← Everyday AI Podcast – An AI and ChatGPT Podcast
Everyday AI Podcast – An AI and ChatGPT Podcast
Ep 771: ChatGPT Workspace Agents: How to Use OpenAI’s Most Overlooked New Feature
00:40:30
Loading summary
Transcript3 lines
- [00:01]
A
This is the Everyday AI show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business and everyday life.
- [00:16]
B
OpenAI releases dozens of products and model updates each year, and I think their recent release of workspace agents 2 weeks ago May have been the company's most overlooked meaningful update, well, ever. Why is that? Because now teams have a simple drag and drop agent builder that connects to all the apps you use. It can be shared across your team, you can talk to it in Slack like an employee, and it has read and write capabilities to get work done. And maybe even better than a human employee, you can actually go back and observe and audit each step of the Workspace Agent's work work. Chances are, though, you probably didn't hear much about this. That's because the Workspace Agents release was kind of sandwiched between two of OpenAI's more viral or hyped up announcements of the year during the same week, which was GPT 5.5 and their Images 2 model. So unfortunately a lot of the normal attention that would normally come with an update like Workspace Agents kind of fizzled and faded away. But fortunately I'm going to be recapping all of the agent action that you might have missed on today's episode of Everyday AI, because we are putting AI to work on Wednesdays in our weekly Wednesday demo series. So here is the big picture. Well, the big picture is you might have missed this and you should have been paying attention because Workspace Agents, like I said, got released. It was one of OpenAI's biggest release weeks ever. I would say. Actually it was their second or third biggest release week ever since the company started, you know, with chat GBT in November 2022. So if you missed it, I don't blame you. Because between GPT 5.5 and images 2, you might look at something like the new Workspace Agents and say, well, this wasn't, you know, even OpenAI's biggest release of the week, but I think it actually may have been. So these new agents are powered by Codex. That part's important, and more on that later. And they can be shared across teams used in Slack, and they have complete observability. And there are pros and cons to using these new Workspace agents versus setting up something similar in Codex or even using a normal GPT that you might normally use. Also, one other thing to note, well, why it's noteworthy for today at least, is pricing is set to change today. Yeah, everyone kind of had A if you were on a paid teams plan, you kind of had unlimited use and unlimited access. But OpenAI did say that starting today they're going to be shifting to a usage based pricing system. So we're going to have more on that in today's newsletter. But maybe today will be the day because of that price change that a lot of people are maybe hearing about ChatGPT's agents for the first time. So on today's show, here's what we're going to do. Here's what we're going to go over live. We're going to show you how to build a workspace agent from scratch and understand their capabilities. You're going to know the pros and cons of using workspace agents versus setting something up in Codex and even versus the kind of quote unquote older or old school GPTs. And then I'm going to show you some of the best practices on how your team can start using workspace agents today. All right, let's go. Welcome. This is Everyday AI. My name is Jordan Wilson. If you're new here, this is a daily, unedited, unscripted live stream podcast and a free daily newsletter helping everyday business leaders like you and me keep up with all the AI advancements. Because like this workspace agents, you might miss some. I tell you what's important, how to use it to grow your company and your career. So if that's what you're trying to do, yeah, it starts here. But make sure you go to our website at your everyday AI dot com. Aside from there being an entire library of free gen AI resources, you can go listen to 750 of our back episodes, read about them, watch the videos, all of that. But make sure you go sign up for our free daily newsletter. We're going to be recapping the highlights and what you need to know from today's show in a more digestible form, as well as all of the other AI news and updates that you need to know to stay ahead. All right, so let's talk workspace agents. You probably didn't hear a lot about them and I already told you some of the reasons why. Yes, they were sandwiched between OpenAI's, some of their biggest releases. But the other reason why is, well, they're only available for teams and I think that a lot of the maybe, quote unquote hype or attention online is sometimes driven by individual power users. Right? So maybe those people with a $200 pro account or, you know, those on a normal paid plan. So unfortunately A lot of these capabilities, although you can kind of replicate some of the basics in Codex if you do have an individual paid plan. But for the most part, I think that's one of the reasons why we haven't heard more about these because they are right now only for business and enterprise customers. So you do have to be on a team plan to take advantage of them. And I guess that's why they're called Workspace agents. And if you've used Agent mode before in Chat GPT and you're looking at your paid plan and you're like, oh well this says agent, that must be what it is. No, Agent mode is a older kind of capability inside of Chat gbt, originally called Operator, it moved over into Agent mode. So if you're looking at that, you're like, no, that's not it. Workspace agents are extremely powerful. So think of all of these manual processes that you do over and over, all of the different apps that you use. Maybe you're using Skills, right? Obviously Skills have wide support now in certain ChatGPT plans. In Gemini plans, Microsoft has rolled out supports, obviously Claude with Skills, you can actually use Skills and bundle together Skills. All of the connected apps that you use as well as how you would normally use Chat GPT and it's all powered by Codex. Right? So that's kind of Workspace agents sandwich up into a little package. But we're going to do it different this week. We're going to go straight live. All right, we're going to start our demonstration and maybe by the end we might have an agent fully working. If not, I already do have this one pre built to where we can look at the the responses because you know what could ever go wrong doing Building an agent. Absolutely live. Right? All right, so here we go. All right. I'm actually going to show everyone how to do this kind of step by step. So live stream audience, if you could do me a favor, let me know if you can see my screen. Podcast audience appreciate you as always. But if you want the video version of this, make sure to go to our website. Your everyday AI.com you can watch this video and all of our other Wednesday tutorials. So there is a new section on your sidebar. You're going to go into agents. You're going to see this on the left hand side. So I actually am on a cleaner account. I have another teams account where I'm doing a lot of the majority of my actual use. This is one of my team plans that I use more for testing. But where you would normally have your chats and the different modes like Deep Research and Apps, all those things. On the left hand sidebar, you're going to see a new section called Agents. All right, so make sure you look for that. There's a little plus button that you can click on there. You can click on Browse Agents and then you can go in, you can see your recently used agents, agents that were built by you, and then a directory of agents. So you can share agents. And the agents that are shared will be accessible there in that place under the directory. So we're going to get a little bit more into the governance side, the sharing and the memory, because I think that's all very powerful, but that's the basics. So it is kind of similarly set up to GPTs, how your company can have a GPT store, you could build GPTs for yourself. The same thing with workspace agents, even though they are made for only teams, you obviously can have agents that only work, you know, with your data that aren't available to everyone else. So there is a sharing element as well. All right, so we're going to go straight in and we're just going to create an agent. All right, I have my prop kind of typed up here and we're going to watch this live. I'm going to explain it to you, how it all works. So let's go. I have a very, a very simple prompt here. So there's different ways that you can build these. You can build these conversationally, you can start with a template, you can build them from scratch, but I'm just building it conversationally right now. So all I said, I said I want the agent to go into my Beehive account over the past five days. So, yeah, we do a daily newsletter. We use Beehive. That is our platform. Beehive has an MCP server. So this is fast, right? So it's already kind of done? Not actually, but it's already outlined my agent. So before I could even finish reading the prompt. So here's what my prompt is. The agent builder is kind of ready for me to start approving. So it says I want the agent to go into my Beehive account over the past five days and give me a list of the five most clicked stories each day over the past five days. So 25 in total. Then I want a pull out of anything that might be considered a new AI feature or an LLM upgrade that can be used immediately. I want this to run every day at 7am and email me the findings. So this is just an example, but this is a process that I go through every single week, right? So we have our new show on Fridays where we go over, you know, we call it the Friday Features Show. We go over seven of the most useful AI updates that you can actually use use. So a little different than our Monday news story because a lot of that now is caught up in drama and, you know, you know, funding and all these other things. So Fridays are practical AI updates and LLM upgrades that you can use. So the problem is, is there's obviously way more than seven. So this is one thing that I would normally do manually. And I want you to put yourself in my shoes, right? What are those tasks that you have to do manually over and over that you're like, this is silly for me, this is one of them. But I always like to back, you know, my hunches up with data. So that's why I normally do this manually. Before I had my workspace agent is I would go into look into each email newsletter that we would send out to see what are the most clicked stories because I might think AI Feature B is the popular thing that most people care about. But if no one's clicking or caring about feature B and it, oh, turns out it's feature circumstances, that's all database evidence that I can go into my daily newsletter each and every day and look at. The problem is, is that takes a lot of clicking, that takes a lot of manual effort. And let's be honest, I think one of the biggest upsides of using things like workspace agents or any automated AI workflows is, well, you take out that the likelihood of you getting distracted, right? Because for me, I would probably go in here, I would see something in, in our email and I would say, oh, a ton of people clicked on that. Let me go start planning an episode on that for tomorrow or for next week, right? That's what would happen to me. So not only are workspace agents great at the context carry, right? So carrying this context between different apps that you can set up, but also just, well, you aren't going to get distracted. All right, so that is my prompt. The agent has already built a plan for me to approve. And you'll see here I didn't even need to tell, which is really good. The agent builder that the Beehive is actually a custom MCP that I already connected, so I didn't have to choose everything conversationally, but it already decided on its own that it needs to use the Beehive MCP connector that I have already previously set up and my Gmail account. So all I said is email me something, but it was smart enough to look through. It knew I had the be the the Beehive custom mcp, I can't talk this morning and it knew that I had already connected my Gmail. So that's a great benefit of just being able to talk about your goals conversationally. You don't even have to know or remember necessarily what you have hooked up in your account. All right, so it gave me a plan. So here's the plan. It says this agent will review your Beehive data from the previous five days, rank the five most clicked stories for each day, and highlight new AI features so I can ask for edits. So one thing, and I know this from the testing, one thing I did a little bit wrong is I need to clarify the last five episodes, not the last five days because the last five days would include weekends. So I'm going to say don't include weekends. I'm going to say go back to the last five newsletter posts that have been published regardless of day. All right, so if you don't like the plan, just like I showed you, it's going to go ahead and modify that. So I sent that and now the Agent builder split out into two kind of panes. So if you are familiar with the custom GPT builder in chat GPT you'll be familiar with this setup. But essentially you can talk conversationally on the left hand side for edits in natural language, you can upload files, you can click the app button for context, bringing in tools like your memory, Web search, images, etc. Or connect it to other apps. So if I do in the future want to add another app to this, I can do that fairly easily in a conversational editor. And then on the right hand side, as the Agent builder builds the workspace agent, it is going to be previewed here on the right hand side. So now my next step. It says one important setup is still blocking. It says for the daily 7am email, should the agent send it automatically from your Gmail or pause for approval. So I'm just going to say send it automatically. AI moves too fast to follow, but you're expected to keep up. Otherwise your career or company might lag behind while AI native competitors leap ahead. But you don't have 10 hours a day to understand it all. That's what I do for you. But after 700 plus episodes of everyday AI, the most common questions I get is where do I start? That's why we created the Start Here series, an ongoing podcast series of more than a dozen episodes you can listen to in order. It covers the AI basics for beginners and sharpens the skills of AI champions pushing their companies forward. In the ongoing series, we explain complex trends in simple language that you can turn into action. There's three ways to jump in. Number one, go scroll back to the first one in episode 691. Number two, tap the link in your show notes at any time for the Start Here series. Or you can just go to starthereseries.com which also gives you free access to our inner circle community where you can connect with other business leaders doing the same. The Start Here series will slow down the pace of AI so you can get ahead. Again. This is like having a super smart developer that's building you an agent that can run on a schedule, it can run on a trigger, and you just have to sit there and say, go do my work. You just have to think of redesigning your workflow. I think that's where most teams are going to find. The biggest benefit from using workspace agents is just reimagining how work can work. All right, so now the cool thing, right? So if you have used Codex like me, you can actually kind of watch this agent work, right? So it says it's thinking, there's kind of a cursor, you know, floating around. I can see what it's doing. So we're going to give this a minute to cook, right, Quite literally, and then we're going to check back in on it here in a little bit. But we're going to go over some of the other features that we need to know. So let's make sure we cover the basics. What are these workspace agents? They are much more than talking in chat GBT because they can follow your instructions, they can use all of your connected tools, whether they're apps that you have Already connected, that OpenAI provides support for custom MCPS that you can set up. That's model context protocol. So many of the most popular apps and softwares today have MCP connections. So it's not just for the apps that OpenAI has act, you know, has official compatibility for, it's for your own apps. And here's one of the big benefits. They run in the cloud, right? So you've probably heard a lot of talk, even myself, right? Codex, I have codecs running around the clock, right? I have all these automations. Same thing with cloud code, right? Cloud code and cloud co work. But for all of those things, you have to have number one, your machine running, number two, you have to have the apps open. So that's a big benefit of the workspace agents and is they run in the cloud and then the other big one is you can talk to it and command it in Slack just like you would an employee. So if your team is a big Slack user, that's going to be a big benefit. So you don't even have to go into chat GPT browse through your agents, right? If it's not scheduled because you can have it run every single day or you can trigger it manually, right? So when a new, you know, customer support ticket comes in, maybe you want to look through your CRM, maybe you need to look through Google sheets and then maybe you need to, you know, fill out a customer spots. So if your team does run in Slack and you have a Slack channel where a new, you know, customer inquiry comes in, or a customer support ticket as an example, you could fire, fire off your workspace agent or it could obviously do that automatically based on the trigger of receiving, you know, any of that information. So these triggers can be automatically or you can talk and kind of get the workspace agent running inside Slack. But essentially this replaces the older paradigm of, well, you just having to go in and give chat GBT a prompt and then have it go do work. So the difference here is there's a trigger. It can be scheduled, you line up the process, it uses the different tools that it has access to. It also has persistent memory which we'll talk about here in a minute. Then you get an output and you can review it. So this is best for repeatable multi step workflows, not one off chats. So your one off chats, you know, maybe you'll still use GPTs for that. That's a big difference between workspace agents and GPTs. Well, number one, the workspace agents have read write ability and by default GPTs can't actually write. They can't go into your Gmail and send an email, they can't create a calendar event, right? Some you can do draft versions, but you can't actually autonomously do that actual work. And that's where workspace agents actually set themselves apart. Next, here's kind of the five parts that make it useful. So you think of your workspace agent as having a role or a job. In my very specific instance here, all it's doing is it's doing some basic research for me, going into my newsletter, clicking through, looking at the performance and then also thinking it has a layer of thinking. Because not all of these top click stories each and every day are going to be actual features that you can use today. So it's doing a level of synthesizing thinking and then also a level of research. So anything that you can normally do inside of Chad gbt, this workspace agent can do, so you give it a role or a job. Then there's a trigger, whether that's a manual trigger, a scheduled trigger, or inside of Slack, then there's different steps in a process. It has tools that's kind of the system that it works within. And then there's also guardrails on the back end for observability and traceability. And the big thing here is skills. So when you're building this and I'll show you some of the more customizable features inside of these workspace agents, any skill that you have in Claude, in Copilot, you know, now even Google Chrome supports skills, right? Kind of an open protocol of, you know, essentially a set of markdown files. You can import a skill, so if you export it from anywhere else, it's going to give you a zip file. So in the same way that I conversationally gave the agent builder a prompt and said, I want to do all these things, then it was smart enough to connect it to my current apps. You can just upload a zip file and say, hey, this is a skill I was using in Quad, right? But maybe it couldn't do everything you wanted it to. And then it will use all of that skill, it will import it and then connect all of the different apps. So what can they do? The biggest thing is repeatable work. Because if I'm being honest, if this is a process you do once a year, it's not probably worth building a workspace agent on it right now. These are for those tasks, the manual tasks that you are having to be the human duct tape. You are having to do the context carry, right? Going in, checking, you know, you get an email, there's a new entry from your website, you're going into Google sheets, you're looking at this person, then you're going into your CRM, you're seeing what's the latest, you're doing some manual research on the company, then maybe you're updating a proposal that you have saved in, you know, Canva or Figma or Gamma, whatever it may be, and then you're uploading that and you know, sending it to your team for approval in Slack, right? That might be a two, three, four hour process that I just described that Entire thing, you can go to either draft or even publish it if you want to get spicy without any work, right? So this is for the repeatable work. So whether that's for prep docs briefing, I think that's a very popular use case. You know, kind of like your daily briefing, looking at your, your calendar, your email, your Google Drive, or if you're on the Microsoft side, it connects to, you know, Your outlook, your SharePoint, your OneDrive, but kind of just helping you manage your day to day, your meetings, all of those things done for you. Analysis, coordination, you know, creating content, obviously triaging, you know, certain things based on a set of predetermined rules that you can build conversationally. So I really would say that there's no hard limitations right now with workspace agents, mainly due to the fact that you can schedule them, they can run autonomously in the cloud to your connected data, you can edit them and they have custom MCP and skill support. So this is kind of like the agents that I think a lot of us have always wanted, but we just haven't really had access to. So there are different ways to build them. I showed you one way, which is just conversationally, right? So that's way number one. Another way that you can do it is by using a template, which is great. The third way is you can build it from scratch. So if you are more of a builder, that's a great way to do it. And then last but not least, OpenAI has said that there's going to be a custom GPT conversion. So what does that mean for the future of GPTs? I'm not sure. OpenAI did call workspace agents the next iteration of GPTs. So we don't know if they're going to eventually get rid of them, stop supporting them. Maybe they'll always be around, I'm not sure. But that is kind of the last way to build it is OpenAI did say that they will be releasing soon a dedicated tool that converts all of your GPTs into workspace agents. The big hold up there is, well, what about for people with non team accounts, right? That's going to be one of the biggest stick ups. And let me just tell you the straight facts here, right? For me, I have a team account that I just use, right? And one of the reasons why is because even on my, you know, $200 chat GPT Pro account, I don't have MCPS, right? Yeah. Don't have custom MCP access on that certain level. So I've always had different tiers. But For a lot of people, you might just want to get a team plan. You have to have a minimum of two seats. And if you look at that right, so you're paying, I don't know, at that point, like $60 a month. Then you get workspace agents. You technically have double the accounts because you have two accounts. You do have to have a minimum of two accounts, even if it's under one team. So even if you're on something right now, as an example, like the Claude 100amonth plan, let's say you are a power AI user and you're like, well, I'm not going to do this because it's only for teams. Well, for $60, you're saving money. A single team or business account gets higher rates than $100 a month plan. So you probably have double the rates for almost half the price. So I'm just putting that out there. That's not biased, that's just the facts. Go, go do that yourself. I've done it plenty of times. I've been on all the different quad plans, the $20, the $100, the $200. And you get better limits on the business plan. For Chat GPT, it's a little bit better than the $20 a month plan. So you're getting better limits anyways. So just putting that out there. So those are the different ways that you can build, you know, four different ways, one of them not out yet with the GBT conversion. Then you connect your apps and skills, you can preview it, you can do test runs, you can refine it like you already saw me do. And then you can share and publish it to your team or you can use it individually. So you just build in plain English and then you test before you share. Last but not least, the controls and governance. This part's big because you can actually go through and see every single thing that this agent has done. Right? You don't have that right now with GPTs, you technically don't even really have that capability in Codex, which I absolutely love. So I think the play here with workspace agents is for teams that are already embedded on a ChatGPT business plan or a Chat GPT enterprise plan, obviously, because that level of control is huge. There's also RBAC or ROLES based access control. So you know, especially if you're a heavy Microsoft organization or if you do need that very strict governance because with great power for workspace agents means great responsibility. So it's not like you are going to have, you know, swarms of agents going out and performing work and no way to actually go through and trace and observe what they actually did. So the control and governance side is a pretty big piece that you can go through, literally. Right. I like, I say it's like showing your work back in grade school. For math, you can go back and see every single run that an agent has done to know if something did go off the guardrails. Maybe you need to go and tighten that up. If you didn't get the best responses last week and you need to improve it, you can go see where it maybe went wrong and how you can improve it. All right, so let's go back here and take a look. Let's see how our agent did. All right, it's done. Cool. So now I'm showing you the view for your agents that are already built. So like I said, this is a different account. I don't use this one a ton. This is more for testing. But now I have my Beehive AI story digest. All right, so this is on my main agents page. I can go in and edit this, I can copy the link, I can go in and start sharing, etc. So a couple other things to point out here as I'm showing you now, the inside, you'll see right now in the upper right hand corner, it is it. This is on a schedule because I told it conversationally to run every single day. So I can go in and I can edit this, I can run it now, or I can add a new schedule. So I can click on this as an example and say maybe I want this on weekends or maybe I do only need this Thursday. Actually, I really only need it Thursday. So I'm just going to go ahead and update this. I only need this Thursday. And we're going to say at 7. No, we're going to say at 5:00pm, right. Because this is for my Friday show. So this is going to be an email that I'm now going to get at least just on Thursdays at 5pm because there's always a lot of releases on Thursdays. All right. And I'm going to update my schedule there. Perfect. So the other thing is I can see the latest runs. There's two different places here. There's one on the left hand side that says review the latest digest. So if this does run every single day versus having to go search for the end, right? Because my end output for this example was going to. My, excuse me, was going to my email. So instead I can click on this and chat with the agent and you know, it'll Set a prompt that says review the latest Beehive Digest and show me the top clicked stories and AI updates. These are kind of starter prompts and you can add these to your agent. But let's go ahead and look at two other things. So this is your activity. There hasn't been a run yet, so let's go ahead and test this. So let's go ahead and let's just click run. Now. I do have one that's already done. So now you'll see in the top it says running Automation. All right, which is. I'll probably just jump in. I have the email already done. But two other things I did want to talk about. You can always see which apps are connected there. So at the bottom you have your activity, your apps in your memory. So you can see right now the only apps I have connected are my Gmail and my Beehive. Let's say I wanted. Because you can take advantage of all of the capabilities. Let's say I wanted to create a slide deck of these seven AI features. I can use chat GPT images too. I can create a slide deck and then I can have it save that slide deck as an example in Google Drive. So maybe if I want to improve upon this workflow, I can do that. And then you'll see those apps added as you go through and modify them a little bit more. Last but not least, memory. This is huge. So there's nothing in here yet because our first actual run of this version is running right now. But eventually, in the same way, like projects have dedicated memory for anything that's happened in there, you're going to have a persistent memory that's going to build up over time within your workspace agent. So what does that mean? That means that next week when this runs, it's going to know the seven, you know, tools or I think technically I asked for 25 because I asked for the five biggest stories over the past five days. So it's not going to repeat anything. It's going to have a running memory of what it's done each time that it runs. All right, so that's kind of the big picture overview we went hands on. But maybe let's quickly look. Let's quickly look here at an output. All right, so now I'm sharing my actual email. This is the message that was sent. So it went through, it told me the most clicked, right? So it said this. You know, as an example, this version had 309 unique clicks, right? So it's going through each day, it's telling me the most click stories like I asked. Right. I'm scrolling through to the bottom. Here we go. It went through and it kind of did the thinking and the synthesis and the real cool thing here is it also went and found additional sources on the web which I told it to so it could verify all of this information from the email from inside Beehive. So it see now it ranked, let's see, it gave me the four that I should focus on the most. So I'd go through and I would ask for, you know, hey, out of the AI features and LLM upgrades, send me the top 10. So I didn't tell it to, I just said go through and find the five stories from the last five. The five most click stories from the last five episodes. So it went through the 25 and then it suggested kind of the top four. So if I were going through and edit this again, I would say send me the top 10 that are actual AI features or LLM upgrades that you can use today. It did send me four. So it said the biggest ones that were the most clicked, the Most interest was GPT 5.5 instant, which just came out yesterday. The new Claude Finance Agents, Manus Cloud Computer and Copilot and Outlooks goes Agentic. So there's a very quick overview of how all of that works. So now let me wrap up by zooming out a little bit. Should you use this, what's the difference between GPTs versus this versus Codex, etc. Well, I will tell you this, especially if you're on a team plan. I would not be creating new GPTs because we don't know what's going to happen. We also won't know pricing on how these things are actually priced until later today. All Right, that's when OpenAI, when they announced these workspace agents, they said May 6th. Right. Essentially you have free reign until May 6th and then they're going to introduce some sort of usage based pricing. So not sure if there's a, you know, default kind of rate that you'll have included or if it's going to be 100 usage base. All right, so we'll know more on that. But let's take out future pricing aside and just look at the utility of utility of these three different things. So looking at GPTs versus Codex versus workspace agents. All right, so technically some benefits of GBTS is well, you can use them in any normal chat inside of chat gbt, whereas agents are not really like that. They have predefined workflows and yes, in theory you could turn whatever you're normally doing into a chat into a. Into a workspace agent. But still, the biggest benefit of GPTs is within 1 conversation, you can kind of flip back and forth between a bunch of different GPTs, and you don't really have those capabilities right now inside of chat GPT with workspace agents, I'm actually double checking, unless they updated something, which I don't think they did. Oh, no, I stay corrected. All right. That's why I do it live and I always check. I don't think this feature was available when they first came out, but now I'm looking. If you click the add button, you actually can flip between different agents, which is really cool, right? That's the fun part of doing these live. And, you know, sometimes I forget to test every single feature during my prep, but I always want to make sure that I do it while we. While we talk. So. All right, in terms of benefits of GPTs, then there's not a ton, right? Unless they're already really ingrained in your workflow. If you did a bunch of custom setup on these GPTs, right, because you can connect via third party, via APIs, all those things. But like I said, eventually OpenAI is going to bring you a way to easily convert those into workspace agents. So I wouldn't start building new GPTs if you're on a team plan. Instead I would look at, well, either build them now inside workspace agents or wait for that conversion process. If you're on an individual plan and you don't want to spring for the workspace agents, maybe you continue to work to use GPTs or you use Codex, right? So these workspace agents are powered by Codex. So for those that were watching live, that's how you know it was kind of using my screen and you could watch it click around, right? That's the power of Codex for me. Let me be honest, I very rarely use ChatGPT anymore. Little, little secret, right? The only time I really use Chat GBT is if I know I'm going to need to access something via mobile, right? So if I'm working on something throughout the day, I know I'm going to be out and about, you know, that's when I'll actually use chat GPT because then I can have it on the mobile app. Because right now, one of the biggest reasons why I don't just use Codex for 100%, everything is, well, no mobile app. Although the team did kind of allude to that was coming. And we do know if you don't know Codex, you know, I think it's gotten maybe not quite the most accurate representation of what it is because it was originally a cloud based developer tool. Now it is. If you've ever used Claude Cowork or Claude Code, Codex is all of that and more in one unified platform and it is for everyday knowledge work. So I'll just go ahead and put my personal plug. Codex is better than Claude Desktop, Claude Code, Claude co work by far, not even close. Codex in my opinion, much better than the normal quote unquote version of Chat GPT. So if you're a heavy Codex user, maybe you're not going to find a ton of utility out of workspace agents. But if you are a heavy enterprise team, a heavy business user that uses Chat GPT right now, workspace agents are going to be your go to that and the fact that they all run in the cloud. Whereas at least right now, for the most part, anything that you build in Codex, because you can build, right, they may not be called workspace agents inside Codex, but in your automations you can build out the exact same thing. But the downside is yeah, your actual physical machine has to be on and the Codex app has to be running for these to run. So that's kind of a very informal quick back and forth between GPTs, codecs and workspace agents. But I'll leave with this. Whether you're setting these up in Codex or setting them up via a workspace agent that you can share across your organization, the big call out here is it is easier than ever without being technical, without knowing how to write any code where you can just go and start converting all of your day to day manual knowledge work processes like that's why I wanted to give you an actual example of how I'm using these workspace agents now with a custom mcp. This is a task that, you know, doesn't take me very long. It might take me an hour without AI to go into those last five, you know, beehive posts, scroll through, see which ones are the most clicked, click on it, verify, see if there's anything I missed, synthesize it, say which ones are actually available now versus which ones are available next month, right? So maybe that process took an hour. But the biggest thing for me is while the context carry, carrying that context over, but also the ability to get distracted. I think that's one of the biggest, you know, fridge benefits that people don't talk about is human distraction. In the day and age of social media, the Internet, even AI, right? You're oh wow. I could, you know, do A, B and C while I'm in here. No Go build, redesign your workflows front to back and then give it to Workspace Agent and then you don't have to really worry about it. All right. I hope this one was helpful putting AI to work at Wednesdays if you want. If I should do a more in depth show on Workspace Agents if you want to see more on it. If you want to see more on Codex, whatever it is, let me know in the comments. I always do. Go and check these Spotify comments or if you're listening live on LinkedIn. So thank you. If you haven't already, please go to your everydayai.com Sign up for our free daily newsletter. Thanks for tuning in. We hope to see you back tomorrow and every day for more Everyday AI. Thanks y'. All.
- [39:02]
A
And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going for a little more AI magic. Visit your everyday AI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.