know, I had it analyzed everything. Then I said, every day at 6:52 AM, please use ChatGPTsearch to deeply research the latest in AI news for today. It's for the everyday AI show, so please keep in mind all of the info in this chat's context window to better understand the types of topics covered. For today's AI news, please do not tell me trends or speak in generalities. Only focus on pinpoint specific updates in the new on newest AI features, model updates, and actual news stories that are gaining traction in the AI industry. Reply back with 10 bullet points of the biggest AI news stories. Be extremely specific yet do not waste words.
Jordan Wilson [00:39:29]:
The bullet point should be factual and succinct. Please research at least 10 different websites with chat, GPT search, and or browse with Bing. Please do this every day at 652. And then it created that task. It didn't you know, it truncated it. It made it smaller and shorter, so I had to go in there and edit it because I that's the other thing too. Right? Chat tbt right now, this is beta. I wish it wouldn't even, you know, try to modify your task instructions, but I get some people aren't gonna do a very good job and it will actually improve it.
Jordan Wilson [00:40:03]:
For people like me, like, I know what I'm doing. So now I have to go in and manually change the task instructions each time, but then you'll see. It said, got it. I'll get back to you with the latest news at 6:52 each morning. And then this morning, it was working a little better, a little earlier. And then it gave me 10 AI news stories that are just, from today. These aren't necessarily old, and it's based on the topics that we already cover. So I gave it context.
Jordan Wilson [00:40:30]:
Alright. So, I hope this is helpful. And let me see. What questions do you all have? Alright. I'm gonna I'm gonna quickly, wrap this show up. I I I said I would answer questions. It's sometimes hard, if I'm being honest doing this on the, on the livestream to look at questions. I wish y'all could, like, mark it as question because otherwise, I have to go through and do this.
Jordan Wilson [00:40:55]:
Yeah. Samuel says, on weekdays at 7:28 AM, remind me to watch Jordan live. Love that. Thank you. Big bogey. Yes. So he's saying this is kind of if if this, then that. So, yeah, there's a very popular and and and that's how I'm testing it as well.
Jordan Wilson [00:41:11]:
Doing some conditional logic. Right? Like, hey, task. Go look up this. If this, then do this. If that, then do that. Like an easy example that I was testing out, and it worked fairly well. You know, I gave it 10 different stocks every morning. Go check the price of for each one, I said, if it's over this price point, tell me why.
Jordan Wilson [00:41:33]:
If it's under this price point, tell me why. Do research. Give me 5 bullet points. Right? But, you know, pretty wide ranges. So if any you know, as an example, if if if NVIDIA's stock pops or drops or if Apple's pops or drops or Google's, right, I I just wanna know. I don't wanna have to monitor those things. Right? But if it's just, you know, if it's upper, you know, if it's up or down, you know, 1 or 2%, I don't care as much. Right? But if there's something big that's happening, right, if NVIDIA shoots up 5% or if Apple shoots down 3%, I mean, we're talking tens of 1,000,000,000 or 100 of 1,000,000,000 of dollars shifting.
Jordan Wilson [00:42:07]:
Right? And that's probably gonna impact how we all work or it may. Right? So, yeah, that's that's another good example of how you can think of this is think of these as conditional logic flows or if this then that flows. Because think of what you do every day. This is what you do in your own brain. Right? You're researching things for your company and, you know, your competitors, and you're saying, oh, well, if they, you know, if they launch a new project. Right? If they if they update their services, right, then then I need to do this. Okay. Well, see how you can start building that into ChatGPT task.
Jordan Wilson [00:42:39]:
They're not great right now, but you gotta take these baby steps along with the technology. So when it's running, you can keep up. Because if you don't crawl, when OpenAI and ChatGPT is crawling, you're gonna get left behind when everyone else has the capabilities to run. That's what we do every single day. We help you keep up with the technology where it's at, because it's not perfect right now. Alright. Hopefully, I didn't miss any questions. If I did, I'll go back.
Jordan Wilson [00:43:10]:
Oh, great question, from YouTube. 1 a Lee. Can you use GPTs within the tasks? So yes and no. I've gotten it to work twice. I don't know if that was a bug. I don't think it's supposed to work, but I got it to work twice. But I'll say right now, no. In the future, maybe it will.
Jordan Wilson [00:43:31]:
I would love for it to work with GPTs, with projects, with with Canvas. Right now, it doesn't. With o one, you you know, the the the reasoning model. Right now, it doesn't. Alright. Alright. Let's see. Is this how you, Jordan, will be planning your Monday shows? Yes and no.
Jordan Wilson [00:43:53]:
Right? I already do these things a lot. I've already built a lot of these type of flows. So, yeah, I I I will be using it. I won't be using it exclusively. Right? But for the most part, yeah, like on Mondays, I do the, you know, AI news that matters, but I use perplexity. I use ChatGPT search. I use Google Gemini, Deep Research a little bit. Right.
Jordan Wilson [00:44:17]:
But, yeah, I I will definitely and I have already set up some, task flows that I'm going to be using on a daily basis. Yeah. So Douglas Douglas also asking can you use an at to tag a gbt in a task? I tried. So the problem is when you do that, it's a little different in task mode. Let's see. Should I should I try to do this? I'll I'll try to do it. I'll just show you all an example since multiple people are asking. So I'm gonna go into check GPT task mode.
Jordan Wilson [00:44:49]:
Alright. And I'm gonna say, you know, at, you know, 8:17 AM every day use, and then I'm gonna put the at. I'm gonna say use, this to recap an episode. Right? Alright. So all I did was at mention a GPT, but I'm trying to use tasks, and you'll see it doesn't technically work. Yeah. So when I try to do it, the GPT takes over from the task mode, but I have gotten it to work. So I I did an example, you know, I have a custom GPT, and it it's it's getting confused.
Jordan Wilson [00:45:28]:
Right? So the task generating ability of this mode is getting confused because the GPT is taking over. But I did get it to work once, but it took a while and it's not working consistently. So I'll just say right now, no, not consistently. You can't use GPTs. Once you can, that's gonna be extremely fun. Alright. So, I hope this was helpful y'all. As a reminder, Monday, Monday, Monday.
Jordan Wilson [00:45:55]:
Be there. Be there. Be there. Alright. Next week in a couple of days, we are doing our 2025 AI predictions and road map series. I'm lucky enough each year to talk to 100 with an s. So both interviews I do live on the Everyday AI Show. You know, companies that I meet with, a lot of people just wanna, you know, quote, unquote, pick my brain, right from big companies.
Jordan Wilson [00:46:19]:
I gobble up all this information and I steal it from the smartest people in the world. And I have a a couple files that I store all this information, and then I have all my, you know, AI tools help me, you know, make sense of all my predictions and my thoughts. But this is a long time in the works. You are not gonna wanna miss our 5 part series starting January 20th. Not just random hot take predictions, but laying out the road map for you to take advantage of what I know. I'm gonna give myself I'm projecting. I got a little spicier this year than last year. So I'm gonna say I'm gonna be about 80, 85 percent right.
Jordan Wilson [00:46:57]:
I think last year in my 24 predictions, I think 21 of them were technically right, and the other ones were, like, maybe a little too ambiguous. Right? But you do not want to miss starting January 20th, 5 part series. So, please go join us then. Also, if this show was helpful, please click that repost. So I actually have 10 business use cases, that I already built out for kind of this, task stacking. I knew I wouldn't have enough time to go over them today, and our newsletter's already long enough. Alright. So, I already have it made, so, you know, you're not gonna have to wait.
Jordan Wilson [00:47:32]:
Like, sometimes I make you wait like a week. I'm like, oh, you know, go repost this, and I'll give you this. And I'm like, oh, frick. I didn't finish it. Alright. This is done. So if this was helpful, please, on LinkedIn, go click that repost button. So if you're still listening on the podcast, thank you.
Jordan Wilson [00:47:45]:
Appreciate your support. If you wanna see the 10 business use cases that we already built out, got them ready to go. Just click the show notes, for this. I always put a link, to the, LinkedIn, stream for this. So go find that. Go click the repost. I'll send you, 10 business use cases that take advantage of this tech stacking concept. Alright.
Jordan Wilson [00:48:09]:
Thank you for tuning in. Please go to your everydayai.com, and please join us tomorrow and everyday for more everyday AI. Thanks, y'all.