AI for Business with BCN

Understanding Microsoft Copilot Studio

• BCN

Share episode

Peter Filitz, Mark Rotherham and Andy James discuss Microsoft Copilot Studio and how businesses can leverage this new AI technology. 

We cover:

  • A broad overview of Copilot Studio and its functionality.
  • Integration with Microsoft 365 Copilot.
  • Getting started with Copilot Studio.
  • How BCN can add value and help businesses.
  • Licensing and cost considerations and integration with third-party applications.
  • Case studies to help illustrate the value of Copilot Studio.

Thanks for listening!

Connect with us on LinkedIn or visit our website.

Peter:

Hello and welcome to today's podcast. Today I have the pleasure of speaking with Mark Rotherham and Andy James. Welcome, gentlemen. Hi. Afternoon. Good to have you. So today's topic of conversation is all about Copilot Studio. What is it? How can it be used? And how can you get started? We're going to be discussing today how businesses can leverage this new AI technology in their business. So, Mark, why don't you tell us a little bit about BCN's positioning of AI?

Mark Rotheram:

Yeah, sure. So there's a lot going on in AI at the moment. It's been increasing over the last 12 months at such a rapid pace, it's actually quite hard to keep on top of it. So what we're trying to do is kind of go in with a really simple three-step approach. The first and generic approach is to look at Copilot 365, which is very much focused on end user productivity, out-of-the-box use cases from Microsoft that they've delivered and basically empowering all of the consumers of M365 to be more productive. It's a great way of leveling up across the workforce a bunch of generic capability and skills. It's not necessarily an addition to the generic layer, it could be next to or instead of. So we have a lot of people looking at just building business copilots out without actually going down the Copilot 365 productivity approach. It's in this business process kind of segment using Copilot Studio. The final and third one is where we start to kind of push the envelope, push boundaries a bit further. And this is where we start to do business systems with co-pilots and large language models kind of bleeding into them and leveling them up. So here we use things like Azure AI Studio and Azure Analytics to actually embed and use the latest and greatest artificial intelligence capabilities, like all the large language models, to actually transform the way that the business operates rather than improve the business processes behind the scenes. So the last one's more transformational, it's more disruptive, and some might argue more important than the first two. That's kind of how we segment up the three different parts of AI and how we uh envisage customers will start to look at them. You know, they'll be looking at one of the three different areas or a blend of them as they kind of uh go to market and look at what AI can do for them.

Peter:

That's great. Thanks so much, Mark, for that overview. So I guess just if we were to put it into simple terms for the laymans out there like myself, you've got essentially co-pilot for 365, which is really focused around end user productivity. Then looking at co-pilot studio is more centered around the back-end systems and looking at how AI can augment and support those for driving more efficiency within the business. Is that right or fair to say?

Mark Rotheram:

Yeah, yeah, it's it's very much focused on I use the word business process again. So Andy will be able to talk about this in more detail in a moment. We do a lot of work around how we can make business processes be more automated, be more efficient. And Copilot Studio helps us do that, adding in large language to the mix to make it even more intuitive and easy to use. And it's all around taking your unique data and building out a use case that's specific for your business. What we don't want to do is compete with Microsoft here. If there's a generic capability coming along, Microsoft are likely to go and develop that as part of their copilot ecosystem. So yeah, this is where we're adding discrete value for a business around their business processes, and a lot of the time it is knitting together the back-end systems to bring that value and to expose it right channels.

Peter:

That's great. Thanks, Mark. So I guess over to you, Andy. Why don't you give us uh an overview of Copilot Studio and what it really is, how it works, and how does it work with Microsoft 365?

Andy James:

So Copilot Studio is kind of the rebranded uh Power Virtual Agents. Power Virtual Agents were the chatbots that were integrated and kind of out of the box that came as part of the Power Platform package. And it allowed users to create a prompt, put it into a chatbot, and then deploy it kind of internally or within websites. And it just gave a really nice interface where a user could ask a question or engage in some kind of conversation like you would with any kind of chatbot, but it was tailored to you and your business needs. So it was really nice and easy to put together. You could have a list of prompts and then a list of actions, and you could build out how you wanted the uh the chatbot to interact with you. There's a load of other features around entities. So if you have a reference number in your business that's in a certain structure, then when that is presented in a in a question or in a comment, the chatbot can identify that that's a reference number, think, ah, I know that reference number, I know where that database is hosted, so I can go and get all that information, and then I can answer questions based on that record. So that was kind of the the previous iterations of Power Virtual Agent, and they were quite manual to put together, and it was very much you ask a question, you get an answer, and action is taken, which was great for what we needed at the time. But what we'd seen with Chat GTP and OpenAI and all the other kind of uh revelations that have come along with AI over the past months and years is that actually we could tie in co-pilot and generative AI to give a much broader experience and more functionality to those chatbots. So what Microsoft has done, they've taken the chatbots, the power of virtual agents, and turned them into the Microsoft Copilot Studio, where you can now create essentially your own co-pilots and tie them into relevant information for what that particular copilot needs to do. Like Mark was just saying, if it's something that's broad and it's going to be against a product, then Microsoft are probably going to cover it off, so that's great. But if there's something that we need to do short-term, specific and bespoke to us, then we can use these to create our own kind of copilots. So in a in a broad sense, that's kind of what they are. For kind of an example to give that a bit of context, we can tie the copilots into all kinds of different plugins and different AI models, but we can also now give them the ability to use the OpenAI prompt and generative capabilities. So you can ask it a question, and we can, having plugged that uh copilot, that chatbot into let's say a website, or we've given it access to some specific documents. When you ask it the question, the chatbot can review those documents of the website, whatever kind of context we've given it, and it can give you a chat GTP-like generative answer that is bespoken specific to what you've asked it. Um, if the same person or a different person was to ask the exactly the same question, it would get a slightly different worded answer. It uses that large language model and that generative AI to give you a thought-through and responsive answer based on the context that we've we've kind of plugged it into, which is fantastic. There's all kinds of opportunities we can then tie into that. We're using it to help with some tender questions. So when a company goes for tender, there's a list of questions that need to be answered. You know, what's your process for project management? How do you mitigate risk? Uh, what are your policies and processes around uh environmental impact and so on and so forth? Once we've answered a number of these tender questions over time, we're building up a catalogue of information, you know, so we have a long history of these tenders, a long history of what our environmental impact policies are and the like. So actually, we can plug that information into these chatbots, into these co-pilots, and when I ask the co-pilot this tender question, it can create me a new and bespoke answer based on the tender that I'm working on, but taking into account all of the similar questions that we've asked in in Tenders Gone Past. So instead of just getting an answer from maybe Chat GTP that sounds really good, but there's no real meet it to it, the answer it gives us is based in our history, in our in our reality, and it will refer to some of those things that we've done in the past. The information it gives us is always kind of a first draft. We always want to make sure we're checking over it, making sure we're happy with it. But just the time saved in generating all of these first drafts for us in the click of a finger is just phenomenal. So massive amount of productivity increase and time saving with them.

Peter:

Yeah, no, that's so true, because I mean, I'm sure it's for most businesses the same challenge. Finding the data you need quickly and effectively is always a challenge, especially with the rate at which it's changing, growing, and having distributed systems in which these data pods live, so to speak, it's sometimes really difficult to pull it all together. So having a means by which you can call on the data you need quickly and effectively, and it's surfaced in a format that is usable and good to go is a real game changer. I guess question for Mark obviously a lot of hype around the the Microsoft 365 Copilot uh that was was recently launched. How does um Copilot Studio work with the Microsoft 365 Copilot? How does it knit that together?

Mark Rotheram:

Yeah, so the way that Microsoft positions is quite clever. So with Copilot 365, they've given this whole raft of generic capabilities out of the box, great use cases, and embed it into every single one of their office tools. And given you copilot chat as well, which is the kind of generic search. But what they've done with Copilot Studio has made it almost effortless to extend beyond those set of use cases. So as we create a new bot, a new capability in Copilot Studio, we can very quickly publish that bot into Copilot 365 and it becomes part of that broader ecosystem. And this is really powerful because then it becomes one channel, albeit in very different places, uh, where you can go and ask Copilot for help or assistance for information, and it will always fall back on that copilot you published for the relevant information that that can generate. So if we did hypothetically, we want to create a copilot for weather, we could go and uh create a copilot for weather in copilot studio, we could have it ask it to give some points of view, maybe adding traffic as well, and then you could publish that into 365 copilot, and all of a sudden it'd have an extra use case that we could then consume by the business. And that's a pretty generic use case, but where it gets into kind of a bit more easier to think through is things like uh reservations, seat bookings. You know, if I wanted to book a seat in our threading office, we could have a little copilot that's integrated through copilot studio that just does that one job, holds a conversation, makes a reservation, and has that nice generative feel to it, all where you get a different flavor and a different feel for each kind of interaction. And you could just access that through your copilot for 365. But like I said earlier, you don't have to do that. You could very quickly just create that reservation system and have it run independently in Excel. So what Microsoft has done is given us kind of a modular approach to copilots where we can choose to bring them all together under 365 or not. It really depends on what we want to do.

Peter:

That's great. Thanks, Mark. That's a great overview. Now, I guess the big question on everyone's mind is how can businesses get started today with us? Mark, can you talk us through the starting point?

Mark Rotheram:

The best thing to do is find someone curious, which is pretty much what we did when we first found out about Co-Pilot Studio in November, and log in. Um, citizen developers have got access to it today. You know, you can go and start a free trial and have a little play. It's not the most complicated thing to do something simple. So find someone curious in your organization, get them to go in, create a box, maybe point it at your website and see what it does. You know, it's a very, very good first step. That will do two things. One, it'll show, I guess, how quickly you can expose value out of large language models, but it'll very quickly show, I guess, it'll bring a load of ideas and how we can go further. And that's where your citizen developer will probably get so far that that's also the nice segue into where we can come in to help.

Peter:

That's great. And and just on that, uh, I guess Andy, one for you. How can BCN add value and help uh businesses today?

Andy James:

Well, in so many ways, in uh in relation to Copilot Studio specifically, like I say, we've got experience with building these out. We've had previous projects that have used AI in power platform elements and power of automate. So we're already well versed with the Microsoft AI stack and how it integrates. Across BCM, we have.NET teams if we need to, we have Azure specialists, so we can tie together a project with pulling in the specialists that we need for any given part. With Copilot Studio being predominantly Power Platform, you know, there are elements that can tie into the Azure OpenAI services and cognitive services. So again, we've been using those where required. A lot of it comes with the experience that we've had with the technical side. You know, we've already mentioned the tender writing chatbot to help answer some questions based on some set data, and it's all quite contained and nicely put together. But one of the real powers of chatbot, or the copilot, sorry, is that it plugs in and it's integrated with the Microsoft Estate and the Microsoft ecosystem. So asking a question can automatically trigger a Power Automate flow. Power Automate flows are ridiculously powerful. They can touch all kinds of different systems and third-party systems and integration. They can trigger all kinds of business processes and services that are required. Similarly, there's plugins directly for Copilot Studio to plug into other systems directly. So these things can kind of grow out. So one of the benefits that I think we bring is not just the technical knowledge because we've done it and we've plugged these things together and that's great, but we've got the experience of thinking on the next step. So it might be that the room booking is a case of, yeah, we just need a room booking, no problem. Well, we can have a chat box that I want a room in Reading. Yep, that date, job done, and that's as far as it goes. Well, actually, what happens when we book that? Do we also need to pick a parking space? Okay, well, let's maybe take that out a bit further. Well, if we're booking the room, do we need any equipment for it? What type of booking do we need? Do we need any, you know, actually we need to book a workshop? Brilliant, let's book the meeting room, but we need a bit of time in advance. So we can look at the business processes to see what the end goal could be, and then make sure that anything that's put in place is scalable. One of the limitations that we find when we're picking up a project that's been started by kind of citizen developers is it's built to do that one job, and it does that job really well. But when you then want to take it to the next step, sometimes we almost have to undo everything that's been done and start again. So even if you know a simple co-pilot bot to book a room, we'll always make sure we know what the end goal could be, so that we can then make sure that in the future, if you want to and you need to, we can then scale, we can make sure the integrations are there. So it's that level of experience that we've got building things technically, so we know what we technically can do, but it's the consultancy side of it to think about well, what else is available? What other things could we strive to do? And then we'll make sure that we architect and put things together in a way that we can scale towards those things over time and in the future.

Peter:

That's great. Yeah, so true. I mean, and and I think it goes back to as always referring to this journey that business is on, but it is about creating that roadmap for that journey, understanding the longer-term objectives, what the art of the possible is, um, so that whatever you do today, you know, is future-proofed for tomorrow. And that is so important because it is a service, a technology that is always evolving, and you need to make sure that you can build that future-proofing into it. Now, I guess always commercials and cost comes into every decision, every new technology or project. Can you tell us a little bit about the licensing and and how it's licensed and how this differs from the more traditional power platform? Sure.

Andy James:

So licensing with Microsoft is always straightforward. Most of the power platform is available via a normal license. So if you've got enterprise licenses, e3s, e5s, you can use standard power platform elements and connectors. So if you're paying for those and you're not using Power Automate or PowerWraps, then you're already missing out. It's there, you can use it. Once it's built, it just runs on your normal license. Elements of the power platform do require an additional license to be able to run. And that's when we get into using premium connectors. So that could be plugging into a third-party system, using AI is a really common one, and sometimes the data storage solution that we use. So SharePoint is a free place for us to store data, but Dataverse SQL require premium connectors. So that's generally how the licenses work with PowerPlatform. It's either part of what you've got if it's a standard thing, or you need to bump it up if you need to use a premium connector. Where certain things fall out of that is around the use of AI and the use of the Power Virtual Agents as work, but the co-pilots that we can now create instead. They work on more of a credit basis. So as an organization, you buy a load of credits or a load of messages, as it is with the with the chatbots, and they sit centrally across your organization for anyone to use as they need to. So the issuing of a license for a person to use the chatbot is free. You can give as many or as few out depending on who you want to be able to access Copilot Studio. That doesn't cost anything. But once that person has that license and they're interacting and they're developing, what they do then draws down messages from that pot of messages that you've bought. So depending again on the type of message and the type of interaction that the person is having with the chatbot will depend on how many messages. So just asking a question that doesn't use the AI generative capabilities, it's you know it's triggering a standard response or it's triggering a flow, that costs one message. But if we're asking a question and then have to generate the response to then come back, that takes two messages. So at the moment, I believe it's around £165 a month for £20,000, I think it is, messages.

Mark Rotheram:

£25,000, I mean.

Andy James:

£25,000. I was so close. So about £165 a month for £25,000 messages that sit across your organization. And then the people with that interact with those bots would then for each question they ask would either take one or two messages from that pot, depending if it used the AI generative services. That's also where we can come and help because we can look Um kind of time and motion studies look at demand to try and work out the number of questions that are going to be asked, the number of messages that will be used to see how many of these chunks of 25,000 messages you would need each month to support across your organization. But that's how the licensing works with Copilot.

Mark Rotheram:

So if you think about an internal chatbot, copilot, and for an internal questions, that's one great use case. Your messages are probably not going to be too significant. Publish it to the internet and have an internet-facing version of that chatbot, you could open up a whole raft of new interactions and cost. So one of the key things that Andy mentioned earlier was the solution design of these things. What one of the key bits of value BCM brings is experiencing the implications of doing a bad solution versus a really good one. And being able to explain some of these concepts and repercussions. Yeah, it's very easy for a citizen developer. I'll go and create a chat box, I'll go and publish it to the website for that to get out of control and for it to be ungoverned. And I think it it the keyword there is governed. You know, we bring experience to govern out of um shadow IT, out of citizen developers, where the capability has become business critical. So it's finding that line between citizen developers doing something useful for themselves versus that spiraling into being a business process that you actually need documentation, governance, control, you need to understand things going to break because you run out of credits and things like that. That kind of next step is really where BTN come through and help provide the value to our customers.

Peter:

Yeah, so true. And you know, as as you've hit the sort of nail on the head there, given it's all subscription-based costs, making sure that it's governed and controlled uh to ensure that it's financially viable is is obviously important. Andy, you touched on something earlier which I thought was quite an important point. Copilot Studios. So you can use Microsoft's co-pilot and tap that into third-party applications and use Microsoft AI services essentially to surface data and information from those applications as well. So it's not just for Microsoft only, because I think that's the other important point because not all businesses are purely Microsoft, right? Everyone has other products and applications typically integrated. Can you elaborate a little on that?

Andy James:

Absolutely. Yeah. So, and this is kind of where the power comes from, essentially the power platform, in that you have the chat bots and you um have power automate and the integration that comes with that. Within Copilot directly, there are plugins, and that's an ever-growing list of things that you can plug into it that you can then also plug in across the power platform. But if I give you a quick example, we're creating a ticketing system or a link to a ticketing system. So there is a ticketing system that the client has and they're using, they have been for a while, but it's a very manual and awkward process for people to find out what their updates are. We are creating a chatbot that when you ask a question, you give the IT reference number. And as I said earlier, we can identify those based on the structure of the numbers. The chatbot can grab that out of a whole string of text, it knows what that number is, it understands it, and then it queries against that third-party ticketing system to find out what the status is, when it was last updated, what the last customer-facing update is, and then gives that back to the chatbot to then present. If the update hasn't been made for the last X amount of time, so if the ticket hasn't been updated for two weeks, then we can automate notifying the engineer to say this ticket needs to be updated because the person's asking, they're chasing, they want to know what's going on, and we don't have a good update for them. So there's a whole raft of other things that we, like I say, from the chatbot we can integrate with either directly through the plugins, as these systems are having connectors made for them, or that we can do through Power Automate, which we've been doing for years and years and years now, but now we've got a really intuitive way of talking to that PowerAutomate flow. In the past it just kind of happened in the background, it was faceless. But now there's a way that you can have a conversation with those automations to get that information and then interrogate that information directly in the chatbot. So it's it's just bringing all these things together and making sure that we're structuring these in a way to be most efficient as possible.

Peter:

That's great, and uh it's truly awesome to see the change and the advantage this this development in technology is bringing to businesses. And you know, it's businesses of all shapes and sizes, so it it is really revolutionary in terms of how it's changing the way in which we work and operate. Well, that brings us to the end of today's podcast. Thanks, gentlemen, for joining us. It's been a pleasure speaking to you, Andy and Mark. No worries, Peter. Thanks for having us. Thanks, Peter. Cheers. Just uh a word for the audience. We hope you found this uh informative and and educational. You know, our aim is to bring these complex technologies to you in a way that is easy to understand and you can see for yourself how easy it can be applied to your business and bring about efficiencies and further success. Please feel free to visit our website www.bcn.co.uk. There you'll find a wealth of information around the products and services that we can help you with. Thanks for joining us today. Look forward to speaking to you next time.