The BCN Podcast

Microsoft Copilot in 2025: What’s Landed, What’s Next for AI and Agents

• BCN

Share episode

Halfway through 2025, the AI revolution isn't slowing down it's accelerating in ways that are transforming how businesses operate. In this enlightening conversation, Peter Filitz sits down with Fraser Dear and Johan Venables to unpack the dramatic evolution of Microsoft's Copilot ecosystem and what it means for organisations navigating digital transformation.

The discussion reveals how Microsoft has strategically repositioned Copilot from a standalone chat tool to the central hub of the Microsoft 365 experience. This shift mirrors their earlier Microsoft Teams strategy, creating a unified interface where AI becomes the primary gateway to all productivity applications. Users can now generate documents, presentations, and insights without switching between applications, fundamentally changing workflow patterns across organisations.

Most striking are the real-world productivity gains businesses are experiencing. Rather than incremental improvements, organisations are seeing multiplicative efficiency boosts of 300-400% for specific tasks. But this isn't about replacing humans it's about enhancement. Employees report having more time for high-value work, reduced stress levels, and unexpected skill development, with non-technical staff gaining coding literacy through natural language interactions with AI assistants.

Looking ahead, our experts predict the emergence of more specialised AI agents working in unison, creating an orchestrated ecosystem of purpose-built assistants. The next frontier may involve a shift from reactive to proactive AI systems that anticipate needs rather than just responding to prompts, essentially moving from "co-pilot" to "autopilot" in certain scenarios.

For business leaders, the message is clear: the AI and automation revolution isn't optional. Organisations need to start their adoption journey now while implementing proper governance, maintaining transparency about AI use, and ensuring human oversight.

Visit bcn.co.uk for more information on our products and services and how we can help your business adopt these new technologies.

Peter Filitz:

Good afternoon and welcome to the BCN podcast. It's Peter Filitz here. I'm your host and I have the pleasure today, with Fraser Dear and Johan Venables, joining me on our conversation, where we'll be talking all things AI. I'll let the gentlemen introduce themselves. Fraser, over to you.

Fraser Dear:

Hi everyone. It's Fraser Deer here again. I look after BCN's AI and data innovation activities, so this topic is always right at the top of the normal things we talk with our customers, so we thought we'd bring it to you today.

Peter Filitz:

Excellent Thanks, Fraser, for joining us. As always, a pleasure, Johan. Over to you.

Johan Venables:

Hi all, I'm a cloud consultant for BCN. My role is basically to help and support our customers with their digital transformation journey and also help them and support them within the Microsoft 365 environment, especially around productivity and security.

Peter Filitz:

Excellent thanks, Jan, for joining us. So can't believe it. We're already halfway through 2025. It has absolutely flown by. Earlier this year, January, we set out to see what was coming this year from an AI perspective and, with us being halfway through the year, I thought it a good opportunity for us to sort of take a step back, evaluate what we've seen to date, understand how it's impacted our clients and the businesses we work with, and then also see what's on the horizon. So quite keen to get both of your thoughts on what we've seen in terms of advancements, specifically within the Microsoft stack of services around Copilot. Quite eager to also understand how businesses have responded. Who wants to take that one away?

Johan Venables:

Well, yeah, so recently around May, microsoft launched a new update for Copilot 365. So if we had to go back on the first release date, we had Copilot chat, which was free and available to everyone with a Microsoft 365 subscription, and then they launched Copilot 365, the licensed version, where you get Copilot embedded into all your Office applications, so you didn't have to copy and paste in between Copilot chat and your office applications. It's nicely embedded, so it made it really easy to engage with AI. And now recently they sort of flipped it around a little bit and they are now fully focused on the Copilot chat interface, which now brings all your data and applications into one interface. It's very similar to remember when they launched Teams. You know you had data everywhere, conversations, emails everywhere, so they brought this one unified hub that connects you to your data and your conversation.

Johan Venables:

So it seems like Microsoft has taken a very similar approach of Copilot, where you now, if you log into your Microsoft 365 portal, instead of seeing all your applications and your recent documents, what you'll see is Copilot Chat as your landing page.

Johan Venables:

And I think the reason why Microsoft has gotten done that is sort of to expose more people to it, because we still see a lot of businesses being unaware of Copilot Chat, its capabilities, so I think they're kind of like forcing it down upon people to go oh, what is this?

Johan Venables:

Let me have a go and see what it's all about. But not just that. They've also sort of just made it easier. So instead of you now going into where to generate a document or going into PowerPoint or going into your emails to summarize important emails, you can do all of that in the Copilot so-called let's call it hub, where you can get information from the web or look for data within Microsoft Graph. So graph is data like emails, documents, conversations, et cetera, and the content it creates. What you'll notice with Copilot at the end it will ask you do you want to draft up a document, a Word document, or do you want to convert it into a PowerPoint presentation? So you can easily and quickly generate content without having to go into these different Office applications, and for me that's a very neat feature.

Peter Filitz:

Yeah, it gives you a great sort of starting point, so to speak, and, as you say, it provides that sort of seamless workflow and integration with the rest of the productivity apps. Fraser thoughts on that. What's your take on some of the changes we've seen?

Fraser Dear:

Yeah, I totally agree with Johan. It's becoming as Microsoft said. You know, their vision for Copilot was. You know it's the UI for AI and we're seeing it everywhere. It's the landing page. When you open up Excel, copilot pops up immediately. What are you trying to do today? So they are absolutely delivering on that kind of vision.

Fraser Dear:

I think the other thing that's happening which is really interesting to see with Copilot is the kind of rollout of those Microsoft level agents. So we're seeing things like researcher and analyst, which are designed particularly for personal productivity enhancements. Rather than you having to sit and do some basic research about a business or a basic research about someone you're about to talk to or a client, or you can just throw it in and Copilot will go and look across all of your chats, all of your data and all of that publicly available data and surface it to you in a matter of minutes. It's really really powerful and, again, that's exactly what Microsoft said they were going to do Get Copilot in front of users, get people using it for personal productivity and then grow that out with agents, and we're really seeing that taking hold now.

Peter Filitz:

Obviously, andy. One of our esteemed colleagues was at the recent Power Platform conference and I believe, from what he's shared with us, the big play, so to speak, is looking at how Microsoft can better expand and leverage agents and the orchestration of those agents. Could you give us a little insight into what we're expected to see on that front, fraser?

Fraser Dear:

Yeah, it's interesting the concept. You know, if I think back a year or so ago, the concept of maybe even two years ago now, oh, let's all go and build large language model, that's what you need to do. It's all bespokepoke, it's all custom. The reality now is, you know, copilot is tunable. Copilot is kind of customizable within a business sense. So that idea that you need your own business platform unless you're doing something really super special, I would argue that the models that are now available are so advanced that with some light tuning in a low-code way, it's going to meet, you know, 50%, 60%, 70% of your use cases.

Fraser Dear:

If there is something super special, yeah great, let's go and do something custom. You know, we can use AI Foundry, we can create a custom application, we can create a custom chatbot, we can create custom experiences and agents. But one of the things that Microsoft are really starting to kind of deliver upon now is this multi-agent piece, and this is where agents are linking together. There's, you know me as a user and I have my helper, but then that helper has lots of little kind of microservices sitting around it, so when it needs to jump into another agent that's sitting waiting for that call. It's ready then and available for use. So it's starting to see this. You know again this language came out around a year ago this agentic service that would sit around that Cop.

Johan Venables:

So, fraser, I did a bit of reading on multi-agents and so from my understanding is that you'll have an agent that could do let's say, you've got one specific agent that can do a draft, another one that can do the sort of analytics, and they will all have specific roles. And so when I looked at Microsoft Keynote, what really frightened me a little bit is that they say organizations may end up with millions of agents. Now controlling that is obviously going to be IT's worst nightmare, because I'm going to use Teams again as an example. Remember when Teams came out, people just created channels left, right and center and then you had data sprawl right. So what's your thoughts around that sort of managing agents?

Fraser Dear:

Oh, it's, johan, I totally agree.

Fraser Dear:

You know, one of the things that really came to me when we were looking at the researcher agent was isn't it amazing that researcher knows everything about everything I've ever done?

Fraser Dear:

And isn't it amazing, once that document's produced, I can forward it on to you, johan. But the catch is you weren't supposed to see some of the content in there because it was private to me and so that the governance and management of these orchestrated agents is going to be really interesting, because, of course, each agent will have access to different data sets, so it's understanding how they interact and understanding the data risk that that may pose. And, and you know, there are tools and services available to support that. So classification is going to become really important because, again, if I've got some special secret project, is it appropriate that that's then included in that research paper that I'm asking it to create? And, of course, all of those subsidiary agents that's then set around that master agent, if you like. You know what do they have access to? Is it my permissions or is it your permissions, or is it organizational permissions?

Johan Venables:

absolutely, and I see there's going to be a lot of talk. Well, there's a lot of talks about out-of-the-box agents that you can just get from the sort of app store. You know that people from different businesses will generate and create and publish which can be shared, so that could also be a little bit of a nightmare to control to make sure that people are using the right agents in the organization.

Fraser Dear:

It's understanding what access that agent has and the data that's contained within that access credential. That's going to be critical to manage and I think it will require maybe another agent to manage that. Hey, who knows. But there's going to need to be some changes required inside internal IT organizations to be able to manage and control the sprawl of what Microsoft likes to do is give it into the power of the people.

Fraser Dear:

And so you know, creating your own agent inside a SharePoint site is great, but should that be shared as a business tool? You know it's back to that my agents for my personal productivity, and then the creation of something by me for my team Cool, great, self-serve. But the creation by me of something that's shared to an organization that's where it organizations really need to be very, very mindful that there could be some significant risk in that and that comes back down to the a high adoption journey, so to speak, and making sure that you and your business, along obviously with your data, are ready for it and I know we've got got an episode coming up where we'll talk in a little more detail around the adoption journey that we've seen work and the approach that businesses should take, so we can discuss that more.

Peter Filitz:

then Tell me keen to get your guys' thoughts around the impact AI is having on businesses today. Obviously, I know you guys are very much on the forefront in pioneering some of these technologies into businesses. We work with Fraser. What are some of the key benefits you've seen from the project you've been involved with?

Fraser Dear:

It's a really good question, peter, because I think there's a lot of hype out there. Oh, we're seeing return of investment of X and Y and then, of course, when you actually deploy it into an organization, they're like right, where's that return of investment? What I would say is, in almost all cases, whether it's a return of investment on speed, efficiency, just capability, in some cases actually, actually organizations are seeing significant return. So, you know, we're seeing numbers like 3x, 4x. It's it's not a percentage of, it's a significant advance and it doesn't necessarily mean oh well, if it's a 3x, that means I need 3x less people.

Fraser Dear:

Microsoft is positioning copyle in particular around personal enhancement, not personal replacement. So it's saying okay, great, that person can do that task three times faster. So what am I going to be replacing that task with? Because, let's face it, no one's sitting with, you know, a queue of nothing to do. So it means that they're going to have more time to actually do the higher value tasks rather than those lower value. You know mundane, repetitive pieces but, to answer your question directly, I think it is anything from significant portions of people's day 20, 30% up to multipliers of enhancements.

Peter Filitz:

Yeah, so time saving, which, as you quite rightly say, you know gives people back more time to focus on the more important elements of their job, the elements that really drive value our customers and their staff is using CoPilot.

Johan Venables:

There's a really positive impact that it's having, because they now have enough bandwidth to do the things that they want to do in a much happier place, less stressed, which is also good, because they get CoPilot to help them get organized, help them to generate new content quite quickly rewrite policies, you know and they're learning new skills as well. I've spoken to a customer who doesn't understand code at all and now they basically keep talking to me about oh, I understand, this code can do this and this and that, which is amazing to see, because they use natural language to understand things that they do not understand. So it's upskilling, it's giving them a healthy workplace as well and it's giving them that bandwidth to do more for the business.

Fraser Dear:

Yeah, I totally agree, johan. That ability to say I know what the right question is, but I don't know who to ask you of. Copilot gives them that. It gives them the ability to say this is what I want, how do I get there? And copilot will help them unpick that with whatever it is, whether it's something like I don't know how to get excel to do this right, right the way through to you know, I'm building a NET application and I need this code refactored from Java 8 to Java 16. It can do it, and they might not know how, but Copilot can help them get there.

Peter Filitz:

Yeah, so helping upskill the workforce without necessarily having to invest significant amounts of time and money into further training and hiring, you know, hiring additional staff and expertise, which is great. So, looking ahead, obviously things are changing and moving at such a rapid pace. What do we anticipate the next six months has in store for us and our businesses that we work with?

Fraser Dear:

If we had access to the Microsoft roadmap when it comes to the agents they're going to deploy, I think we would be a very, very happy bunny. But the reality is that the pace of change and Microsoft delivering exactly what they said they would do, what I would expect to see is a proliferation of brand spanking new high-end, a-class agents coming into the Copilot experience. You know, we've got researcher, we've got analyst. What's going to be next? I wouldn't like to predict, but whatever it will be will be advanced, it will be powerful and it will be game-changing, I think, on the Copilot platform. That's what I would say, johan.

Johan Venables:

Well, I was speaking with Peter actually earlier on the call and I did have some sort of prediction, but we felt like, no, let's wait and see what happens first. Don't want to scare everyone, but I think it's going to be awesome. You know, at the pace that it's moving, it's going to be another game changer, that's for sure. Just to give you an idea, I predict it could be an automated AI that's going to help and support you with specific tasks without you actually asking it. I think that's where it's going to go, so you'll have an assistant that could sort of almost think to itself but then guide you through your daily tasks and making sure that it's there to assist you.

Fraser Dear:

Yeah, it's the proactivity rather than the reactivity.

Johan Venables:

Exactly, yeah. So instead of steering it, it's going to. It's there to assist you. Yeah, it's the proactivity rather than the reactivity. Exactly, yeah. So instead of steering it.

Fraser Dear:

It's going to just be there to support you Almost. Co-pilots prompting the user rather than the user prompting co-pilot.

Johan Venables:

Yeah, autopilot.

Peter Filitz:

Yeah and gentlemen, how should businesses prepare for this deeper AI integration into their operations?

Fraser Dear:

To me. It's you, you know. I got asked this question the other week a quite large event and within that event, there was a number of different customers customers who were already on the journey and customers who were not. Anybody thinking, oh, this ai thing is going to go away, it's just a fad, I'm sorry to say it's. It's here to stay. It's that next evolution in business applications and in productivity. So if you're not on the journey, get started and get started soon, because you will be left behind.

Fraser Dear:

I think the other thing that businesses should also really be considerate of at the moment is about their own internal strategy and documentation and support for users, because what we're seeing now is you know yourself. You just have to look at LinkedIn or whatever platform you're on. It's like well, I know that was written by AI, because you can tell the difference and I think you know things. The kind of foundational pillars that Microsoft put out right at the beginning accountability, fairness, transparency these are fundamental and I think if businesses are using AI, they should be clear about that, they should be transparent about that and they should be informing their customers that we're using AI to respond or whatever the interaction may be. But further they need to kind of ensure that their team members, their business users, understand that this is not magic, it's a predictive model based on the information it knows. And so, again, we're back into that kind of conversation, that internal understanding of what's actually going on with these tools.

Fraser Dear:

You know, I think that is really really important. Bias understanding where the bias is originating from is really really important. And, of course, guess what Human in the loop? Just because AI has written it doesn't mean to say it's right. So you know, fundamental understanding and guardrails around the organization. No business would input a brand new business system without thinking, well, how is this going to impact our experience or customers or whatever the business is doing? This is no different, and so the same level of governance, if not a little bit more, because of course this is, it's an all seeing potential power. You need to build governance and security and control around this, as well as change, as well as transparency, as well as transparency, as well as understanding, testing, validating that output and response.

Johan Venables:

Absolutely and constantly. Also, monitoring and auditing on what AI tools people are using is very important as well. So it's not just what you've got in house, what you've built yourselves and brought into your business applications, but identifying what people are doing, making sure people are aware there's a process in place, there's a policy for the company on safe use of AI. If there's a product or tool that you feel that will boost productivity and will help you in your workplace, run it past your IT department. Have the conversation with them so that they can do all the checks and make sure it's a safe tool to use, because there's so many AI tools out there that I'm losing track. Every time I have a conversation with someone who's techie. They talk about a new AI app and sounds all good and great, but what does it do at the back end? How is it governed? So that's also very important.

Fraser Dear:

I mean one practical piece of advice. I kind of say under my breath but kind of mean it If an organization has an IT team who are a little bit skeptical about the tool, give it to them and say go break it. Be the red team, go and try this tool and play those ed cases out in a secure, safe environment with people that understand what they're attempting to do and see if you can get at it, because actually that's a real practical way of bringing those people on board and also starting to understand what the parameters and boundaries of the tool are within your organization. Excellent.

Peter Filitz:

Some really wise words there from our two experts today. Gentlemen, I think that brings us to the end of today's conversation around what we've seen over the past six months and what we've got to look forward to. I think it's fair to say a lot of change is afoot and a lot of really exciting progression in this space, which is certainly revolutionizing the way in which we work and our clients work. Please do stay tuned to our podcast series. We'll be doing a deeper dive around the AI adoption journey on our next episode. Please don't forget to visit our website, bcncouk. There you'll find a wealth of information around the products and services we provide and how we can help you and your business adopt and take advantage of these new technologies and services. Thanks so much, gentlemen, for joining us today. Look forward to seeing you on the next one. Thank you, Thanks so much gentlemen, for joining us today.

Fraser Dear:

Look forward to seeing you on the next one, thank you. Thanks so much.