Sarah Miller: Flipping on Copilot chat at work gets them started on their adoption journey, getting used to how to interact with the technology. But it also helps the organization stay safe because you have that enterprise grade security still wrapped around it.

Tom Arbuthnot: Hi and welcome back to the Teams Insider Podcast. This week we're going deep into Microsoft Copilots but in a very practical conversation about what customers are actually doing. I talked to Sarah Miller, who's product marketing manager for Copilot in the UK. Really great conversation with her. Thanks to her for taking the time out.

We get into what agents are, what the use cases are and how you can be successful. And also many thanks to Crestron who are the sponsor of this podcast. Really appreciate their support of Empowering Cloud. On with the show. Hey everybody. Welcome back to the podcast. Exciting conversation this week. We're getting a little bit wider than Teams.

We're going to talk about Copilot and obviously that's massively influencing us in the Teams space, but the whole Microsoft 360 stack and I guess the whole world at the moment. I'm very fortunate. Sarah and I ended up on a webinar together in the last couple of months and got chatting and she's kindly agreed to come on the podcast.

So Sarah, do you just want to introduce yourself and give a little bit of context about your current role? 

Sarah Miller: Yeah, thanks Tom for having me. I am Sarah Miller. I am the product marketing manager for Copilot here in the UK. What that actually means is it's my role to help customers understand the value that, um, assistive and generative AI can bring to their organizations.

As well as realizing that value. So what can they do now that they have it to implement it to drive up that adoption? 

Tom Arbuthnot: I love the realizing the value bit. I think that's the, uh, that's where we really are in kind of end of 24 and into 25. It's like everybody knows AI is here now. Everybody understands it's going to have an impact.

But how, how do I as an organization drive value out of it? 

Sarah Miller: 100%. I mean, that's where the rubber hits the road, right? It's nice to have a pretty new tool, but until you're using it and getting that value out of it, Um, you don't really get to experience the aha moment that we hope for. 

Tom Arbuthnot: What does your day to day look like?

I mean, that seems like a huge challenge. We were just talking off camera about how much is changing and how you communicate that. I know you, you have a customer user group. You work a lot with customers. What does it look like? 

Sarah Miller: Yeah. So, you know, I'll start with the external part of my role because that's the more fun bit because it's where I get to.

Present to customers to talk through and sort of dig into what they're trying to achieve with AI. And I always say it's really important to be clear on your north star goal. Um, because having the toolkit is great, but unless you know what you're solving for, you know, that's what you want to always ladder back up to.

So I'm presenting to customers. I'm doing workshops. We host events called promptathons where we invite. Business folks to come and, um, get hands on with some demo devices. So even if they don't have it in their organization, they can get hands on with the technology and have a play with it. And, um, they solve business challenges.

They win prizes at the end of the day. So this is all in the category of trying to get. Copilot into the hands of business people, the people who are going to be using it, though. You know, we love our IT folks. IT is how we get it into the hands of the business. We really need to have our business counterparts.

So people across different disciplines, marketing, finance, legal to get their hands on the technology so they can really see that aha moment. So that's external facing. Um, you know, I'm also working with marketing to make sure that we're harping on the right value, especially for the UK, um, where, where we are currently, there's so many great things that the UK government is coming out with.

We want to make sure we're, we're harping on that and how it's really a priority for the UK. Um, and then of course, in internally, I'm curating all the news that's coming our way and making sure that our people are equipped. to have the best, most up to date conversation. 

Tom Arbuthnot: That's awesome. Yeah. So just, just the, just the two or three roles.

It sounds like 

Sarah Miller: on a good day. Yes. 

Tom Arbuthnot: Um, and what is Sarah, like I mentioned your perspective on the UK is a global audience on the podcast, but like, uh, is the, uh, from my world in UC, actually the UK was quite progressive in terms of first of VoIP phones and very progressive on that. Where are we on the global scale in terms of AI adoption and AI interest?

Sarah Miller: I always say we're, I'm so lucky to live in this country, which is so able and willing to push the envelope on what's next. And that's because I think organizations are specifically really interested in driving up, you know, our GDP per capita there. They want to push the envelope. They want to stay competitive.

I mentioned the UK government earlier. On, I think it was like the 13th of January, they released this really impressive AI opportunities action plan, which is basically the UK government saying, let's roll out the red carpet for AI. We want to stay competitive in the global market. So that means that we need our organizations to invest today.

So I love working here and being here because I do feel like we're kind of pushing the envelope every day. 

Tom Arbuthnot: That's awesome. There's a few places I want to go in terms of what Individual customers are doing where they're seeing value, but maybe we should start at tail end of last year and definitely into 25 agents.

Agentic is the term not just with Microsoft, but in the wider industry. Can you kind of decode that for us? What does that mean? In practical terms? 

Sarah Miller: Yeah. So what is an agent? This terminology that people are starting to toss around and I'm so happy they are. What we like to say is every person has a Copilot.

This generative AI assistant, um, and every business process has an agent so you can think of an agent as a specialty AI assistant that's been pre designed or prompted preemptively to do a specific set of tasks, or maybe its scope has been directed toward a specific set of documents or resources. So maybe a good example would be like a legal contracts agent.

Um, Whereas you might want to use Copilot, um, to create, to spin up a legal contract or to check a legal contract. If this is a process that you're going to be doing repeatedly and you have a standard way of doing it, rather than going back to Copilot and prompting for this process every single time, you can basically design an agent where you can tell the agent, this is the way this process functions.

This is the standardized format. I want to see it. Um, I'm going to be asking you questions like X, Y, Z, and then you've created a specialty AI assistant and you don't need to be some sort of coder who's created this assistant. You can just use your natural language to create that assistant yourself. Um, so agents we've seen have, have really catapulted into the market because when we're thinking of those aha moments or ways to, to realize value, tying it to business processes is one of those surefire ways that.

We see organizations starting to really take up generative A. I. 

Tom Arbuthnot: Yeah, I love that. And I think it's I'm I'm I'm really aligned to the that business value conversation of actually a lot of the, you know, AI talking kind of consumer land is like chat based like you can drive to do things, but it feels much more practical to grab onto.

You have people doing this process. That's like, move this from here to here. Summarize this, fill in this form, do this thing. Um, and the fact that the AI can augment or replace that capability and free that person up. And from a business point of view, it's great, right? It's like consistency, it's procedure and it doesn't have to be like Yeah.

The AI comes up with, you know, something more creative, like come up with a marketing plan. It could just be very line of business type scenarios. 

Sarah Miller: Yeah, it's a much more streamlined way of using AI. And I also think it helps both the end user. So the people in your organization, you know, we say Copilot is the UI for AI.

That's that common user interface that you can interact with AI. Um, so therefore that familiarity and as they learn how to interact and how to prompt to get the answer that they want, they only get better because it's consistent. And then from a governance point of view, you have this one pane of glass with which you can manage all these different business processes.

Whereas before maybe you had like different point solutions that that you were trying to cobble together. That would be really hard to manage 

Tom Arbuthnot: that makes sense. And what is that like drilling down on agents? What does that look like? Because you've got Microsoft doing first party agents. So take like a dynamics product.

For example, they will frame agents in Teams. We've got facilitator interpreter coming. Those are agents. But those are built ready to work for the customer. And then what you're talking about here is more line of business. Is that, is that Copilot and Copilot actions? Is that Copilot studio? Is it proper code?

What's that look like? 

Sarah Miller: Yeah, you, you raise a good point. So a couple of things in that there are agents that are now included. Out of the box, right within a Microsoft 365 Copilot license. And those are agents that we saw just so frequently being built that we thought, Hey, like, let's just put it in the hands of our customers prebuilt because known if these are the ones that are most common.

So those are things like a prompt excellence agent that helps you improve your prompts. Um, there's, there's other facilitator agent. You know, there's other agents that we just heard most of our customers asking for. Um, then, so those are the ones that come in the box, but then when you think of like the spectrum of agents, there's simple agents, people have used the term declarative agents, and this is where you direct the agent toward a specific set of information.

Um, or documents or maybe a SharePoint. So this might look like an HR agent that you've directed toward all the HR policies, right? Um, and when a person is interacting with that agent, they can ask how those policies apply to them and they can ask specific questions. So that's a very simple agent that's just restating or declaring the information.

That the agent has been directed to, then as you get a little more complex, you can have agents that are doing tasks. So that could be like a service now agent, for example, where you're asking the agent to complete a specific action. And only when prompted does the agent, you know, complete that action and it can walk you through this level of steps.

Um, those agents tend to be created using Copilot studio because it's been a little more, um, complex or you want to be a little more prescriptive. With the if then scenarios. Then you can go all the way to the end of the spectrum, which are more autonomous agents. And these are agents that can be designed to work in the background, not necessarily prompted, but can be working throughout.

Um, and we've seen, you know, some, some of our public facing agents that, that we've shared about. Our agents that, for example, bring in new leads, they're doing constant research on the market and on customers and they raise new leads to the sales teams and that's, that's value add so you can build agents along that spectrum ranging in their level of complexity.

Tom Arbuthnot: That's a great way to understand, yeah, that, that spectrum is really useful to say, where would my use case fit? If there's something that's running in the background, constantly doing things proactively, is it a bit of a process driven or is it actually it's, it's meeting notes, right? Everyone needs meeting notes.

There's the, there's your, you know, intelligent recap or your facilitator. 

Sarah Miller: Absolutely. Exactly. 

Tom Arbuthnot: So Sarah, that's a really good understanding of kind of the portfolio of options, but if we can, let's get a bit more specific. I know you work with specific customers. Like what are some actual use cases you've seen where customers have driven value and what they've done with the AI?

Sarah Miller: Absolutely. I think a great example of this would be Pets at Home. So they're a pet care company. They have veterinary services. They're also a retailer and they have used a bevy of different agents across their organization. So, for example, one of them is an integration of their operations where they use Azure and an agent that is surfacing their Azure services to help surface data that's united from their physical stores, their retail website.

And their vet practices. You can imagine someone who's maybe a store leader who's on the shop floor. Maybe they want to have access to that on the go. They don't want to have to, you know, look at a specific, um, database or they don't have to want to, you know, interpret a graph. They can just ask this agent to pull these stats really quickly.

And in the background, that agent is pulling from all these different sources, but into a format that has been pre programmed. Um, another way that Pets at Home has been really powerful is through fraud detection. So they're using this AI agent to assist in retail fraud detection, so that their, you know, fraud detection team can start from something greater than zero.

They're sort of being fed cases that already look suspicious. 

Tom Arbuthnot: Okay. 

Sarah Miller: So there's a lot of power there. I mean, other ones are, you know, recommendations based off of the, um, the, the tasks that have been done with a given patient or with a different customer, it can recommend the next best action. So we've just seen them fully embrace this.

And you can hear like it's across so many different teams and so many different functions. And that gets back to this idea of. Every business process has an agent, right? So it can, it can range across different departments. 

Tom Arbuthnot: I love that database tip example. I've seen that in the context of the space where it's like actually agents spend a lot of time just swivel chairing between like interface A to get the address, interface B to check the delivery code and interface C to check with DHL or FedEx where it is.

And actually an AI given the right access can read that information, bring it together and answer it. I guess it sounds like the same scenario there. 

Sarah Miller: Exactly. You, I'm thinking of the, the faces that people make when we can kind of showcase this technology and say, Hey, yeah, like you can hook it up to so many different, it doesn't even have to be a Microsoft product.

It could be one of those third party databases. that you want to reach into and that we know you need to do your business, but to surface it through Copilot means that it's that one consistent experience. 

Tom Arbuthnot: And how have customers been about data access and AI? Because all those scenarios we talked about are like, we've got a lot of data.

We don't know how to make sense of it. So I'm guessing that's a hot topic. 

Sarah Miller: Yeah, it's definitely a hot topic. And I think the first step that most organizations take is they do some sort of audit via their IT team. They're saying, Okay, what is our data security looking like right now? How is our data labeling our sensitivity labels?

How are our permissions to kind of make sure they have their ducks in a row? Um, and once they do that, they probably do it an extremely small scale pilot with some users just to kind of stress test those permissions and make sure they trust the technology. And I think as we've sort of proven that credibility that the people who have access to given documents.

Only those with access will be able to surface information from that document. With Copilot, but people who do not have access to a document Copilot can't surface them any information. They wouldn't have had access to, um, so once they kind of see that in action and feel a little reassured we all we've been advocating for.

And what we've seen the greater success with is organizations that then just go full force because. It takes some resource to have that adoption and change management project. So doing it at once and with the full estate of people is, is what we think, you know, drives that greater impact because people can share with each other.

They're all going through the same journey. You're spending that admin and time and money. To do that full adoption change management project that one time. 

Tom Arbuthnot: Yeah, it feels very analogous to kind of the power platform powerups power automate movement. I've seen in some organizations like they get a single use case and they build and build and build.

I know there's certain orgs that I've. Talk to you that use it in all sorts of ways and about all sorts of line of business. It feels like that kind of thing. Like I found a use case it proven. I can find another one. I can find another one. 

Sarah Miller: And it's so true. And oftentimes it's those, the people who are using the technology, the people who are solving challenges or, you know, coming across prompts that were excellent that we want to encourage them to share with one another.

And. That's where the real magic happens. They have greater, uh, credibility with one another than than any one of us or the I. T. team trying to imagine their role. You know, they're living it. So when they get to discover something and share with each other, that's when you start to see real scale and excitement.

Tom Arbuthnot: Yeah, you mentioned at the start of the pod, the partnership kind of between the line of business people in the business marketing sales, whatever it may be, and I. T. And needing the people in the business to, A. Want it because let's be honest, like you want the business to pull through rather than IT to push and B. Them knowing the use cases like, like you just spend time be like, what's painful for you?

What are options? And I think that's where IT can be the absolute heroes in this scenario is. Getting educated, which is a fast moving space at the moment, and then helping tease out. And I think not going for the most complex use case day one, just being like, here's a here's a win. We can have with AI.

Let's take the win. And let's, you know, then kind of advocate that around the business. 

Sarah Miller: Yeah, take those small wins. We say during promptathons that oftentimes a prompt which can save. One minute, 10 times a day is a much better start to using Copilot than having this one masterfully written giant prompt that maybe will solve a one hour problem, but you know, it takes a lot more trial and error to get to that.

So when we're starting out with people to start with those. Small problems, those maybe low decision tasks or repetitive tasks. Those are some quick wins we can, we can start with. 

Tom Arbuthnot: It's a fun kind of bugbear of our industry that, you know, we like this amazing capability was built and immediately we called it prompt engineering.

It's like if there's any way to turn people off, like call it engineering. 

Sarah Miller: I totally agree. There's this problem with acumen and language that I do think keeps people from trying. When really we're just teaching people a new way to interact with technology, um, you know, with a traditional search engine, it was keyword search.

And I think people almost take for granted that they just now intuitively type, not full sentences into their search engine. They're typing, we're trying to tag words, right? So in the same way that people have learned how to interact with a search, search engine. Maybe they'll learn that they can just speak naturally.

Maybe it's voice. If they're using the voice capabilities or typing naturally. And because it's a semantic index. This interrelation of concepts rather than a pure like keyword tag, then Copilot will capture anything and everything that's related to that concept that you're talking about. 

Tom Arbuthnot: And like, Sarah, you spend a lot of time with organizations getting them started.

Like it's, it's a lot, right? There's a lot going on. There's a lot of options. Um, and I think. There's a lot in the box with Microsoft 365 and whatever the customer has in terms of license level. Where would you suggest someone listening starts? 

Sarah Miller: So we've recently announced something called Copilot chat.

And for those eagle eyed who, and I'm sure you're one of them that this already existed, right? It's web grounded chat and it's free and you can surface it right within Teams. You can surface it on the web. Um, I would say flipping that on for everyone is a first start. And the reason for that is, even if people don't have the full magic of Microsoft 365 Copilot, where they have their work context as they're interacting with the technology, or where they can access Copilot in Outlook.

You know, Word, PowerPoint, even if they won't have that full functionality flipping on Copilot chat at work gets them started on their adoption journey. So it gets them sort of getting used to how to interact with the technology. Um, but it also helps the organization stay safe because you have that enterprise grade security still wrapped around it.

So I think starting first thing to do is flip on Copilot chat. Because then you know that everyone is using generative AI that is safe and within the confines of this secure environment. Um,

Tom Arbuthnot: and that's pretty much all the M365 sku, so you don't need to be like E5 super premium, like, you know, even things like frontline, they have access to that Copilot.

Sarah Miller: That's right. That's right. If they have a Microsoft login. Yes. So, um, so that's number one. And then I would say number two is. Start to approach your business stakeholders to understand, Hey, where do we have big problems we want to solve? Who, who wants to get their hands on this technology? Um, and then you can start doing scoping out those Microsoft 365 Copilot user groups.

Um, but as I said, I think getting past that IT, um, audit, let's say of like information and security. Is really important. We don't want to get stuck there because getting them in the hands of people in your business stakeholders is so, so, so important that mapping needs to happen. So you might as well start today.

Tom Arbuthnot: I mean, the other thing I've seen is, uh, it's quite, it's been surprising how that message of Copilot chat and the branding, you know, the three or four brands we've had in the last two years, it hasn't really got through like, and there's no way people inside your organization aren't using. The consumer options like, like they're, they're going to do something.

So you're giving them a compliant safe in the box option. And then I, I find it very hard for any organization to credibly say we haven't turned it on because we're not ready for AI. It's like, well, I mean, you're very trusting your users aren't picking up a consumer option then. 

Sarah Miller: Exactly. There's, we have a stat somewhere about how.

So bring your own AI is not just for Gen Z. We see it across every single demographic that's using AI in the workplace. Um, even if it's not been sanctioned by their organization. So that's a risk, right? And unless you're using this sort of enterprise grade product that is going to protect anything that you enter into Copilot, for example, we're not using anything you share to train the model.

And that's the difference, right? Is like you can trust that as you're using the power of generative AI, we're not going to then allow that to be exposed outside of the bounds of your organization. I think that's, that's the key message. 

Tom Arbuthnot: Yeah, that's really important. And I think you've been, I don't want now actually.

Um, everybody in Microsoft has been working on this, but like the ability to report and understand how it's being used seems to have come on leaps and bounds. And also you've got a rev of the success kit as well, I think. 

Sarah Miller: Absolutely. So the Copilot dashboard has been really in development to help organizations.

You know, understand which apps are there. Are there people using most with Copilot? Who are those hero users that we can recruit as the evangelists and the champs that we want to share out? And there's also some really, really like deep data there about where there could be potential use cases as well.

Um, so that's the Copilot dashboard. And this goes back to this idea of. Business value and creating business value. It's not real unless you can measure it. Right? So that's, that's what we've done to try to help put that toolkit in the hands of organization. Um, we also have this free resource online called the Copilot success kit.

It's almost silly to call it one resource because what it is, it's, it's a wrapper around a bunch of different resources from um, Use cases and persona mapping through to launch day materials. So, you know, you don't have to design your own poster. We've done it for you, or we've at least started it for you all the way to different videos and things that you can learn from on how to roll out Copilot to your organization.

So that Copilot success kit is huge. And we've actually just Totally added a bunch of information to it, including the, um, agents overview guide. So, you know, in line with all these updates and exciting news around agents. So there's, there's a lot going on online. Um, but for those of you who are interested, check out the Copilot success kit.

Tom Arbuthnot: Awesome. Well, Sarah, thanks for taking the time to jump on the pod. I really appreciate how. Eloquently you decode what's going on and, uh, if it's okay with you, maybe we'll, uh, we'll, we'll, we'll give it six months and you can get us updated again on what's new and exciting. 

Sarah Miller: Sounds great. Thanks for having me.

Tom Arbuthnot: Awesome. Thanks.