Video: Transitioning Custom GPTs to Plugins | Duration: 2468s | Summary: Transitioning Custom GPTs to Plugins | Chapters: Webinar Introduction (4.48s), Webinar Setup & Agenda (74.805s), GPTs to Plugins (199.99s), Plugin Investment Strategy (338.31s), Plugin Governance (549.385s), Migration Demo & Editing (654.235s), Q&A and Resources (1392.795s), Q&A Wrap-Up (2024.285s), Closing & Next Steps (2417.53s)
Transcript for "Transitioning Custom GPTs to Plugins":
Hey, everyone. We're super excited to be chatting with you guys today for another webinar specifically focused today on transitioning custom GPTs to plug ins. We have a lot in store to cover, including why we're doing it, a demo of what's happening, and an overview of the timelines. Before we get started, just wanna quickly introduce myself. I'm Hannah. I'm a product manager at OpenAI. I focus specifically on the enterprise and admin experience. And so work really closely with a number of you guys and listening to a ton of your feedback on all of the controls, observability, governance that's required end to end to make sure that you can deploy our products and capabilities across your organizations and facilitate through critical migrations like this as well. I'm excited to be joined today with Karen, who's one of my engineering counterparts. Karen, do you wanna introduce yourself? Yeah. I'm Karen. I'm an engineer here at OpenAI. I'm working on the enterprise team with Hannah. Today, I'm going to be walking you guys through a demo of how this GPT to plug in migration, transition is going to happen. Amazing. Karen's a superstar engineer, so we're thrilled to be here together and to show you all that we have done to make this migration as easy as possible. So why don't we get into it? So, just a little bit of housekeeping to start here. So, there's two types of ways to interact with us. As you guys can see, this is live and so we're super excited to be answering some of your questions live at the end of this. If you wanna be able to ask us questions, make sure you do that in the q and a part. That gives you the ability of actually, like, upvoting questions. And so when we go to start to take questions from the audience, we're gonna be pulling from the q and a tab. If you just wanna chat with others, get feedback, share ideas, make sure to use the chat tab for that. So this way, any of the, like, critical questions that you want answered, we'll make sure to cover that in the q and a. From an agenda perspective, so we have a lot to cover over the course of the next forty minutes or so. First, we're gonna start with why we're transitioning custom GPTs to plug ins. Second, we're gonna go through a migration demo, show you a number of the different features that you have to be able to do this as seamlessly as possible, and also what are the admin controls that are supported for it. Next, we're gonna go through a review of the key dates and resources. And then last, we're gonna make sure that we save enough time to answer any questions that you have. So first, while we're getting into it, you guys all know that's why you're here. We're migrating custom GPTs to plug ins. Before I get into all of this, what I wanna say is that I know that many of you guys have built some tens of hundreds, some, you know, tens, some hundreds, some thousands of GPTs across your business. And these GPTs, we know, are critical ways in which you and your organization get work done today and codify critical workflows. So I think we know that this is a core part of the business today, and this migration is designed to keep and preserve those critical workflows, and providing you a path to do it as easy as possible so that those critical workflows that you've built can be moved into a primitive that's gonna be more durable over the long term, which is our plug in. So, we have a ton of empathy that because we know that this is something that you guys rely heavily on, and we also know that it's not easy to do a migration of this scale. And so we've really been keeping that in mind as we developed the entire product experience around it. Why are we doing this? Well, if we were actually just, running back in time, so, you know, go back to 2024, this is really when GPTs were created. GPTs were created in a very chat first era where their primary part of doing a workflow was a question and response and being able to scope that question in response to things like a knowledge base or defined instructions. A lot has changed since 2024. So, two years within the the pace at which our products and experiences have involved has meant that there's been a tremendous number of advancements. And first and foremost, one of the biggest ones that we've done is that we are starting to move towards an agent first world. And so you can see that with our products like Work and Codex. One of the critical things is that GPTs that were built are not compatible with an agent first work work experience. And so as you guys might have seen, those GPTs don't, operate when you're in our work tab or when you're in our codex tab. And so one of the big driving reasons why we're doing this is so that we can actually take those GPTs, we can bring them to a primitive like plugins, where plugins are supported across every surface that you're using, whether you're using that in chat, work, or codex. And so the idea is to make sure that these critical workflows can actually both persist into this agent first workflow. And then more importantly, plugins give you the container that allows you to be able to make these workflows more powerful over time. And so plugins allow you to be able to map your your custom instructions and your files into the skills. And then it also allows you to be able to map any apps that you have or any connectors into apps within the plugin. And the plugin experience built also on our agent harness allows you to be able to string together more and more complex workflows. So it really gives you the right container to be able to ensure that these workflows are not just preserved, but actually become a lot more powerful as our capabilities advance. And so that's really the driving reason behind why we wanna make sure that we can get you guys to the right primitive for these workflows going forward. It's to ensure that you have a workflow that's agent first. And then the second is to make sure that we provide you a more powerful experience for it. As part of this migration, you guys might have also seen that we are investing a lot in plugins. I know I've talked to a number of you guys over the course of the last week or two. One of the biggest points of frustration that I actually heard was that there is frustration that GPTs were not actively supported across OpenAI. There were a number of feedback requests that you guys had had about wanting things to be done differently, And this is really because GPTs had started to become a smaller and smaller focus of their organization as we moved into these agent first experiences. The cool thing about plugins that we're excited about is that this is the right primitive that has an engineering team behind it that will continue to be invested in. So when you guys have feature requests and things that you wanna be able to improve around the plugin experience, there is a team that has the ability to actually support this, make sure that this is the right, primitive for you guys, and that we're fielding that request. And so as part of that, we've made a lot of improvements that you guys have probably seen over the course of the last few weeks. So first and foremost, you have the ability to easily create a plugin, and edit that plugin through a conversation. And so you can do that within chat or work. You have the ability to just add plugin, and it's as simple as that you were to previously build a GPT. You can now do that within a conversation. So you can, you know, include any of the knowledge files that you wanna pull from. You can write any custom instructions. You can mention in conversation any apps that you wanna be included or what you want the workflow to do. And you can actually create and publish that plugin directly through that conversation. It's a much easier creation experience. The second thing that we've improved is the plugin sharing experience as well. Once a GPT is migrated to a plugin or you create an entirely new plugin, you can now have the ability of actually sharing that, with any users or groups or publishing it to the workspace. And so this sharing settings are the same that's actually going to be aligned with our desktop app experience. So we're really unifying the way that plugins operate across web and desktop, which should hopefully provide a much more coherent experience for plugins holistically. And last, the thing that we're also excited about is if you join these webinars in the past, we have always talked about how, we are excited about the ChatGPT admin plugin. So you now have the ability of using that plugin to also manage your plugins. And so by using the ChatGPT admin plugin, you can do things like being able to ensure that you have the right permissions for plugins across your organization. Also do things like being able to create reports, of being able to identify what are the top GPTs used across your organization that might have been used over the last thirty days. You can identify what those GPTs are, and then you can also use this plugin to actually trigger some of the bulk migrations. And so a lot of these experiences are designed to make this as easy as possible. And also, hopefully, you've started if you've been using the ChatGPT admin plugin, hopefully, you can start to realize actually yourself how powerful plugins are and can be. And those are the types of, like, experiences that you can create for your organization as well. So the Chat gbt admin plugin that you hopefully you have tried, hopefully it feels a lot different than if you were to just leverage a gbt. And that's really, like, the the type of experience we want. So that's really these are some of, like, the big investments that we've made so far. And it's something an area where we're continuing to improve on that admin, just holistically on the plug in experience and the admin experiences around it. One of the top questions I get is how do you actually govern plug ins as well? Just starting here on the left side, we have permissions, RBAC permissions for plugins. You have the ability to enable who can use plugins. The new creation experience is also enabled. So anyone that has the ability to use a plugin also has the ability to create or edit a plugin. So by turning that on, you have the ability to do those three things. This is very similar to the GPT permission that existed where you had a top level permission to be able to create and manage DPTs. You also have a number of different sub permissions that you can choose. You can decide who across the organization wants to upload plugins. Uploading plugins is taking effectively a zip file and uploading it that way. It's another form of creation. So if you wanted to be able to enable your users to upload zip files to create these plugins, you can use this permission. You also have another control of being able to upload plugins with custom MCP servers, which can be higher risk, which is why it's a separate permission. And then you can also see the corresponding permissions for being able to share plugins and then also publishing plugins to the workspace. And lastly, creating plugins with MCPs. So anything that involves potentially like a higher risk plugin, whether that's with MCP servers, you have the associated permissions with that. And then as, as you have with both the plugins that we support and also plugins that are now gonna be created across the organization and published to the works workspace, you have the ability of being able to actually manage how that plugin operates as well. So without further ado, I would love to actually hand it over to Karen and we're going to show you what this migration experience looks like, and some of the tools that you have available to you as well. Okay. Cool. Give me one second. I'm gonna share my screen and give a demo. Where'd it go? Sorry. One sec. Okay. I'm going to give a demo on how GPTs are migrated to plugins, and I'm also going to be showing our new capabilities around plugins such as conversation based creating and editing, and also the admin controls that we have to manage your workspaces plugins. When you go to the GPTs page, the first thing you'll see here is a banner telling you that GPTs are going away and they're migrating the plug in. When you click Migrate Plugins, you're going to land on the page that lists all the GPTs that you've created, and then you can see which plug ins that you've already migrated, and you can open those plugins and which ones you still need to migrate. And so I'm gonna click, edit GPT on that first one so we can see the configuration of the GPT that this demo migrate. So this is a basic demo GPT that, has an attachment of a list of prices and a couple of conversation starters that are asking questions about what's in that knowledge file. Yeah. So it's basically a very simple list of prices. Going back, you click migrate to plugin. What happens is that, it does some validation around, whether the GPT is ready for migration. I'm going to circle back to that, in a little bit after this basic demo completes. But basically, it copies everything from the GPT to the plugin. So instructions, attached files, and associated apps. What you can see as this is going is that this is a fully deterministic migration experience, and that's designed to be able to ensure that everything is copied over one for one. And so what we've done behind the scenes with that one click migration button is that we're effectively mapping the constraints of that GPT to the correct place within the plug in as well. And so, one of the questions that I've heard is whether it costs any additional credits to be able to do this migration experience. That's not the case. There's no inference that's used to be able to do this migration. It's one click mapping, very deterministic, and that's to be able to preserve as much as possible from the original GPT. Once it's migrated, can I take it back? Yeah, sure. Once it's migrated, it takes you directly to the migrated plug in page. When you first migrate the plug in, it starts out private, so that you can review your plug in, make sure everything looks good, before you decide to share it with your workspace again. Just looking at this, we can see that, the name was copied as well as our starter prompts. You can click on the skill and see that, the instructions on the reference file were copied. What we have here is a demo of edit plugin. So edit plugin takes you to this conversation based edit workflow, where you can, if you're not happy with the plugin and you just want to make improvements, you can do this via conversation. In this demo, I'm saying, okay, this is no longer a GPT. I'm just going to remove GPT four from the name. I wanna add one more starter prompt. So I'm just kicking this off, and you can see it, going for a bit. This is intended to be similar to, when you are working with a GPT, the option as well to be able to talk to a model and create your GPT that way. And, hopefully, this experience is intended to be, a very similar thing. So kind of while this goes, I'm going back to here and I'm showing you our sharing capabilities. So, you can share it to a specific person. So I can share it, to Hannah. So she can also review it, and test along with me. But I can also, let me go back a little bit. You can also share it to anyone with the link or you can choose to have it published, which means that when you go to the plugins page, you can see that plugin listed instead of having it, just shared. If that makes sense. Yeah. And so I think this is the like, one of the new features that we're supporting with plugins as well, and that we're continuing to invest in is the ability for you to be able to share this plugin very easily directly from, the web experience. And you can share it with any of your users or groups. Or if you have the permission, you have the ability to publish it to the workspace as well. One of the things that we're working on as well is eventually gonna be able to provide the ability to not only just share it, but you can also decide who you shared it with also has editing capabilities of it. So right now, within the plugin, what's supported is that the creator of the plugin and any admins have the ability to edit that, plugin, and then eventually that will become more granular to support other people to be able to manage and edit the plugins as well. Yeah. That's coming through. Okay. Okay. Going back to the edit flow, you can see that I opened this in work, but this plugin creator also works in chat as well. I do see a question, in the chat as well where it's not clear whether you can share outside of your organization. So right now the permissions are restricted where you can only share the plugin within the organization. You have the ability of sharing just within users or groups or publish to the workspace. That's actually one of the biggest pieces of feedback that we've heard as well of wanting to ensure that only these critical workflows across the organization can be used within the organization. Quickly, I'm switching a little bit. I'm going to show you the plugin creation experience. If you go to the plugins page and you click on the plus button that is on the right, it'll open, the plugin creation experience. So you, don't just have to I'm sorry. You don't just have the ability to, edit plugins, but you can also create new ones. And I'm sorry. This jumps back a bit back and forth, but, the edit, this is the result of the edit. And you can see that the edit was successful. The GPT name was removed from the the GPT four was removed from the title, and our new requested, startup prompt was also added. And, in this demo for, you can create a plug in. So I'm making one that is kind of like the exact same thing as the GPT just, just to show that, just just as something, like, similar. And I think what what's really cool here is, this is really, like, showcasing just the power of the conversation based experience and how flexible it is. And so it starts to just show all of the different ways in which you can both create a plug in and then also you can edit it as well. Yeah. So sorry. What am I what am I doing here? Okay. Okay. Yeah. This is what I was gonna do. So, when you once the GPT has been migrated, you can see, on your GPT's page, this is the way we migrate it. And now, Manoa shows migrate to plug in. It shows open plug in because it's, been migrated. Earlier when I was talking about the migrate to plug in experience and this validation stuff that happens, I'm going to come as we go back to that. As you can see, we do three validations on your instructions and files, your apps and actions, and your access and sharing. And those are to just make it super clear to you, like, what's being migrated and what's not compatible with plugins. And, you might ask, like, why are things might not be compatible? And it's because, like, sometimes GPTs might be created a while back, so they might be older. And so what happens is that maybe, like, you create a GPT with an app that has since been, disabled by your admin. So it's, like, certain, like, changes have been made since the GPT got created, and we just call this out, like, very explicitly. So in this example, we have this calculator app that's disabled. So you can migrate it, but, note that when we migrate it to a plug in, we don't actually you won't actually be able to use the plugin. It's gonna have a disable button, calling out that out, because one of the required apps, is disabled. And, similarly, there's another one I'm gonna show, which is with custom actions. Custom actions, is not a thing that's supported in plug ins because they're not an industry standard for how we represent, how we make network calls. MCPs is what we're going with. This also calls out custom actions won't work in the plugin. Another thing is access and sharing as well. This is because your admin controls, you might have, for example, sharing enabled on your GPT, but your workspace might currently have showing disabled on your plugin. That's just because we allow this level of granularity with your admin actions. And so, this won't block the migration, but it is to call it out that because GPTs and plugins are different and you have different admin controls on them, you might see just to make sure you don't encounter things that are unexpected once the migration finishes. Yeah. And this, this maybe if we just pause there. Yeah. I think the the way of being able to actually easily see these error messages is something that is really important here because what we want to be able to help you do is that once you actually migrate a GPT to a plug in, the GPT no longer becomes editable. The reason for that is that the plugin that you've edited or migrated it to now becomes the source of truth for any future enhancements. We wanted to make sure that before you actually go through that migration process, that you can be sure that there's going to be no issues found across that migration. So this is really important for you to be able to identify if there's any issues with who you might be able to share it with, with any of the apps and actions and and making sure that you're aware that none of the custom actions will carry over or that there's no issues with the file sizes as we're moving over to plug in. So definitely, I would encourage you guys and your teams to to pay attention of this error handling state, because once you do, if you wanted to proceed with the migration, that migration, for example, won't have the ability to edit that GPT anymore. But all future enhancements, you can do through that conversational based editing experience as Karen was mentioning. Yeah. Let me see. Okay. And then the last thing with the GPT migration is, when you land on this GPT's home page, what happens if someone else has already migrated the GPT to a plug in, you as the user? Let me hide this. You as the user will be able to see. It's gonna say, like, use plug in, and it kind of encouraged you to do that instead of, going back to use the GPT. You can choose to use the GPT because we're giving, like, several months of time before we actually finally, deprecate GPT so that, that transition can be easier. I'm gonna switch videos really quick because this is a little bit chopped up. Drum is it? And then this is, this one is to demonstrate, like, our admin controls. Sorry. Again, this kind of jumps back again to the plugin creator experience. So you can see that, a brand new plugin was successfully created from scratch via conversation, with our, attached file. So I forgot to include that in the last video, so I'm starting there. Okay. So this is actually, like, our admin controls. This is from, admin.openai.com. And here is on the GPT's page. You can see that, you can see not just, like, your own GPTs, but also GPTs created by other people in your workspace, and you can migrate GPTs on their behalf. So I'm in this demo. We click on one of those GPTs. You can see it takes you to the same model, to validate the GPT. And then once it's done, you can see the GPTs the migrate GPTs configuration. Quickly, you can also see when you click on the permissions tab, this is where, you manage your, like, your granular controls that, Hannah had mentioned a bit earlier. So, if you you can turn off the use of plugins for, like, a specific role. So if you don't want people in a certain group to be able to use plugins, you have that option. You can also control, the ability to share and publish plugins specifically, as well. There's one more demo I have, which is our admin plug in. So you have the option to, administer your GPT migrations through the UI, as well as, using codex, to, help you with this migration. So in this demo, we're running a report, to say, okay, what are all the GPTs I need to migrate and can I prioritize by the most important ones, by the ones that are most actively used? And so you can see, like, this, it generates this nice guy, this a nice summary. And the plugin's a great way to kind of ask, like, these ad hoc questions, like, that the UI won't necessarily give you, like, right away, such as this example for prioritizing the most highly used ones. Yeah. This is, I think, by far the coolest feature, so I would definitely start playing with this. It has the ability without with very, very little effort on your side to be able to pull all of the DPTs that have been used across your organization, the top ones that need to be migrated and in just a very easy way to do it. Yeah. The follow-up for this question can be like, can you find the best owner for all these high priority ones to help migrate? So, if you want, like, the to find, like, the creators, of the plugins or maybe someone who is, like, a frequent user of it, for example. So this result, finds the current owners of these high priority GPTs. Yeah. And then, coming up is, well or soon, we're gonna have the ability to also, like, bulk migrate, a bunch of, MGPTs through this plugin as well. That's all I have. Then this is the way in which you will be able to actually make it just a lot easier to do this bulk migration. So with the plug in that you're seeing here, the admin plug in, you can identify who the responsible party for doing that GPT migration. And then as Karen is showing, which is like the next action there is like, can you automatically migrate them for you so that it becomes very, very easy on your side. And so that bulk migration feature is one that will be supported within the next few days for everyone too. The I saw a few different questions in the channel. So, like, one was, who has the ability to actually do this migration? So we know many of your organizations have, you know, hundreds of these, and so this is not expected to be an admin only action. The people that have the ability of migrating the GPT is either the creator of the GPT or an admin within the organization. Those are the two different personas that can support it. As Karen showed, anyone that's a creator will see the ability to actually do it for any GPT that they own, will be able to trigger that migration directly through the GPT Discovery Service. Anyone that's an admin has the ability to go to the GPT, page within admin console or Chat gbt workspace settings, and you'll see that full list of of GPTs and you can do one click, migration. Or you can leverage something like the the plugin that Karen was showing of actually, like, pulling that and just effectively automating the migration process entirely. I think those those were all of the demos that we had in store for you guys today. I think we have probably about ten more minutes to take questions from the chat. And so, do you wanna cover that first before we do, Q and A? Yeah. Yeah. Just turn your mic off. Hi. Hey, everybody. Hey, everybody. Sorry. Hi, everybody. I'm Julianne. I'm from the OpenAI customer education team. I just wanna share, a couple of key dates and resources before we jump into to q and a. The first thing is that the migration timeline has changed from our, original plan. So we've moved out the date of GBT creation cutoff a month. So you have an extra month, in that migration timeline. So the rollout happened yesterday or the banner happened yesterday. The ability to, to complete the migrations happened yesterday. October 26, you will no longer be able to create new custom GBTs. And then December 11 is the official GBT retirement date. If you've, had conversations with your account team here at OpenAI about some different dates, this might not apply to you, but that is the timeline that we're working with. If that changes, I will send you an email that that changed. And then, after this webinar, we'll send you the recording. And then both in that email and in, the weekly email that we send every week to admins, we will continue sharing these resources. So the first is the migration guide. So we have a one stop shop and it now includes the admin plug in or will by the end of the day, that will help admins go through what their options are for migration. We have our FAQ on our help center that also has the key dates and resources that you need. And then we're ramping up our enablement for users about plug in creation now that we have the ability to create it within chat. So, again, look out for your email. All of those resources will be shared. I'll share with you an email template that you can send with your team, but keep putting questions in the q and a tab because we will use the output of that q and a tab and the chat to inform all of our documentation going forward. So appreciate you all sharing, sharing your feedback and your questions. And now I'll hand it over to, Hannah and Karen to go through the the q and a in more detail. Awesome. Thank you so much, Julianne, for sharing this timeline. So we're going to just be picking the questions off from the Q and A tab. So if there's any questions in there that you either want to ask or that you want to upvote, please make sure to just upvote them. We're just going to go write down the list in it. So starting from the top, so do plugins have a cost associated with them? So to create a plugin, it does not cost anything. The only cost that's incurred is using the plugin, but it depends on what model you're using when you use that plugin. You can use a plugin that's created in chat on an instant model if you want. You can also use it in chat on any of the reasoning models or you can use it in worker codex. Normal consumption fees apply to the use of plugins, but there's absolutely no fee for creating and migrating these plugins when you go and use that deterministic workflow from migrating from a GPT to a plugin. Second question here, I'm seeing we can't replicate all of the main functionalities we added to custom GPTs to plugins. It contains files and docs. So the plugin has two main constituents of it. So it's skills and then it's connected apps. So skills have the ability of being able to attach files and custom instructions to it. So you should be able to migrate any of the files that were serving as a knowledge base in your GPT to the plugin. So you will have the ability to replicate that functionality. Then also with the plugin, you have the ability to interact with any like apps or services that you want that workflow to interact with as well. So that's also one of the reasons why we've chose I saw a lot of comments in here of why we're migrating to plugins versus skills. The reason why is that plugins give you a container to be able to actually, improve that workflow that you want and to be able to add multiple skills if you want to include additional capabilities for that plugin or if you want to include the ability for that plugin to interact with many services. Plugins is a very extensible container for it. Your organization can continue to create skills if you want, if you want just like single file markdowns for custom instructions. But if you really want to build like an end to end workflow that you want to distribute across your organization, a plugin is like a very natural home to do that. And so we only wanted you to do the migration once, and so doing that to plug ins gives you the most extensibility going forward with it. I also see as part of this question is, will our previous chats done with custom GPTs be erased? Absolutely not. We're not erasing any of your data. Your chats for your GPTs have the ability to be maintained. When you go to use a GPT, you will see and if it's migrated to a plugin, you will see the ability to continue in a plugin. And so you will continue to be able to use any of your GPTs to the retirement date, but we will strongly start to urge your users to use the plugin if there's a migrated one, and you will be able to go to past chats that have been used with the GBT. Going down the list, can we share the meeting minutes to attendees? So, yes, this webinar is recorded so you will be able to access the recording and share with anyone that you need. Hopefully, the demos that Karen showed you gives you the sense of just how easy that migration process is, really just one click deterministic. And then also the capabilities that you have of accessing it through either seeing the report from the admin side or being able to use the ChatGPT admin plugin to be able to trigger, reports if you want it and then also eventually being able to do the bulk migration. That is definitely available for you to be able to send across the organization. What cybersecurity concerns can you address with plug ins? Custom GPTs were very non technical, low risk solution that seems to be going away. Yet, that's a very fair question. I think the GPTs were a nontechnical low risk solution. The capabilities of that GPT were also very limited as a result of it, and the ability for those quite like the the harness that it's built on, it has the ability to have, really just like questions and responses designed for it. As we move you to plugins, there are additional capabilities that that plugin can support. It also comes with the ability of being able to do more powerful end to end automations associated with it. And so would definitely recommend if you guys haven't already to go through any of the risk reviews that you need to be able to ensure that plugins have the ability to be used across the organization to support this migration workflow. What we found is that they we do have the ability to you know, the incremental risks aren't significant in exchange for the end to end power that you have for it. And so again, really, we wanna make sure that users within your organizations have the ability to use the strongest capability of doing this end to end automation so that you get the most value and plugins are the most natural workflow for that. I don't wanna, you know, belabor this example, but hopefully, it's one that you guys have used a lot, which is the ChatGPT admin plugin. And you guys can kinda get that sense of how powerful plugins are compared to GPTs when used and and how they can be used for your organization as well. I'm also seeing questions about custom GBTs that include a large amount of knowledge data, and that changes frequently. So custom GBTs with knowledge bases can map directly to a plugin as well. You can use the conversational editor to update those files, in a pretty easy way. So whenever you have changes to any of the knowledge base that you want to draw from, you can just, upload that file through a conversation and have it edit or modify there are file restrictions on size, which is why we provide the ability to actually see the, a breakdown of the any errors that you might have the potential, to hit. And so one of those is a file size limitation, and so you will see that error if any of your files in your GPT hit that limit. We are also actively working on our side of of seeing how we might be able to extend those so that there you won't hit those errors anymore of being able to ensure that you have the, like, full parity of bringing your files over. Going down, so, seeing more questions here about how do plugins consume credits versus GBTs and should we expect elevated token consumption with this change. So if you use the migrated plug in in chat in the same way you would use the GPT, there will be no additional changes to the way in which it's charged relative to how the GPT was being used. If you do use that plugin within worker codex that is built based on, translating the number of tokens that were used to complete that request into credits or dollar spent associated with it, there may be an incremental fee associated with it. And so the key thing to note here is that we don't charge anything for the plug ins. It just really determines on what surface that plug in is being used in, what model is being used, and that ultimately is what determines the consumption fee associated with it. There's a great question in here of understanding why OpenAI is routing GPTs into plug ins as opposed to skills and agents. All of the materials talk about why plug ins differ from skills, and I wanna know why I should make a GPT and not an agent. This is a great question. What we're finding is that so there's really, like, two types of, like, workflows that are starting to be created in this agent first world. The first is something like a skill or plug in, which is like a codified workflow. We're bringing the GPT into a plug in. Again, within that plug in, there's going to be skills and then also apps, and you'll have the ability to create skills or plugins going forward. We're also investing a lot in our agent based experience as well. So some of you guys have probably used Workspace agents, and the evolution of these agents is really moving towards more, like, proactive and persistent agents that are able to take on more and more work. There is a difference in experience between using an agent that might always be on and that might always be taking action on your behalf, then the codifying of a workflow within it. And so what we wanted to do is that the ideal state for a lot of the GPTs that were created, was to move it to that plug in while also providing you the ability as we invest in things like work space agents and, and more agent or team agent based experiences. You will have the ability to also build workflows for that primitive as well. But what we found is that a lot of the GPTs were not actually suited for that, like, persistent always on agent experience. And so we didn't want to bring a lot of the GPTs to that, but you will have the option to either choose whether you want it to create agent based experiences in the future, or if you want plugins or skills, which is a more natural experience of GPTs that exist. There's a lot more to come on that. It's something that we're spending a lot of time thinking about, which is, like, what are the right, agent experiences for enterprises as well? So we're super excited about that and and some of the investments that we have planned. I think that that actually hits most of the questions here. So, again, this will be sent out and recorded. We know that this is a big process for you guys, and so we're super excited, to be, working with you and making sure that you have the right tools to be able to support this migration. To the extent that you have questions, like feel free to reach out to any of your account teams. If you have any feedback on things that you want to see more of, we'd love to take that feedback and make sure that you guys have the right tools needed to be able to assist with this. But hopefully, this demystifies quite a lot of it, and shows you that, you know, there are the tools to make this as easy as possible. So really excited to see the next evolution of your workflows be built within plugins. And hopefully, this brings the ability to actually drive more end to end automation across your organizations.