Microsoft Teams Insider

Microsoft 365 Copilot Chat, New Features, Business Value and How to Drive Adoption

Tom Arbuthnot

Gareth Bleasdale, Senior GTM Manager: Modern Work and AI at Microsoft, explains what Copilot Chat is, how it differs from consumer AI, and practical steps to adopt it safely across your organisation.

  • What’s included in the free Copilot Chat with Entra accounts
  • Security and governance: enterprise protections, encrypted web queries, audit and e‑discovery
  • New capabilities: unified history, memory and personalisation, in‑app chat in Word and Excel, GPT‑5 access
  • Adoption playbook for IT, business decision makers and end users, plus Microsoft‑funded programmes and resources
  • Where agentic patterns fit: web‑grounded agents on the free tier and pay‑as‑you‑go options for tenant‑grounded workflows


Check out Gareth’s webinar on 17 November 2025


Thanks to Landis, this episode’s sponsor, for their continued support of Empowering.Cloud

Tom Arbuthnot: Welcome back to the Teams Insider Podcast. This week we are talking about the free included Microsoft 365 Copilot Chat. We have Gareth Bleasdale, who's focused on Copilot chat adoption specifically. We talk about what is in the box with Copilot chat, the new capabilities integrated into office, the new GPT five model as well, and also what's coming down the road and how IT orgs can use Copilot chat to minimize security risk of consumer AI tools and really drive business value without an additional license.

Thanks very much to Gareth for joining the podcast. Also, many thanks to Landis for all their support and their sponsorship of the podcast. On with the show.

Hey, everybody, welcome back to the podcast. Excited to have this conversation with Gareth. Normally, uh, in the past, Gareth, and I've been talking converged comms and Teams, but Gareth's roles slightly changed.

So we're gonna be talking, uh, Copilot chat, and specifically the business value of Copilot chat, which is included with all Entra accounts, which we'll talk about. Um, but first, Gareth, it'd be great to get a little bit of a background on, on you, your previous role and what the new role is. 

Gareth Bleasdale: Yeah, indeed. And pleasure to be on Tom.

It's great to be back here as well. Um, I guess, gosh, uh, it's been an incredible rollercoaster here at Microsoft and I think the last few years I've been spent really driving, uh, the core of, uh, what is Microsoft Teams, so all of the telephony side. And then moving into Teams rooms and scaling that out across the enterprises.

And even though we've seen these drops of AI coming in in terms of, of course, the virtual backgrounds, but also the voice isolation and then moving into leverage Copilot within meetings. But really this year I'm laser focused on how do we drive that motion to to drive consumption for Copilot chat particularly, and how do we encourage users to really make the most of that incredible tool set they've got at their disposal.

Tom Arbuthnot: And this is awesome because your, your role is not to go out and sell people on M365 Copilot. Obviously there might be a, a positive side effect there if you're doing Copilot chat, but you are specifically full-time focused on just customers already have access to this free value. How do we help them adopt it?

Gareth Bleasdale: Exactly. And you know, I like to joke, we've, we've kind of given these tools for free. Um, we've got adoption assets that are available for free and indeed for, particularly for our larger customers, we provide funded execution to help drive adoption and change management for free. So it, it's a bit of a win-win situation when you look at it.

Tom Arbuthnot: And, and, and, and that's kind of the blessing and the curse of, of Copilot chat in a sense, isn't it? Because it's free. There's no typically no project around it. So you are having to kind of, uh, get organizations to realize that just because the product is free doesn't mean that they don't need to do some, some comms and adoption to drive the value out of the free product 

Gareth Bleasdale: Exactly. And what we're hearing from kind of busy IT teams is we're putting together a package of asks around Copilot, paid and the agentic world. Uh, but actually we've not got a mechanism to ask for internal resources to drive adoption and change management. So often they're coming to us or or to our partner communities to ask for help on how on earth do we do this.

Tom Arbuthnot: Awesome. Well, let's level set first. 'cause um, God bless Microsoft branding. Um, we've got, we've got Copilot the consumer product, which is obviously free and consumer. We've got Microsoft 365 Copilot Chat, which is the free tier with Entra and and M365 or O365. And then we've got Microsoft 365 Copilot, which is the paid Copilot license.

Right? 

Gareth Bleasdale: Yeah, indeed. I don't think we've won ourselves any favors with our taxonomy, but uh. What you'll see is effectively Copilot chat is both a feature, uh, so it's what we call the experience of consuming Copilot, uh, within the Copilot application, but also the inclusive element of Copilot that's part of your Microsoft 365 SKU.

So, uh, traditionally when we talk about it, we will talk about Copilot chat, uh, specifically as that free service. Uh, or the inclusive service. And the great way to differentiate that versus your consumer offering is by looking for the little green shield on the top right hand corner of your browser or where you're, where you're using Copilot.

Tom Arbuthnot: That's good to know. And it is quite, it's, it's quite important. 'cause we, we tend to use shorthand or certainly I use shorthand of like, Copilot free and Copilot paid or, or Copilot free and Copilot premium. I've even been saying now, and it's like, but Copilot consumer is obviously actually has a free and a paid tier as well.

But for, for the sense of this podcast, what we're talking about is you are signed in in an Office 365, or a Microsoft 365 account. You've got that enterprise protection and you're using the Copilot chat that comes with your Office 365 or Microsoft 365. 

Gareth Bleasdale: Exactly. And that's the tool that is designed for work.

Whereas your Copilot consumer is designed for booking holidays and having chats and, uh, typically not sharing, uh, Excel spreadsheets of data and PII and the likes. 

Tom Arbuthnot: Awesome. And, and so, and, and Copilot chat has gone from. Uh, like it was, you know, Bing chat. If we go back far enough, it was a paid add-on for a while, then it was included and, and now it's included and it's got some quite powerful capabilities rolling out now of actual app integration.

Maybe you could take us through what is in that free Copilot Chat tier. 

Gareth Bleasdale: Yeah, indeed. And by the way, I love that evolution that it's gone on. And I think one of the challenges that, uh, that we as humans often face is that, that incredible pace of change, the, uh, tangential increase in scaling power that these models have.

And we need to continually adapt and understand what's available for the tool today and what can it do in, what will it be able to do in three months, six months, nine months. Um, but I guess out of the box today, you can think of Copilot chat as. AI powered chat, agents and IT control. Those are your kind of three. Behind that

Also got some asset creation capabilities and pages, which allows you to, uh, use a Loop type component to manage, uh, prompting and sharing of information. But really chat agents and IT control is, uh, the core of Copilot Chat. 

Tom Arbuthnot: Awesome. And like, so, I mean, I think most people are familiar with that kind of.

Chat type approach now of like app or website chat box, ask and answer type scenario. So in the Copilot chat we are not getting multi turn reasoning of researcher or deep research. We're getting kind of basic back and forward stuff, um, agents, which we'll talk about in a minute, but there's some pay as you go elements to that, isn't there?

Um, and, and just recently as we're recording it, it was announced that you'll have Copilot chat in some of the office apps. Um, with the ability to engage in the current, so like in Word or Excel, open current open documents. So not being able to reach into all of your graph and your OneDrive and your SharePoint, but kind of current active documents 

Gareth Bleasdale: exactly that and what a boost that is in terms of how you actually consume these tools and use them in your day to day work.

Um, I guess. Inside of that chat application, you've also got what I would term hidden capabilities. So the ability to do translation, uh, the ability to do image recognition, and also a really powerful dictation tool. I, when I'm, whenever I'm processing complex information, I will often, uh, click the dictation.

Uh, element of Copilot chat, talk to it and then ask it to form that into something that's reasonable and helpful. Uh, hopefully you won't need that for today's podcast, Tom. 

Tom Arbuthnot: Yeah, we've pre done that, so we're all good. So, uh, like, uh, we can go more into features, but I guess the first kind of important thing is this is all under kind of Microsoft's, you know, data security best practices.

I mean, the reality is. There's a lot of, uh, consumer AI tools out there, you know, Gemini, Chat, GPT obviously, um, you know, Le Chat and Claude and, and tons of others. And, you know, those, those org, those customer, those organizations are reporting, you know, in some cases hundreds of millions of users. So people are using these tools out there.

Um, and I feel like that there's no way that some people in your organization aren't using IT tools, so aren't using chat kind of chat type AI tools. So this gives them a way to do it in a safe way, essentially. 

Gareth Bleasdale: Yeah. And you know what we, we recently ran an AI tour in London and we did a bit of a pulse survey there.

And the pulse survey identified that 71% of users have used consumer. AI tools at work. Now we're not saying that consumer AI tools are inherently insecure, but when you are using work data, you, you need to be cognizant of a few other things, and that's why we really push the idea of kind of it sanctioned AI tooling.

So. You want to make sure you have an appropriate level of responsible AI, uh, that you are using tools that don't generate social relationships, that could perhaps turn toxic with an end user, uh, that are blocking certain types of inappropriate content, uh, but also that are managed so that you can control where data is being sent and where it's being stored and where it's being used.

Um, in a similar way. Um. We, we use WhatsApp at work sometimes, or, or, and that can be quite a dangerous thing. You really need to be careful on what you are sharing outside of your cart managed tools. Um, so, so within the Microsoft ecosystem, we start off with this baseline that your data is yours. You use Copilot chat, and we will not be training the large language models on your data.

Now that's important if you are sending holiday requests in or uh, doing something in, in the consumer world. But it's really important when you're updating customer information or you are, you are uploading an Excel file of, of confidential data to then reason over. You don't want that being used to train the wider world.

Um, and so, uh, we also have baked Copilot chat into that Microsoft 365 platform, uh, which means that you can both secure it. But you can also govern it. And this is where it really deviates from that consumer model. You know, it's secure in that it's within your Microsoft 365 service boundary. It's, it's protected by the legal protections that your contract with, uh, Microsoft has.

Um, it also allows us to do things like to, uh, uh, encrypt our web search queries and to de-identify them. But then whilst we're encrypting, we're still making sure that it's governable. So all the prompts. All the responses, all the web queries are available for audit, for e-discovery, so you can comply with your regulation and that all, uh, loops into your retention policies and into your purview management and et cetera.

So it's designed to be secure and governable. And it's also designed to align, align with kind of our responsible AI controls to make sure that you are safe. And that's the way that we deliver a safe model for ai that you can then work with, uh, uh, confidently and your users can be com comfortable and confident to execute with.

Tom Arbuthnot: Yeah, and that's such a big one. I mean, we like, uh, pretty much all the free level con consumer AI tools, like they would include the option for that company to train on your data because that's just how freemium works. And it's tough because we are talking about Copilot chat being free in the sense of it's included in your licensing, but it's not free.

Like consumer tools are free. It is, as you said, all the data protection, all the control. And, and, and frankly, if you're not giving your users a safe. Option. They're just trying to get their, their job done, a, a, a and then they're gonna be trying to make judgment calls about, well, this is about a product, but it doesn't have any customer names in, so can I summarize this?

Or, this is customer names, but not emails. Like, can I put it like, like, you just don't want them having any type of conversation like that. It's just like, well, this is, this is safe. This is our chosen tool. This is what you can use. 

Gareth Bleasdale: Absolutely. And, and this kind of feeds back a little bit to the, the cloud revolution for me, if you're old enough to have been around when we started moving from zip drives and, and, uh, USB sticks to share information to, to, to using cloud-based services.

And you had these terrifying scenarios where people were loading really quite confidential documents and sharing them on third party repositories. And the problem is how do you manage deletion of that data once it's out there? How do you control it? If another person leaves the business and starts to, and wants to take that information elsewhere, how do you prevent it?

By having that corporate sanctioned AI tool, you prevent all those kind of complicated and multifaceted risks. 

Tom Arbuthnot: That's, that's awesome. And, and I guess in terms of the, the, the Copilot chat evolution, like the, the other thing you're gonna get from users potentially is like, well, I can do other things, you know, in the consumer tools, but Microsoft recently have really.

Uh, kind of caught up with where, say, Chat GPT is in terms of model parity and, and capabilities and, and a big one is now even I believe, in the free tier you can get access to GPT-5, is that right? 

Gareth Bleasdale: It is. And I think Eve you've hit on something that's really important, which is. We are in a bit of an arms race for AI, and there are a plethora of different tools that are really constantly releasing new features, enhancing capabilities, improving their reasoning engines.

So what does that mean if you are trying to stay in control of a, a large data estate or a large tenant? Um. And so I feel like Microsoft kind of has a responsibility to make sure we're punching ahead and really driving forward with our development of our chat ecosystem so that it is a compelling offering and so that we're not incentivizing users to use third party unsanctioned tools.

So I, if you look back, even just as, as most recently as, uh, as May or June, I think in May we announced that we had, uh, doubled, uh, or increased our response time by a factor of two. And we'd increased prompt, uh, quality by something like 12%. And that's measured based upon the feedback that users give us.

Mm-hmm. So it's a very reasonable metric. And then in July, we announced the launch of a global history, which means that you can move from different expressions of Copilot chat and have a consistent conversation with the chat bot. 

Tom Arbuthnot: Yeah, that, that's a big one. Through both the apps and the main Copilot chat app and the web app as well, isn't it?

So it felt like kind of V, whatever on of this journey. V1 was, there's a, there's a Copilot experience in Word, a Copilot experience. Were Copilot experience on the web and they were all sort of disjointed. This is, uh, you might hear, you know, universal, unified, some of the different announcements, but essentially it's kind of bringing that conversation together, both for Copilot Chat free and for.

Copilot M365 Copilot chat, but it, it, it makes the user experience more consistent. 

Gareth Bleasdale: Yeah, and I'd kind of turn that as it's the kind of experience the user would expect when they go onto these tools and it, and it, it's important for making sure that we provide that consistency. So, so, so that takes us to the beginning of, uh, June, July.

Then later in July, we, we announced the launch of memory and personalization, even in the free tool. That's something that a lot of our users aren't aware of. So you can kind of pre-populate specific details and, and terminologies within memory, uh, to allow it to re refine how it res, how, how Copilot chat responds whenever you have that conversation with it.

Um, I, I always, I, I set mine up with, uh. I am a, a go to market manager with an IQ of 150, and I want to be clear and lucid and articulate and use British English, uh, which hopefully kind of expresses a little bit of my character. You never know. I, I, I don't have an IQ of 150, but that's what Copilot Chat is helping you with.

And, and then of course, uh, August came along, Tom, and to your point. GPT-5 nk we sim shipped that on the same day. 

Tom Arbuthnot: Yes, in fact, I had GPT five. I've got paid, uh, Chat GPT license as well, kind of for comparison. And I got GPT-5 first in Copilot. Now I'm an M 365 Copilot user. But like it was amazing to see Microsoft so on top of the model innovation, whereas historically we'd have to wait a few months for that to come over to the Microsoft side.

So again, a, a testament to Microsoft really pushing hard on, on, on keeping bang up to date in terms of the models. 

Gareth Bleasdale: Indeed and the, the feedback on GPT-5 has been fascinating. If you read kind of the, the wider forums, uh, because it wasn't necessarily a massive jump forward in terms of the model quality, but what it did do is reduce things like psycho fancy, which is where the model just agrees with you.

You know, the, I, I like to think that. GPT-5 is more about telling it straight. If, if you are wrong, it will not entertain your delusions. It will, uh, uh, be more likely to respond in a way that's more correct. And then it has the, these fascinating tools as well that I allow you to either get a really quick response for simple queries or route to a more advanced reasoning engine to do a little bit more of the type of chain of thought that you might expect from something like researcher.

Tom Arbuthnot: Yeah. Yeah. And again, like, uh, you are, you are, you're getting the, the, the, the, the UpToDate model on the Copilot chat tier without having to jump up to premium for that. 

Gareth Bleasdale: Indeed, indeed. And, and then that brings us bang up to date for, uh, September, October, which is where we announced Ga, general availability of Copilot chat actually inside those Microsoft 365 applications.

And that for me, really is the game changer. I mean, it, it's a simple change, but it, from an end user experience, it makes all the difference. 

Tom Arbuthnot: Yeah, I mean that, that's, that's a, a again, for the, for the free tier, suddenly being able to engage with my Word document directly is, is a really nice experience. And that, that's, that's some of the benefit of Copilot chat.

It, it can live, uh, inside and alongside the applications, whereas a, a third party tool obviously doesn't do that. 

Gareth Bleasdale: Indeed. So if you were to then roll that back up and think, well, what are Microsoft try trying to achieve with this? For our customers and where we're trying to get to is the point where every user wants to consume AI chat using the secure IT sanction tool so that we can go wall to wall, and that every user starts to build that AI literacy.

That's so important. 

Tom Arbuthnot: That's awesome. So, so we've kind of gone over some of what the tool is like and, and people listening in either will be an enterprise or maybe partner architects like. How do we get that conversation moving to the customers and what, what are you guys doing as Microsoft to kind of help help customers drive the value?

Gareth Bleasdale: That's a great question. Um. When we're, when we're talking about Copilot chat, I kind of think about this in terms of three personas or community cohorts. So your, you have your traditionally termed business decision makers, your decision makers that are looking at what's our strategy for AI and, and for those we really want to.

Be clear that Copilot ca chat can and should be your inclusive low cost to free tool, uh, for delivering AI transformation in a really secure way, in a way that fits with your business and in a way that allows you to scale out AI literacy across all your users. So I think that's number one. The second is the IT decision makers and the IT decision makers.

We want to make sure that they're really comfortable, that they understand the risk that unsanctioned AI tools hold, and how they can use tools like Defender to identify that usage and perhaps even redirect that towards the IT sanction tool, which we very much hope would be Copilot chat, and then to make sure the applications are pinned.

Um, we've then got the additional challenges in that, uh, the it decision makers tend to be on the hook to, uh, the adoption process for the tool. Perhaps we'll come back to that a little bit later, Tom. Um, and then you have the end users and the end users. We, it, that's all about skilling. It's about enabling and, uh, ensuring that those users are both absolutely clear that they have the rights and the, the, the, uh, the.

Ability to consume AI in a way that works within their business and is secure. And then developing those skills to enable and those habits to enable 'em to use the tools in anger and to get the most out of them for whatever workflow they happen to work in. 

Tom Arbuthnot: Yeah, and, and, and I've seen in organizations there's really different, uh, user profiles of personas around this stuff.

So you've got a bunch of people that are really leaning into the AI conversation that are probably experimented in the personal time that are proactively diving in, but you've definitely got a cohort of people who just busy doing the day-to-day job and haven't considered that how Copilot chat can add value to it.

So there, there's, there's different groups to target. And again, there's no cost here from a licensing perspective, it's just the, the time and investment to drive that value. 

Gareth Bleasdale: Yeah. And, and those end, end users need a little bit of help sometimes. You know, there's often that. Uh, a joke that goes around, you know, I can't sharpen my axe.

I'm too busy cutting down trees, and it sometimes you've gotta break users out of that day-to-day workflow to show what they can do and give them time and opportunity to experiment. 

Tom Arbuthnot: And we talked about, uh, or you briefly talked about at the start that the agentic is in Copilot chat. So that is, that's pay as you go if you're a Copilot chat user or just some usages included if you're M365 Copilot.

So I, is there an opportunity there to get into the age agent conversation even on the, the, the free Copilot chat. 

Gareth Bleasdale: Yeah, there is, and again, age agentic, the biggest blocker is user behavioral changes. So within Copilot Chat, the free tier, your end users are able to create their own agents using the inbuilt, what was known as Agent Builder, which is now called Copilot Studio Lite.

So that's, uh, within the actual Copilot chat application. And that allows you to, without any cost. Create web reasoned agents. So for example, you might create a, an agent that allows us some, a set of options for, I don't know, uh, creating a LinkedIn post for you or ideating on an event or perhaps, uh, working out some options for venues to have an evening meal.

You know, simple tools that are all, uh, web grounded. And then of course you also have the option to have a pay as you Go model for, uh, corporate grounded agents. So that's agents that might reference a bit of internal information or execute a workflow on your behalf using it, uh, internal information.

That's where you get into kind of the pay as you go approach or the bundle packages for tokens. But I think for most users on Copilot Chat, a great place to start is to. Look at the kind of prompts you're doing on a day-to-day basis, and then perhaps set up an agent that does that for you. So if you are using Copilot to summarize emails, pull together a couple of different scenarios of the types of emails you want it to create and use cases, and just have a go at building an agent to create those.

It just saves that time on a day by day basis. 

Tom Arbuthnot: That's awesome. And, and like historically, a lot of the conversation, you know, from Microsoft has been about the M365 Copilot license. Right. That's the mm-hmm. Flagship experience. That's the paid experience that both in terms of adoption and in terms of kind of reference case studies.

But there are programs now for partners and customers around purely not, we're not trying to upsell you, we're just trying to drive value out of the Copilot chats. 

Gareth Bleasdale: Yeah, absolutely. Um, so we have an amazing adoption site. It's adoption.microsoft.com, and on there you'll find a specific link to a Copilot chat success kit, uh, the great Copilot uh, activity, which is a set of pre documented emails that you can send out with prompting ideas and a whole host of resources.

I, I'm also running a dedicated adoption masterclass, uh, later in November. So, uh, that's, uh, gonna be a 45 to 50 minute webinar really designed for IT teams to get to know how to adopt, uh, Copilot chat at scale. Uh, so the link for that is aka.ms/adoptingchat. It's that simple. So you're more than welcome to nice sign up.

Uh, but also, uh, if you need help deploying the tool, if you've got worries about security, we have, uh, a fast track team that's Microsoft funded that can help. And then if you've not got those motions internally for adoption, we have our unified teams that can help with, uh, driving adoption. And we've even got some partner funded activities, uh, that can help you do that because we're really keen to get that driven out at scale and to help you to drive adoption wherever you're at.

Tom Arbuthnot: That's awesome. It's really nice to have a conversation about, uh, like the, the value that's in the box and, uh, yes, like by all means, you'll probably find some use cases where some of those users will be more power users and might want to go on the, on the journey to M365 Copilot. But. Uh, I, I would echo there's a ton of value in the box and it's a great safe place to start to give a sanctioned tool and, and, and, and drive some change in the organization.

And, and on your journey, you'll find a agent use cases or M365 Copilot use cases. Um, but you might just find there's a ton of value to be had straight outta the box there without any additional licensing as well. 

Gareth Bleasdale: Uh, absolutely. And I often loop back to if I was running the IT department, if I, if I was running my own business, what would I, what would I do?

And I've got a confession, Tom. I hate spending money. I think it's, uh, the most difficult thing, uh, uh, and it, it often causes the odd, dis, uh, heated discussion in our family. Uh, but if, if I was running a team, I'd be thinking I would, I'd want everyone using this. I, in my team, saving, uh, saving time, saving energy.

Little to no cost to the organization. And then we can, we can take all that rich data that we get from, from the, the, the, uh, Teams Admin Center to work out well, which users are using this in anger, what's our metric for prioritizing licenses if we do want to go there. 

Tom Arbuthnot: Uh, and a hundred percent and the security element of it, like I, I want have had a bunch of emails going out, being like, this is why it's important to use the sanction tool.

Yeah. These are the risks. Like I, I've got, I've got my audit trail. So all the positive adoption stuff you mentioned there, but also the, the reality of, uh, you know, particularly as you get into, you know, any size of organization, I guess, but even bigger organizations like, like you need a very clear policy on why it's not okay to use these consumer tools because it, it is a real data risk.

Gareth Bleasdale: Indeed and we roll back to kind of. Uh, challenges like the GDPR regulation and the scale of fines if we do leak data. Mm-hmm. And the, the, the, uh, reputational impact when you have to announce, someone said someone did something very silly with our tooling. That's not something that anyone wants to be struck with.

And Copilot chat makes it so easy to avoid those problems, especially when it's, uh, modeled with using defender to, uh, redirect users that may be going off piece, uh, and bring a sense of control over what is a very complex scenario. Yeah. Amazing. 

Tom Arbuthnot: Well, Gareth, thanks for catching us up. Really exciting new role and, uh, I know there's more coming to Copilot chat later this year, so we'll, uh, have to have you back and talk about those.

But, but for now, I recommend people look out for Gareth's webinar and adoption. And again, just look inside your organization. There's, there's metrics on usage, there's templates of comms. As Gareth said, there's a a lot of value to be had there both for, for you in it de-risking, uh, that people are using, you know.

Maybe consumer tools and for the business, you know, it's, it's a good news story to be bringing a new tool that isn't coming with an immediate price tag that they can get some value out of. 

Gareth Bleasdale: Indeed. Thanks ever so much for having me on the, on the, uh, podcast today, Tom. Awesome. Thanks a lot. Cheers.