Ep 10 - Starting the extensibility story

In this podcast episode, Kevin runs solo to talk about the extensibility of Copilot. he covers the impact of Copilot on organizations, how to prepare for it, and how to extend it. The episode also features interviews with Garry Trinder (developer advocate at Microsoft) and Anoop Tatti (Team Lead at Advania), who share their experiences with Copilot and its extensions.

The episode delves into the importance of thinking about why you would extend Copilot, the different ways to do it, and the value it delivers. The podcast also touches on the importance of responsible AI and how to ensure that the power of AI is used correctly.  Anyone interested in the capabilities of Copilot and how it can be extended to deliver even more value to organizations should listen!

Useful links:

Copilot for Devs (thanks for making this happen, Dona!) - https://aka.ms/CopilotForDevs
Sample Copilot Graph Connectors - pnp/graph-connectors-samples: Microsoft Graph Connector samples (github.com)
Sample Copilot Plugins- OfficeDev/Copilot-for-M365-Plugins-Samples: Microsoft Copilot for Microsoft 365 Plugins samples (github.com)

Garry Trinder's sessions on extending Copilot - Introduction on extending Copilot for Microsoft 365 - YouTube

Miyagi samples for custom copilots - Azure-Samples/miyagi: Sample to envision intelligent apps with Microsoft's Copilot stack for AI-infused product experiences. (github.com)

Suparna Bannerjee's GPT4 turbo with power plat form - Analyse images in Power Apps using Azure OpenAI GPT 4 Turbo with Vision – Suparna's Tech Basket (wordpress.com)

Windows ai studio - https://x.com/omiossec_med/status/1735184534650368354?s=46&t=T_UsQjGMhGGOFKsyPDAf2g

Recreating the Google Gemini in open ai - https://x.com/zohaibahmed/status/1733278841559453898?s=46&t=T_UsQjGMhGGOFKsyPDAf2g

Transcript

[Kevin00:19] 

Welcome to the Copilot connection. Now usually I'd expect to hear Zoe saying there that we're here to share with you all the news, insights and capability of Microsoft Copilot ecosystem across the entire Microsoft stack. But it's just me here. I'm Kevin McDonnell. I'm an MVP, Viva Explorer and the Copilot Strategy and Modern Workplace. AI lead here at Avanade, will be releasing episodes on podcasts on YouTube and posting out onto the socials. Will bring insights from experts from the community and Microsoft and what the different areas of Copilot are, the impact they can make to you and your organization, what you need to do to prepare for them or even start using them now and how you can extend them. And in fact, it is just me this week talking about that extensibility of copilot. I'm going to cover a few different things of why you would do it, some of the options. And then I'm going to play those last two videos that I've been promising for a little while from ESPC as well. And we will try and come back either next week or the week after because there's been a nice couple of bits of news, one of them less related to extensibility. There was the news of copilot for sales and copilot for service has gone general availability. So I'll put some links on that so you can find out more about what's available from those. And also go back to Episode 4 where we had Tricia Sinclair talking on it about the certainly the magic of copilot for service on there. So check those out. It's been a few other bits of news that I might put into the notes on there, but we're going to focus on that extensibility and one of the interesting bits of news on that I see that Jeremy Fake has changed roles within Microsoft and they now have a copilot for Microsoft 365 extensibility team directly there at Microsoft building out those materials within that. So hopefully I did speak to Jeremy about getting onto the show a little while back. Sorry if you're listening, Jeremy completely failed to catch up with you on that. But we'll certainly try and get him on and talk a bit more about that. But let's talk about extending copilot and we're talking here quite a lot when we say we sorry Royal, we in that case is it's just me, but we're talking more about how you can extend copilot in Microsoft 365. But what I love is the story that I'm talking about here is going to be the more general extensibility model for all the copilot. So I mentioned copilot for sales and copilot for service. This ability to do that will be coming in with there as well. So a lot of capability and what you're learning from this will make things really easy from there. There are two core ways to extend, but before we get onto that, the question is why would you expand extends this, why would you actually want to bring more into copilot for Microsoft 365? Well, there is a lot that's within Copilot, a lot of capabilities that it's got there around Microsoft 365 or in the sales and service version in the the capabilities there, whether it's Dynamics or whether you're using it with Salesforce and other tools in there, you've got a lot of functionality, but it's really bounded within that barrier. And focusing a bit more on Microsoft 365, your content has to be in there. Your content has to be loaded through that specific system. So you need to have put all your information into Microsoft 365. But what if you wanted more? What if you say you had ServiceNow or maybe Zendesk and others where you're bringing that external information that you need to surface that into knowledge basis would be a great one. Your HR information, you could have him work day within there. You want to get that external data and have that same capability as your content that you have in Microsoft 365. You might want to bring those systems in and as well as bringing in that data, that contents that's indexed. You might also want to interact with other systems in there. You might want to do tasks on there. You might want to raise a ticket. You might want to capture some information. If we look, I think I spoke at two or three shows about what I loved was that capability that was coming with Copilot for sales or Copilot in Dynamics 365 sales where you could log an opportunity in that flow in the chat. But if you haven't got that within your Microsoft 365 session, sorry, capability and and actually a lot of the Microsoft 365 right now, you can't create that things. You can't send an e-mail, you can't log a Planner task right now. So maybe you want to bring some of those things or external ones. It could be logging a ticket in ServiceNow, it could be logging into HubSpot, for example, making those activities, raising a JIRA tickets that you need to put from there. Those are the things you need to do those plugins for. So you have connectors that allow you to bring that external data and plugins that allow you to bring those tasks. And you may have a completely custom scenario on the interesting conversations I've had recently is, what if you've already got a GPT? What if you've got something in, say, Lama Two as an LLM large language model? What if you wanted to bring that work into Copilot? That's the kind of things that you want to start looking at these extensibilities on there. The way you do this right now is you've got connectors. So the Microsoft Graph has some connectors that can be built in and I'll, I'll share a link for those who are a bit more visual and don't want to just hear me and see me on video waving my hand around. But you've got those connectors that you can index external content. I spoke at ESPC last year in Amsterdam about that where I was talking about things you could build in like the ServiceNow. And in fact, there was announced recently, the ServiceNow has a connector that you can just utilize that will bring that information into there and that could bring your knowledge base, but it doesn't have to be knowledge, it could be data. So the example I gave at ESPC was building a graph connector to use the Pokémon API because you know, everyone needs Pokémon in their copilot. Obviously not that have kids that are slightly obsessed with it still, but there you can go and call an API and you can either hit that on a regular basis, look for changes that have happened, or if it's got a web hook or some other trigger on there, you need to host a bit of code that will go and pull that information and put it into your semantic index. OK, my beginning elite techie for there, so apologies for anyone, I'm going down that dev route. But you're basically writing something that goes and calls a service, pulls that information, loads it into somewhere that then copilot can use. And in fact Microsoft search will also use that. So you can bring that information into search as well and and bring that through Avanade. We're always already using ServiceNow for that. So we can search the ServiceNow knowledge base and I can also go and ask Copilot for example, how do I get a replacement laptop bag and it will use that information from there. But using that graph connector, If I need this eraser ticket, maybe that information comes back and says you need to raise a ticket to get that replacement laptop bag. That's where I need a plugin and so slightly different way of connecting to that information. You've got connectors, where you're bringing knowledge, the plugins you're introducing skills. So you're going and doing that task of raising a ticket. And there's four key ways that you can go and connect into that information. There's something called Teams Message Extensions and there if you've used Teams, you've maybe seen down the bottom you can add a YouTube video into your comments or stream video or I certainly can't think of other ones. I've been talking a lot about Viva goals. You can put a Viva goal, an OK R within your your response back to people. So you can go and search for that and introduce that. Another one could be you can bring a ServiceNow ticket that that is called a message extension and that can also be used within Copilot. So if you ask it to create a ticket in ServiceNow, it will trigger a message extension and bring back that information and it will bring you back something called an Adaptive card, which is a kind of simplified way of being a bit of a conventional layout on there. And why it's nice is because it's designed to work across mobile, tablet and desktop devices. It will scale automatically for that. It will return that information and give you little form where you can fill in that information from there as well. Next way you can do things, Open APR, sorry, Open API, Open AI. You can go in and call Open AI plugins. So if you've been looking at that, that world of open AI that's been around for that dev story for a while, you can go and call those clever plugins. Yes, it will call a different large language model. No, you're not directly calling the the large language model directly. That's in copilot. You you can, but you only have one hit. There probably a discussion for another time. So there's a bit of a limitation, but you can build that extension, load that in there and it'll go and call that information within that using that same framework from there. Another one is power platform connectors. So if you've built a power platform connector previously, you can make that a copilot plug in and and service that. So maybe you could have created a power platform connector over say the Transport for London, which is for those of you not UK based that they run all the transports for London. Yeah, it kind of does what it says on the tin. So for example, the underground buses, all the public transport that runs there and they have some APIs where you could go and ask it about directions from one place in London to another. How would you get there via public transport. So you could build that into a power platform connector and then that's also service through Copilot. And then the final one, and probably the one that's being pushed a lot by Microsoft and you'll hear people speaking about because it's the low code option, is Copilot Studio. You can build an extension in Copilot Studio that will go and create that, plug in and get whatever information you need using what was power. Virtual agents now call Copilot Studio Go and could call a Power Automate, could build in some custom logic if someone said this, do this, etcetera. You can build those blocks and publish that as a plugin. I'm going to put a slight honest caveat for people I haven't seen this actually active. You have to go and request Microsoft to turn it on in your copilot environment. So I know there are ones out there. It's been a bit of a challenge to see that happen in different places. So I think that will accelerate pretty soon and be available from within that. So you've got those four different ways of trigging those those plugins from there. I haven't touched on custom copilots. That's probably a story for another time. So you could use that Copilot Studio to build a copilot. That's not actually in Copilot for Microsoft 365. Completely independent. You could use as your open AI and things like the team's AI library to build something like Copilot for a specific scenario, but not going to cover too much on that today. What I do want to get people thinking about is I've gone through a lot of techie stuff there and how you can do things. I started off with why, but I also want to think about what you do. How do you work out from all these plugins? How do you govern that? How do you make sure they're being used effectively? Can you build monitoring to these these the kind of things that you need to think out about whichever scenario if we're looking at those teams message extensions you could build say as your app insights into that. If we look at some of the other items, you could build it into copilot Studio that's got some analytics built in. Having a look at those different ways that you can make sure that it is being used, log how it's being used for within there, but also how many plugins you've got controlling how those plugins go in there. Because if you've got six different plugins that are all about creating a ticket, how do you know what ticket someone's talking about? How do you make sure that the description is more effective within that. So there's a lot around that governance scenario that you do, you need to be thinking and planning about from it's as well you need to think about how you're going to host that. If it is a a plug in, how is that going to work with a different scenarios, Is it going to have some consumption for Monsieur, is it paper month etcetera on there? How is it going to scale if you add a lot more people on this? How are those costs going to increase? Is it going to be able to cope with that load? But for me, the two most important things you should be thinking about is are you delivering value with this? We've talked about this with copilot for Microsoft 365 about how you deliver value from there. So that is a really key question, how you think about and measure that. And I think that is something we'll take into a future show to go into there. And another one that I'd love to see with the future shows. How do you make sure you're giving responsible AI? Are you using this power of the AI correctly? Have you considered what this would be like if it went out in there? Are you meeting your country's regulations? Are you meeting personal data regulations as well? So thinking about some of those things within there. Now going to deviate for a bit because what I would like to do is to get some of those videos. So two people I spoke to that were really dev heavy from ESPC, one was Gary Trinder and we'll play that video in a SEC and then I'll have a quick chat about some of the other ones. In fact, no, I'm just going to play both the videos. So firstly, Gary Trinder, one of the things I will say is in the show notes and Gary's been working on some fantastic sample Gary's. So whether you're looking at those plugins or whether you're looking graph connectors, he's been working with Weldec to to build out some public GitHub samples for those. So make sure you check those out. And Gary's also been doing a session on the Microsoft community calls which are available up on YouTube. I'll provide some links to those about how you build plugins. So if you're looking at the more technical area, go and check those out on there. But for now and over to Gary and play those videos. Hello, I'm here again at ESPC and I have my Co host of great happier Princess Gary here. Would you like to introduce? 

[Garry 15:51] 

Yourself. I would thanks Kevin. So my name is Gary Trinder. I'm a Cloud Developer advocate at Microsoft and my focus is Microsoft 365. 

[Kevin16:00] 

Fantastic. And you've been doing lots with pilot extensions, graph search, connectors, all the good stuff there. But what does Copilot mean to you? 

[Garry 16:09] 

Copilot, I mean, just in in general is is just, it's meaning a lot to me because it's helping me everyday. It's making me be a better, more productive in in my job. It's helping me on those days when you know, when you have those foggy days, you know you need to do something. You've got a script to write. Yeah, exactly. Yeah. You've got a script to write and you're thinking, I know what I need it to do. But you're, you know, your mental state is not getting you right into the code and you just want to have a nice starting point. And I'm finding that Co pilots are a great way to start that off. 

[Kevin16:44] 

And and so being on dev side, is that more GitHub copilot so? 

[Garry 16:47] 

There's GitHub Copilot. Definitely I'm using that. I'm also using Microsoft Copilot as well. So formally Microsoft Copilot sorry binge Bing chat for enterprise naming and just using it like like a buddy really and and just you know being able to generate code as well as being like you know generic AAI. But I'm using it more and more to say hey, I've got to write a PowerShell script and I need you to do this and can you generate it for me? And there is a script and then I can just define what I want just through text and and it will rewrite it for me and it gets me 7080% of the way there and then I can go. OK. I'm starting from a point where I have code and I can just amend it. I'm not starting from a blank page and and and I'm finding that like massively helping with with productivity as well like little scripts here and there like oh like I've got the. 

[Kevin17:48] 

Sort of things you think are really quick. When you actually try and get your headspace around it, it isn't quite as quick as you think. But now they they become quicker. Yeah, I feel that's really interesting, 'cause, you know, my assumption would be you'd be GitHub Copilot, that would be the main one that you'd be excited about, but that's. 

[Garry 18:04] 

Intriguing. Yeah, it's it's it's using them, using them all and seeing what they can do. There's varying like levels that they've got using different models as well. And so some are obviously better than others. But yeah, that's what I'm finding. 

[Kevin18:19] 

It So what seeing the story of kind of custom Co pilots and extensions but what do you see is the importance of those. 

[Garry 18:26] 

I mean it's it's making them more relevant essentially to to you and your needs, right. So you know we've have the generic AI models that are just trained on whatever they're they're trained on, right. And that's great to to a point where you want you want context of maybe your work and your organization and being able to extend you know these these systems and and bring your own data to it is then OK now it's relevant to me the people who I work with the documents that I work with you know and and that just is gonna make the your responses from these Co pilots just you know more in line with what you're actually doing you need to get done and what you need to get done as well. Yeah. Yeah. 

[Kevin19:11] 

So you've been at ESPC and obviously you've had all this information from Microsoft beforehand and digested it all, but it has been the thing you've learned and seen here that's really made you want to go, oh, I want to do a bit more of that and find out more. 

[Garry 19:25] 

That is a very good question. I think just generally AI Just really generally AI, all forms I I'm learning as much as anyone and it's a constantly moving field. There's always new things and new ways of people using you know copilots for for for what they they do in their work. And it's learning from other people. You know how are you prompting these systems And it's like oh you know what that's a really good idea. I didn't think of using that that type of language in the way that I'm you know, prompting the the the AI and and and just getting those better results out of it. So that's something at the moment that you know, constantly learning and seeing the way that people are using it and just, you know ideas to to build that muscle of hang on. I'm not going to go to a search engine and find this. I'm going to go to a copilot like Copilot, do that work, and then bring me the relevant information so I don't have to summarise it as. 

[Kevin20:22] 

Well. 

[Garry 20:22] 

Yeah, yeah. 

[Kevin20:25] 

You can call the Azure Open AI and get to get you things and bring it back as Jason in the. 

[Garry 20:30] 

In Oh, yeah, yeah, yeah. 

[Kevin20:32] 

You know, there's no, I don't remember days of XML and XSL painful transformations, copilots or the the custom copilots can do that for you. 

[Garry 20:41] 

Yeah. And that's a perfect example just literally before this conference. So building out my demos, I need some test data. I used Copilot to generate, generate that test data. So I I basically said, hey, I've got Contoso here, Contoso Electronics, I have a range of products and it it just gave them the names of the, you know, the the Contoso drone products that we all all know from the demos. And it generated me a full Jason object of kind of like customer support queries that might happen just from those names. And I was like, that's brilliant, can put that into my demo and the things that I show and then more real world and I would have spent, you know ages trying to figure that out And now it's in few seconds and if I wanted it could have been Jason, could. 

[Kevin21:27] 

Have been XML, but it could be Excel, it could be CSV. 

[Garry 21:29] 

Yeah. 

[Kevin21:30] 

Word document You can get that sort of fake data out and be moving ahead with. 

[Garry 21:36] 

The example. 

[Kevin21:37] 

So now I have to be careful here because obviously you see a lot of interesting things coming up that you can't talk about in a public problem. But what do you want to see next? 

[Garry 21:46] 

What do I want to see next? 

[Kevin21:47] 

Or apart from your bed in. 

[Garry 21:49] 

To get some sleep. Yeah, there is, there is that. Well, if I was gonna go absolutely crazy, it would be what do I wanna see next is be able to get to conferences without the travelling and just just magic myself here and be at more conferences with, with, with, with everyone here. It's been great to have the conversations and back to my own bed. Yeah, just let's make that easier. Let's make getting to conferences easier. 

[Kevin22:13] 

So there we go. Microsoft, if you can have a conference Travel Copilot that moves at the speed of light. 

[Garry 22:18] 

That would be very much appreciated. Yeah, as long as I get put back together the same, that'll be all good. Yeah, I'll. 

[Kevin22:23] 

Be worried, but cool. Thank you very much, Gary. 

[Garry 22:25] 

Cheers Kev. Thank you. 

[Kevin22:28] 

So really interesting chat with Gary. I'm not going to witter on. Hopefully Zoe and I can chat about these a little bit in the future, but really loved what's Gary was covering there. The other person I spoke to was Anup Tati who works for now. He was content in cloud and I should have made this in my notes. I think it's at Advania, almost forgot the name, so renamed to Advania and he's been working with Chris O'Brien and others for some time there and doing some really interesting stuff with the chats GPT and the other open AI and these European AI plug insurance. So we talked about particular scenario, but a new. But I'm going to hand over to you. Hello. I'm here at ESPC and I have with me a new Anoop. Would you like to introduce yourself? Yeah. Hi, Kevin, and hi, everyone. My name is Anoop and I'm an MVP in the Microsoft 365 development category. And I don't know, you're not. It's just Microsoft 365 now, isn't it? They've taken the developer away. Oh, yes. Yeah. Thanks. Yeah. Thanks for the update. Yeah. So I am an MVP in the Microsoft 365 category. And yeah, I work at a company called Advania and I work there as a development team lead. Fantastic. So we've got a few questions. We've tweaked them slightly because we want to kind of hit on the a little on the extending copilots, but more you've been doing with custom copilots, haven't you. So first question, what does Copilot mean to you, right, Yeah, Copilot, it's as per the name helps us with different tasks. So for me, the main news has been, you know, in, in, in, you know where where I call, say for example, take the example of GitHub copilot. Yeah. So sometimes you know you want to write a line of course, say for example a regular expression and then you you can just write a comment and then there's a regular expression. So Copilot has been helpful to me in that way. However, Microsoft Copilot and that has got its own advantages as as we've seen throughout this this conference check here Microsoft 365 Copilot or yeah, Microsoft Copilot for Microsoft 365. We're still having fun with the names, it's fair to say. And as as we've seen in the conference, it's been the topic of this conference, we can say that. Yeah. And yeah, that has been helpful to people in many ways who have been trying that in recent times. Thanks. And have you used Microsoft 365 Copilot or you you still hoping for some access? Yeah, I'm still hoping for some access in in my organization. Some others have got access, but not me yet. Yeah. So, you know, we touched on there. We're going to talk about custom copilots and extending them. What do you see is the importance of that, right. So there's a particular scenario that we are working with currently in our organization in which this particular customer they want to surface data which is stored in BLOB storage or surface data in SharePoint and then permission TRIM. So to to to address those scenarios, what we can do is you know we can extend Copilot and not the when I say copilot, there's there's an application on GitHub created by Microsoft called as chat Copilot. Yeah, which is a customer application and this is a standalone Copilot from there, Yes. Yeah. So it's a custom copilot. So we've taken that and then added some extra bits to that. So basically that application takes your input and then calls as your opening eye and then returns the results and that content. That's generally files and documents that's in there. Is it or? Yes, Yeah. So in the BLOB storage for example, yeah, there are a lot of files and then there's an external source as well. Say for example in Sequel, the data might be stored like rows, tables with lots of rows and things like that. Or there might be files in SharePoint. So if you want to interact with these elements, then you can have your own Copilot which will interact with these elements. So I really say slightly hypothetical situation for many clients, but obviously they could get Copilot, they could use the graph connectors to bring that data in use there. How do you have that conversation as to whether build a custom copilot versus extending copilot and investing in that what, what sort of considerations do you see right. So that's a good question. So Graph Connector, obviously you know they it has got its own advantages and it depends on what kind of external data you want to bring into the system. Particular case, the one that I'm working on currently, we decided to use our own APIs that will interact with the data present in BLOB service directly and then get back that data And how are you surfacing, is that a box that you've got as a web page? Are they always that in teams, right. So it's an Azure static web app, so that's the front end and it is using React, It is built using React and then for the back end it's an Azure. I'm just going to pause ever so slightly. Don't always have the most developer audience is your static website, that's basically a standard website hosted in Azure and then you've got React, the JavaScript framework. Is there the kind of mechanism to get that out to people as well from there so as well. So it's all very much website based people logging into this or is it? Yes, yeah, it's, it's all there's proper authentication. So basically if you're signed into Microsoft 365 Single sign on, yeah, you sign into that data. Yes, Yeah, interesting. And what, what sort of usage are people seeing? You know, is this a regular thing that they're using day in, day out is an occasional or still to be worked out? Yeah. So to give an example, we've got chat CPT which everyone is using now. Organizations want private ChatGPT, yeah, because they don't want their data to be present in opening act, they want it in Azure Open. And I guess like in this case, in specific data, not everything. They just want to focus on that area there as well. Is it so correct, so the the information officers within the organization are promoting people to use private ChatGPT and that's where you know we are seeing the the demand. Yeah. So that's my questions which are hopefully will pop up in there. But I think coming back to it, we're here at ESPC. Is there anything you've seen that's particularly resonated or a new announcement that you found interesting? Yeah, from our developer point of view, the announcement that was made on the keynote related to SharePoint embedding or embedded SharePoint, yes. Yeah. So that was very interesting. So the way we can think of that is as you know say you're you have some external apps which which use documents, but they do not have a document management system like SharePoint. So Microsoft provide that as a service to these external apps. So a bit like where you're using BLOB storage, you could use SharePoint embedded and you could bring the metadata into that, you could bring sensitivity labels, you've got search and all those things. But I get and I'm thinking this off the cuff, so bear with me here. But I I guess your same app could then use that as the kind of grounding model for Open AI, and you could have all that SharePoint power to manage those documents. Yeah, with that same ChatGPT experience. It is, yes. Yeah, that's an interesting. I just thought about that. Yeah, brilliant. Yeah, that could work very nicely. So thank you. It's nice. And last question, what do you see next in, sorry, what would you like to see next in the Copilot world? That's an interesting question. I haven't thought much about that. However, you know I would like to see in what direction Microsoft are taking co-filer towards. Currently I haven't used. Like I said earlier, I haven't used it much. So as a developer, you know, I wanted to stay as Copilot and not become Autopilot and take our jobs. But I think it's been good that we've had that, you know, things like GitHub copilot that have helped developers we've seen. I I I feel that if you love the copying stuff because it feels like Microsoft has brought developers back into the pictures again with extending with custom copilot. And I know there's great tools like Copilot Studio and other things you can do with the low code, but it feels like they've bought that, as Donna Sacca calls it, the High code back up as well. It feels like a another good time, another good, which is great, yes, yeah, definitely. And Speaking of Studio, I just want to mention that there's a new Azure AI Studio as well which has been created by Microsoft which will help us create custom Co pilots. So you can create that from scratch. There are various tools available for that and yeah, that's that's really good one to take a look at much of that mentioned here at ESPC. Would it be good to Yeah, definitely something to delve into a bit more and find out more about it. Come join my session, Kevin. Yes, Yeah. You told me to go to the other one. But no, thank you very much for your time, Anoop. And yeah, thanks very much. Great stuff. Thank you very much. Well, thank you very much for that Anupam for your time on there as well. I've put a few links into the show notes that are going to cover on there. One, it was a link from Fabian Williams about the Miyagi samples. So I I touched earlier about some custom copilot, he's some great samples around the kind of custom copilot that you can build. So if you're looking for some ideas this definitely want to look at there another link for those who are more low code oriented. GPT 4 Turbo came out and that had a lot of capability of kind of analysing images as well as analysing text and someone has created a power platform way of connecting to that and surfacing that information. So check out that link as well. And then finally we don't tend to talk about the non Microsoft side, we are focusing the copilot. But there was a little chat a while back about Google Gemini and the capabilities and I loved it. Someone has created at having mocked the release video that actually was shortened and couldn't do the things it said it could do. Someone has created that an Azure Open AI, so do check out that link as well. Thank you very much for sticking with this one. Will be a slightly shorter episode, but we wanted to get this content out and I think we'll have a deeper conversation with Joey, Joey with Zoe and I about this when we can find some time to get together. We've got more experts lined up and some interviews lying up. Hopefully we can get Jeremy Thake on to talk about that deeper side, but please, please follow us on LinkedIn. Twitter checks out on the Podcasts app you're using, whether it's YouTube, tell your colleagues, tell your clients, tell your friends, even tell your parents. Everyone is talking about generative AI. Everyone's talking about copilot, so please share this. Thanks very much for listening and looking forward to speaking to you. More soon. Bye, bye.