Watch the webinar here.

Transcript

Alex Graves

Hi everybody. Welcome to today’s webinar session. Great to have you with us. I can see you’re all just joining from the lobby right now — 41 attendees. Brilliant. This session is going to be about 45 to 50 minutes long, and we’ll be having some Q&A at the end. We’re joined today by Simon and Sarah, so we’re just going to do some quick introductions. Sarah, I’ll hand over to you for an intro.

Sarah Meurer

Hello everybody. My name is Sarah Meurer. I’m currently the Vice President of Global Internal Communications at Elsevier. I’ve been delivering internal comms in one form or another for about 25 years. I’m very lucky at the moment to be embedded in an organisation that’s been using AI for 20 years and is very much at the front of embedding AI in our products and thinking about how we use AI for efficiency and productivity. Excited to share some of our thinking and learning with you today.

Alex Graves

And Simon, over to you.

Simon Andrew

Hi everyone. My name’s Si. I look after employee engagement at The Surgery — we’re an internal comms design and digital agency. I realise I’ve been doing employee engagement for 20 years this year, which makes me feel incredibly old. In terms of AI, last year I took about a dozen teams through some basic AI training in IC, just thinking about how they could be using it, how they get those first steps going, and I somehow ended up as AI guy. It’s improved my gardening, taught me how to do photography. It’s become a way of life for me, and it’s great to be here talking about it with you.

Alex Graves

Thanks both. And I’m Alex Graves, Chief Visionary Officer at Silicon Reef. We’re a Microsoft partner specialising in all things Microsoft 365 — and you can’t say the word Microsoft now without thinking about Copilot and everything that’s going on in that world. I’m here today to give you the breadth of understanding of what I know about how Copilot is changing the world of internal comms, and the conversations I’m having regularly about the intranet.

That leads me nicely into what we’re going to cover today. Six main points of conversation, and we’re going to try and rattle through as much of it as we possibly can. We’ve even got some live demos. With the demo gods being with us again — Simon and I have both had little hiccups this morning — some of you may notice that Sarah is vertical on her phone because her camera stopped working on her laptop today. She’s got set up with her phone in between a couple of bottles, I think she said, in a very professional standard way. But great that we’ve actually got here. Hopefully everything else from a technical perspective works with us today.

The Q&A is open, so please use the chat — thank you. Somebody has just done a laugh reaction. It’s good to know that you’re there, so please use the reactions throughout the webinar today. Use the thumbs up, use the heart, use the surprise one with the mouth open — that’s one of my favourites. Hopefully there’ll be a few surprises, all of them positive and good. Thank you everybody.

Right, let’s kick off into the first topic. To set the scene, the structure will be conversational. I’m joined today by Sarah and Si, both experts in their fields with lots of experience. It’s going to be a combination of slides to talk through and frame the conversation, as well as demos. Please use the chat to get involved as well — it’s a really interactive session. I’m going to start off by inviting you, Si, to talk about some of the insights you’ve learned from The Surgery around AI confidence.

Simon Andrew

Awesome. Would you mind jumping on the slides, Alex, and I’ll explain a little more. When we’re talking about progressions in AI and hopefully some exciting things we are and could be doing today, I wanted to take a moment first to think about employees and where their heads are at, before we get too excited and too far down this route.

We know from psychology that decisions are driven by emotion. We like to think we’re all super logical and rational creatures, but we’re all irrational and emotional — that’s just the reality of it. So we need to think about where people’s heads are at and ultimately meet them where they are.

I like this research because it came out from Pew last year, right at the end of the year, so it’s very recent. Several thousand people across 25 countries, and what they found was that people are anxious about AI. You have a small group of people whose excitement outweighs that — you’ve probably got three of them on the call today. But most people, even if they are excited, have got anxiety alongside that. Knowing that’s where our people are at, we need to think about how we help them, how we get that conversation out in the open, how we address the elephant in the room.

In IC, there’s pressure in terms of supporting the move to AI and taking people on this journey. Let’s start with how they’re feeling. If I think about what we did at The Surgery — bear in mind, we’re a smallish agency, so it’s perhaps a little simpler for us — we sat everyone down and had these open conversations. We said, how are you feeling about AI? What are you worried about? What’s going through your head? We could talk a bit about what the business is thinking about, give some reassurances, but just give that forum to let the steam out, let the pressure off a little. Off the back of that, we set up some communities so people can share where they’re learning and the things they’re learning together, as well as connect some of those advocates with the people still dipping their toe in, so we can get everyone joined up. It’s an important place to start — that head space we’re coming from.

Can we move on? Lovely. In internal comms, we’ve got this unique place where we’re not immune to this anxiety and worry. I like to think everyone at some point has been worried about what’s going to happen with their job with the speed everything changes. When me and my friends get together, we joke about setting up a commune and have an argument about who’s going to be the baker versus the carpenter — I think I’d be better with the wood than the stove, but I’ll try and handle whatever. We do have this anxiety, and we have it whilst we’re having to think about how we support the organisation through this. The stats on screen here are from a slightly smaller study — I think the Institute of Internal Comms Index last year put it at about 48% of IC professionals worried about job future and anxiety. The majority of internal comms professionals are in a space where they’re hearing from the business, “we want more people to be using AI, we need to advance ourselves in this area, you guys are key to helping us move this forward.” But then there seems to be a bit of a gap in terms of, well, this is how we’re doing it, here are all the tools, here’s the laddering to take you forward. So it makes it quite a difficult space to be in. That’s just a little bit of context setting. Sarah lives and breathes IC and can pick up from here and go on for us. Sarah, can I pass to you?

Sarah Meurer

Yes, of course. We’ve very much been on a journey. Just a little over 18 months ago, we introduced an AI tool that was a self-created, in-house tool that was sandboxed and enabled people to put confidential company information in and start playing with the concept of using AI on a day-to-day basis. Since then, we’ve introduced many more tools — even in the last week, we’ve introduced Claude. So we’re very much on a journey.

We also have around three and a half thousand technologists in our business, in a 9,000-size organisation. We have a part of the organisation which is very much about publishing and editorial, then we have operations and functions, and then a sales team. So we’ve got a real federated organisation — people want to use AI in different ways, but also have different experience of it. One of the things we’re looking at is how we very carefully navigate the risk of a two-tier workforce.

When we started to introduce AI at Elsevier, our leaders talk about things like the benefits — things are going to be faster, smarter, we’re going to be more productive, more efficient. But in reality, employees are probably thinking more about: is it safe? Is my job safe? You touched on this, Simon. Am I being monitored? What’s the expectation? We talk about three things being integral to how we take our organisation on this change. Creating clarity — what is our expectation of you? What are the guardrails? What are the tools available for you to use? Confidence — your ability to know how to navigate and use the tools. And capability — day-to-day, how do I prompt? How do I think about building agents? All three of these things are required to build a high level of trust.

Next slide, please. When we look at what’s happening with AI inside organisations — and Elsevier is no different — some are racing ahead, some people are more sceptical, some people are quietly watching, some are pretending it’s not relevant to them at all. There are few and far between of those at Elsevier, given we’re a technology organisation. As a result, it’s starting to feel a little chaotic, and that sense of chaos does have quite a predictable pattern. That’s important for us as internal communicators to recognise and start to think about.

Next slide. I want to show you something called the Rogers diffusion of innovation, and this applies to many things we’ll look at. If you think about your employees in this way, it’s really helpful because you can be more nuanced in the way you approach things. We have innovators — these are people who are experimenting. They’ve probably got the tools themselves and are already using them. I met someone last week who’s been using Claude for over a year, even though it’s only just being introduced to the business. They’re well ahead of many people who are just starting to use it for the first time.

Then you’ve got early adopters. They’re really curious, they can be massively influential, and they’re watching how the leadership navigate this. They’ll move, but they want permission and they’ll do things within the parameters of the organisation. Then you’ve got the early majority — they’re the ones who are very cautious of the hype. They’re asking, is it safe? Am I doing the right things? Do I look foolish? How do I compare to my colleagues? You’ve got the late majority who are generally overloaded. They’re quite process driven, quite risk sensitive, and maybe they’re thinking, “I’m going to wait and see where this goes.”

And then finally — and I hate this word — the laggards. I think that’s a really unfair word, but they’re often resistant and feel slightly exposed, maybe aren’t confident with technology. The percentages here won’t be far away from what we’ll see in other organisations. So it’s a great way to think about, for example, when you’re thinking about training or development, how do you cater for these different groups? And how do you think about peer-to-peer learning, buddying up, sharing that experience?

Next slide, please. We have a responsibility to be really clear about what AI is here for, for our organisation. What are the guardrails? What are the red lines? Where should it be used, and where should decisions remain human? That’s a really important point, because a lot of people fear this sense of: is AI going to take my role? Is my role going to change? But the human element is incredibly important — it’s just about how that evolves and the way we think about it.

Next slide, please. AI-enabled culture is going to be the biggest change programme we will ever deliver in our careers. It is absolutely enormous when we think about the skills upgrade, the technology introduction, the pace of change, and how an organisation has to adapt its processes and systems to adapt to the pace. For example, we were talking last week about how we might start to think about going from annual financial cycles to eight-week financial cycles, because the process of innovation, development, testing and delivery to customer is becoming so much shorter, and that will continue. We need to take a step back, think of it as a programme of activity, think about that change curve, and understand that it’s a long-term process.

Next slide. We’ve got to be careful within this whole programme of change that we aren’t creating this two-tier workforce, where we’ve got AI adopters who are curious, experimental, productivity-focused, thinking about how to task AI and how to make their jobs more efficient — versus the AI non-adopters who are more sceptical. They observe, they don’t necessarily engage. So how do we take everyone on that journey with us? One quote I’ve heard many times, I’m sure you have too, is this idea that AI won’t replace humans — humans who use AI will replace humans who don’t. We talk about that a lot at Elsevier to really drive that sense of curiosity, but also urgency. That’s it from me.

Alex Graves

Thanks, Sarah. Both you and Simon have framed the rest of the conversation really nicely — thinking not just about adoption of technology, but the importance of that change curve and identifying those people in the Rogers diffusion of innovation who are the laggards, who just aren’t ready for the change. But that quote encapsulates that you do need to be ready, and you need to be on the pulse for it right now.

I’m going to shift gears a little and talk about the relevance of AI and the intranet today, because many of you watching this will be internal communicators who are perhaps responsible for looking after your intranet. We have to start with the simple truth that, despite all the growth of collaboration tools we’ve seen in the last 10 years with Teams and Viva Engage, the intranet is still the front door to most people’s digital workplace. No matter what people want to understand — whether it’s finding out what leadership is saying, or finding information they need to trust and be verified — they typically turn to the intranet, that front door to your organisation.

The problem people are facing isn’t relevance. It’s the expectation that “I don’t want to come to the intranet now just to find out news and policy information. I want to come there for purpose — to find an answer, to find something simply and easily, and to help me navigate around my organisation in the simplest way possible” — whether that’s in Teams where people are working. What’s starting to happen is that the intranet is having more relevance than ever before, because it’s a trusted source and a trusted location — and it’s more than just a website. In this hybrid world we now work in, it’s the entry point where you might not go to the office anymore. It’s that anchor where you go to find real information about your organisation, but you feel connected to it. The relevance of the intranet today is more around: “I’m going here to help me connect with my business.”

And how do AI and agents come into that? That friction, that barrier of helping me do something and get something done — AI and agents can really address that. I’m fortunate that I get to speak to many internal comms people about how their intranets are performing. The things I typically hear are the same again and again. Employees go there and they can’t find what they’re looking for, or everything they find is outdated. It’s too complex to navigate. “I don’t know where I am.” So how can we change that experience and change the lens to help people navigate through faster, giving them a better employee experience when they land on the intranet, so they can say: “I know what I’m coming here for. I’m here to find some information — where can I get that done?”

That blend of productivity with employee engagement is starting to have more and more prominence. I like to think of that when I think about a journey from a standard built intranet — where the core purpose is content and publishing — all the way through to the right-hand side, which is a truly engaging digital workplace. What I’ve highlighted here in that row about AI agents and their role: you might start simple — helping me find information and helping me summarise content on a page, all the way through to more employee experience: “I’m coming here to help me navigate my way around and find information, but surface some insights from perhaps another platform, maybe from SAP, maybe from Oracle, whatever it might be.” And then through to that far end of the digital workplace solution, where things are more orchestrated, execution is more seamless, and agents can actually complete the tasks. It already knows what you’re coming there to do in the context of how you work and the navigation you’ve taken to find that agent.

Something I want to address: start simple with Microsoft Copilot. Many of you hopefully will have experienced Copilot and tried out Researcher and Analyst. All three of those, if you rewind the clock 18 months, were painful in some ways. I thought it was a load of lies that it was bringing back to me. Fast forward to today, and some of the things it’s bringing back are a heck of a lot better than they were then. We’re now ready to start experimenting, and many organisations are moving beyond experimentation into the agents that are ready to help you do that self-service and help you come and get something done. These are the sorts of agents that can live within your intranet.

On the far right-hand side, I think these are more digital workplace solutions, where you’re thinking about a custom-built solution that’s perhaps connected to multiple different systems. It may be that you’re on a journey to get there. But the reality is, all of these things can be accessed and launched from your intranet or within Teams — however you need to access it. That evolution from an intranet being a static workplace of news content being pushed out — most organisations I speak to, no matter what size, whether enterprise level or a few hundred users, are starting to think more and more about how agents sit within that.

We’re going to move into a discussion now with Sarah and Simon. Our experience: what have we seen within organisations that helps prioritise and identify when AI is right to step in? Sarah, you had a slide that talks about that — it’s not just for AI’s sake, it has a role to play, it’s coming in and actually performing a task. And Si, we’re going to do a bit of a demo in a bit. How have we seen this evolution of finding the right places where AI can make a difference? Where have we seen it perhaps streamlining processes?

Simon Andrew

I’m happy to talk to that first. For me, it typically turns up when you feel yourself doing a task — I don’t know if repetitive is the right word, but you can see the task is structured in such a way that it’d be easy to get someone else, or an agent, to do this with the right instructions. An example from the agency side: when we finish a project, we’d typically create a case study, which has a particular format to explain why we did this project, what we did, and then what the results were. We’re doing this all the time, so why don’t we build an agent? We’ve got an agent that we can essentially drop all the documents from the project into. We can click on the little mic button, just talk to it for a minute about the project, and it will output this perfect case study as a result.

You’ll typically find in your daily tasks there’ll be things where you think, “Why am I doing this every time? We’re following a similar format, similar approach — I could get an agent to do that.” That for me feels like that natural first step: get it to take the tasks that free up your time and give you a little more strategy space.

Sarah Meurer

I completely agree. Even over the next couple of months we’re doing a lot of work to automate some of our workflows and start to think about how we can take that level of transactional work out of our day-to-day. It’s also very much helping us think about our writing and communications. We use a technique called Smart Brevity, which is from Axios HQ. We’ve done a lot of work with them over the last 12 months to help us be more concise in our writing. Combining that style and tone with, for example, mass information about our strategy delivered through different scripting or particular presentations, enables us to continue to use really authentic, humble, on-brand language, but also make sure we’re hitting all the messaging. That’s really useful.

Right through to — the thing that has saved me hours and even weeks — mass data analysis and insight-driven findings out of mass data, even finding things you as a human might miss. But also making sure you’ve got that human insight to always look over that qualitative information and see if there’s any nuance. Human oversight continues to be super important. So many use cases. And it does beg the question — if it’s enabling all of this, what skill set do we need in the future as internal communicators to enable that? That’s really interesting to me.

Simon Andrew

The data one is huge. Having done lots of employee insight data analysis, it’s absolutely a game changer. But another area that’s less obvious and super useful is where we’re creating things like employer brand guidebooks or even IC strategies. Typically these things live in a folder somewhere and we reference them at key times, but rarely use them in every decision day-to-day. Using agents to own those things can be super useful. If it’s your strategy, your employer brand book or whatever, you get an agent to essentially own that document, and then when you’re putting in your draft communications or plans, campaigns, thoughts, how you’re going to approach stuff, you run it through and it validates against that, corrects you, gives you that steering to make sure you’re reinforcing all the points you spent so much time and effort to determine. You’re really aligning to your strategy and taking it forwards. We can take those things that often live in folders and make them almost living and breathing by putting them in an agent.

Alex Graves

One thing we’ve heard there is that agents are useful pretty much everywhere. And what I’ve seen is that agents for an agent’s sake is not the right approach. We’ve moved beyond that world of using AI just as an FAQ — “here’s a question, give me an answer back.” The world we’re starting to move into — Sarah, I’m going to invite you in in a minute to talk a little about your course you went on last week and what you showed us earlier in your menu — one thing that’s becoming clear is that the growth of agents and how they’re evolving is only going to grow exponentially. From the International Data Corporation, their data provided by Microsoft, is 1.3 billion agents to be created by 2028.

Phase one felt like the introduction of Copilot. If you were an early adopter, or you picked up your phone and used ChatGPT to have a conversation, it was great — “this is a thing that can really help me get stuff done.” Different generations are discovering that phase one experience. Where we’re starting to evolve last year and this year is having human-plus-agent teams, where we think: “OK, I do this role, but I need to be able to delegate some tasks to get stuff done.” There’s some autonomy that can be done here — there’s a task we regularly do. What you said about number crunching with the vast amount of data that may have been done by a person — we can get that done faster and more efficiently with an agent.

Sarah, it’s interesting what you’re starting to experience, and the question you asked about AI strategy. You and your organisation are likely in phase three of human-led but agent-operated. Sarah, can you talk a little about how you’re evolving that with Elsevier, and your experience over the last week?

Sarah Meurer

Very, very early stages, Alex, so I come with the caveat. But I was lucky enough to attend — we had a senior leadership group meeting last week, top 150, and as part of the agenda we had a two-hour training session with Anthropic. Claude was then introduced to that community and will be rolled out over the next couple of weeks across Elsevier. We were asked to experiment with it, and it was very interesting where people started.

One of the things that really caught my attention was — I think you talk about it, is it Frontier mindset you mentioned, Alex? Is that right? It’s just where people start in terms of their thinking. Some people went to redesign a process. Some people went to do something basic by turning some thinking into a PowerPoint deck. I decided to go big last week — I inputted my internal communication strategy for 2026 and asked it to build me a team of agents that would have the perfect job titles and roles to deliver that strategy for me.

It produced something quite incredible — a playbook for an AI agent team. There were nine jobs in my team, and it created a full profile for each job, plus a prompt for that job. I was then able to upload reference materials, depending on what the job was. For example, one of my agents is an IC insights and measurement agent — I’m able to upload our internal comms dashboard. It can also input data from our always-on listening tools, so we’re able to input data from Office 365 and from our eNPS scores, as well as survey materials.

That’s just one example. On a monthly basis it can develop a routine insights dashboard for me — it just produces that and provides insights in a way that we haven’t traditionally had the time or resource to do. It’s been an incredible experience, and this is only last week. So already I’ve created a campaign, I’ve created a measurement dashboard, I’ve redeveloped some of our thinking around our learning lab strategy, which is our commercial acumen strategy for next year. It’s created an entire spreadsheet for me. Just this morning, in 30 minutes, it turned all of our feedback from the event last week into an insights document, 20 pages long. I’m just experimenting and it’s blowing my mind.

I’m already seeing mass efficiency and high-quality outputs in a way that’s really surprised me. Yes, I’ve had to go through everything, and yes, I found things I want to change — but I’d say less than 10% — and the time it’s saved is incredible. So it’s started to make me think, really, my role starts to be about what the art of the possible is, my imagination, my creativity, my ability to make judgement calls about the quality of the output — in a way that traditionally I might have had to go through reams and reams of data and analyse. It’s very much changing my role. It’s very exciting, but I’m very much at the beginning of the journey, Alex.

Alex Graves

What you’re describing there is human-led. As you said, 90% of it you’re happy with, but you’re still leading it — still reviewing everything that’s the output to check it for quality. But you’ve delegated all of these tasks and capabilities that need to be done. I’ve also done a lot of reading over the last 18 months that — you’re starting to see those agents now in some organisations appearing on organisation charts. You’re going to become an agent manager of these agents — they report into you, Sarah, and they get these things done. They’re part of your org chart. I’m not saying they’re going to have names and personalities, but they have a role to fill, and they’re part of the team.

In the next 18 months, that’s something that’s going to be seen more and more within organisations. Phase three, human-led agent-operated, sounds like Sarah is starting to get there. But many organisations I’m still speaking to are getting ready for that — they’re in phase one, perhaps moving towards phase two of agents helping them complete tasks, but not quite so autonomous and orchestrated in the way Sarah just described, where your agents, Sarah, may even be working together, having conversations with each other to get stuff done. That’s that Frontier evolution I was thinking about.

Great. Please keep any questions coming in — use the chat, use the reactions. I’m going to stop sharing this presentation and move into a demo. To set the scene before I do that: what I’m going to show you now is a live campaign demo agent in SharePoint. I’m linking it back to the intranet experience — how you can create what is perhaps the lowest barrier to entry for an agent. I’m going to run through that, and hopefully, again, the demo gods are still with me, and we can run through this nice and simply. Then I’m going to hand over to Si to share his screen and do his demo afterwards. I’m just going to change my screen sharing.

What we’re looking at here is a demo SharePoint environment. It could be your organisation’s intranet. It has some news at the top, news feeds pulling through, some links to getting stuff done — typical layout and structure, with some carousels and some Viva Engage pulling through. Where it’s slightly different here is: yes, we’ve got the search in SharePoint at the top, but I’ve also got the ability to open up an agent on the page and start having a conversation with my agent about “how can you help me?” Typically, as I said, people come to the intranet looking for a policy. So rather than navigating through our policies or finding it in SharePoint, I could just come in here and say, “What’s our policy on remote working?”

This may take a few seconds to work through, but I haven’t had to go and find that — I haven’t had to click through loads of steps to navigate around. From an experience perspective, when I think about how people work now: I pick up my phone, ask it a question, and it gives me an immediate answer. Why should the intranet experience be any different? Here it’s bringing me back a set of information around our ways of working and the policy. It’s applying it to me — who it applies to — and it’s breaking it down in a structured, consumable way. It gives me a link to say where that policy was located. It tells me when it was created and who it was created by, and the reviewing authority. So I’ve got loads of information.

And I trust this — I know when it’s effective from, when it’s being reviewed. I can trust that this is the source of truth for the policies within my organisation. That’s just one example of how the AI-powered intranet is very simple to create, just by dropping an agent into a web part on your SharePoint page.

Another scenario I want to run through — the agent is Copilot. I just saw somebody demoing there. I’m logged in as somebody called Amelia Earhart, and she has a Copilot licence within the organisation, which allows her to experience this. That was a good question I just saw pop up. I just want to make that clear.

Another example of an agent where it can be useful, particularly for internal comms: we’ve got another part of our organisation running an Energy Week. It might be a campaign, a piece of awareness we’re running. It’s complemented by this SharePoint site here with all the things we want to provide information about. It looks and feels like my organisation. I’ve got stories around other green initiatives we’ve got going on. I’ve got the ability to see some content within here as well — nice carousels and layouts within our SharePoint campaign site. But I’ve got my Energy Week agent here. It’s there ready for me — I don’t want to have to read all of those things, so “how can I get involved with the Energy Week?” I can either talk to it by clicking the icon, or just ask it a question. Again, it’s coming back saying, “This is how you can get involved.” I’m going to leave that running in the background, and in the interest of time, skip over and show you how easy it is to create this.

Everything I’m showing today is using SharePoint. Many of you will know a SharePoint document library when you see one. What we’re going to do now: because Amelia has a licence for Copilot, I now have an option to create an agent, and the agent here is simply grounded in this library. So: AI actions, create an agent, and I can start describing — it’s going to be my Energy Week agent. I can now start to build out this agent. What’s the purpose of the agent? Where’s it going to get the information from? Because I created this agent inside this library, it’s grounded within this campaign week. The data it uses — the source of truth — is within this SharePoint site, within this document library.

I can give it some instructions: “This is what I need to do.” The starter prompts you see on the right-hand side are what it typically has. I could give it more instructions, but I could have gone through and created an agent — the parameters it works within, the tone I want it to come back with, an informational tone, telling it it’s targeting perhaps students or staff within the organisation. I can click create on that agent, and I’ve created my own agent for this campaign specifically — not my whole intranet agent, just the campaign agent here, in about three minutes.

All that means is I can then have an agent that sits within this document library, and it allows me to do exactly this by adding a web part onto the SharePoint page — which many of you will know how to do. It allows people to start to interact: “How can I get involved?” And it’s providing answers back from the PowerPoints, the PDFs, the Word documents I’ve uploaded into that library. It’s telling a student how they can get involved with the Energy Week. That’s my simple demo. Think about the scale and complexity here — I’m starting simple. This isn’t a fully orchestrated, big-scale agent that’s going to have any automation behind it. It’s a simple entry point into how you can start to use AI within your intranet to help provide information. Si, I’m going to hand over to you for your demo. I’ll have a look at some of the chat questions while I do that. I’m going to stop sharing.

Simon Andrew

Awesome, thank you. I have a low tolerance for reading lots of information, so that agent totally appeals to me too. Let me share my screen and I’ll explain. I wanted to share an agent we set up recently — it feels like a super practical one that anyone on the call could go away and start orchestrating themselves. As Sarah said earlier, it starts with that vision and ambition.

What we wanted to achieve: we know that employee feedback and employee co-creation create the best communications — make sure it lands well. Lots of organisations have groups of employees set up for this, whether it’s an IC network or an employee experience squad. Maybe you meet them once a month or every couple of weeks. They can check in on your strategy or campaign, review some materials, give you really good feedback to iterate, validate and improve, to make sure you get to the best possible place. I don’t think AI is ever going to replace that actual employee co-operation and insight. But what it can do is give us some of that in the room at a moment’s notice whenever we need it. When we do need to make a quick decision, when we can’t get employees in, we could use AI to give us some of that employee-centric view and help us do that.

The way we do that: we look at an organisation, we look at their audiences, and we ask, who are the key audiences in this organisation? Let’s build some insight into each of those. Let’s create a persona that talks about their goals, their frustrations, their needs, etc. Let’s see what data will supplement that, and let’s build that into an agent. So this agent knows there are five key populations — we’ve got data on each, we’ve got an overview. It can give us insights from all of those.

What I’ve got here is a demo, based around three fictional personas with some fictional data — we don’t want to share anything sensitive for the sake of this. I’m going to give it a prompt and show you what I mean. This is a typical Copilot page, by the way, so hopefully you’re fairly familiar. I’m going to say: “We ran a town hall on AI last week, and we want to help managers talk to the employees about it. Help us do this across…” The great thing is it doesn’t care if I put loads of spelling mistakes in there — that’s always a win.

What it’s going to do is reference the personas and the data we’ve uploaded. It’s hinged off a SharePoint site, so we’ve got a folder for each of those different populations. Built into this, we’ve got frontline employees — people who are looking after customers, may not have email addresses, etc. We’ve got project managers and change managers, people who are in the “how do we do it?” And we’ve got an expert community — those who want to really understand the why and have their chance to challenge and be a part of that journey as well.

It will review across all of those. We’ve given it a fairly strict way of playing back any prompt, any insight from the prompts. So it gives a little “here’s an overview of what you just asked me, here’s an insight from each of these different personas — here’s how the employee might react, here are the things you need to think about.” Then it rounds it all off with a “so what” — when you’re thinking of your plans, here are the risks to be aware of, and here’s what you need to do as a result.

You’ll see I’ve asked it: “OK, how do we engage managers following a town hall on AI?” It’ll give me that playback — “OK, here’s what you said you wanted to do” — so I can make sure it’s understood correctly. Then it gives me that playback by audience. For frontline employees, it’s identified: time-poor, often away from desks, can be highly sensitive to practical impact. It highlights the potential anxiety and worry we might have around this population, and it’s pulled out from some of the data we’ve added in terms of their preferences.

It’ll go through and do this for each group. For the project management office, it’s about the alignment between comms and delivery — that’s where their concerns are. For expert communities, it’ll be slightly different. You’ll all have totally different audiences, but hopefully you can see the application. Then it says, here are the things we’ve got to look out for. For frontline, one of the big things is job threat. They feel they might hear it late or second-hand — not all of them might be able to attend this town hall, for example. For PM, it’s the lining up of the delivery — what happens when the delivery isn’t ready for all this talk, etc.

Then it goes through: here are the opportunities. Here’s how we can use managers as the sense-making layer — show that we’re ready as that translator, give the PM a pack they can use consistently for decision logs, etc. Some really practical options for what you can do as a result. In the interest of time, I’m not going to go through all of these, but hopefully you can see the relevance and where it can add value, giving us that lens from the employee perspective based on data we’ve given it, and ensuring that our thinking is employee-centric.

There’s one more prompt off the back of this — I could follow up and say, give me 10 questions each audience might ask. If we wanted to prep managers for some follow-up conversation, this will go through and, based on all that data, give us 10 likely things you’re going to hear from each of these groups. So we can pre-arm them in the scenario. For frontline: “What does this actually change in my day-to-day work? Is it going to affect my job?” For the PM: “Who owns this initiative? What decisions have already been made?” Very much thinking of the delivery aspects. For the expert community: “What’s the rationale? Help me understand this. What’s the evidence there?” You can see how it takes all those different views.

We can use it for saying, “We’re going to work through this change plan — how might these groups react to it? What are their likely questions? How can we improve these comms?” We could drop in a planned communication draft and say, “Help make this work for our expert communities.” We could give it access to a leader who’s done a presentation for frontline and is going to go out and speak to the PMO, and say, “How do I make this presentation work for PMO, knowing that they’re different from frontline?” So in all of these scenarios, we’ve got almost an employee voice in the room helping us keep things employee-centric. A lot of it is stuff we might know and might do in the majority of situations, but being able to do it 100% of the time, with essentially 100% accuracy, at a moment’s notice — that’s where it becomes super valuable. I hope that makes sense to everyone. Any questions in the chat — we’ll do our best to pick those up. I’m going to stop sharing my screen, Alex, and pass back to you.

Alex Graves

Thanks, Si. And thank you everybody for the questions coming in. I’m doing my best to work across the chat and the Q&A. I’ll pick up on some of them towards the end of the session. Do keep the questions and chat coming in. As mentioned, anything we don’t get to answer today, we’ll pick up afterwards and share along with the recording. Sarah just messaged to say she’s lost video, but hopefully you can still hear us.

Sarah Meurer

Yep, still here, Alex. Still here.

Alex Graves

Great, thank you. Moving on to the next section: some real-world agent cases. This may answer some of the questions coming in around the ability for agents to search across other platforms, and somebody asking about SharePoint — does it only search SharePoint pages? I wanted to give you some real-life scenarios where agents can be created, particularly for internal comms. What I’ve got here on the left-hand side, to keep it simple, are things like helping you create a tone of voice. You can give the agent some instructions, provide some materials — it may be leader comms you’re working on — give it some information, ground it in the context of what people have been talking about, how the leader talks. Those are the very simple agents. They just need some data to give you the output.

As you move along — I’ve just shown you a campaign agent that was simply grounded in SharePoint as a SharePoint agent. On the right-hand side, you start to see things like more employee self-service and more advanced ones. Somebody asked the question — Carmel, you asked about searching across other platforms — this is where the agents are positioned to be part of that digital workplace experience, pulling in information from things like ServiceNow, or whatever it may be — that cross-platform or even cross-team information sharing. You can connect a Copilot agent out to other systems, reference all the data and pull information back in. Absolutely, you can do that. However, you need something called Copilot Studio to do that. As I mentioned in the Q&A, it’s likely to be someone in IT who needs to help make that happen.

All the way through to customer support agents, where you can think about having an agent that’s on your website — that again can be built using something like Azure Foundry — where it can address tickets and issues people raise. It gives you the ability to have far more complex integrations with other agents, and that concept of agents working with other agents — absolutely something your agents can achieve. It just needs a bit more mapping out and complexity. Everything I’ve shown you in the demo, and that Si showed you, you can build out using natural language — you just talk to it and it can build that agent. Anything that involves plumbing — connecting to other systems, connecting the pipes — that’s Copilot Studio, and likely where you need to get IT involved to help you do those things.

All perfectly achievable and possible. I’m conscious of time — got 10 minutes left. How you might go about identifying those: a simple matrix you might want to steal — high-impact quick wins versus low-effort tactical improvements — thinking about the types of agents you might have within your organisation. What Sarah talks about — how she’s built an army of agents to help her make decisions and analyse data to implement strategy — you probably think that’s going to be a high, high impact. But the reality may be that the time and effort is low, because you’re just giving it a set of instructions, and all the long, hard work is already done because you’ve already created the strategy document yourself. You’re giving the information to AI. So the strategic bets may be faster to achieve than you actually think, because you’ve got a lot of the groundwork covered. Just a matrix to help you understand how those agents — from simple all the way through to advanced — can be mapped, and you can realise the benefits and the impact they may have.

In the last five minutes I want to cover off — as I start to think about leaning into the future. We talked a little about something called the Frontier mindset, and now that you’ve seen some of the demos and use cases, I hope it’s becoming clearer what some of those things you can think about using, moving into that intelligent intranet, moving towards the digital workplace experience. The points I want to cover off here — some of these are Frontier features from Microsoft, and Sarah and I both talked about that Frontier mindset: innovation. How can we lead it forward? How can we change from being reactive, just using AI to generate content — “create me a set of information, create me a new PowerPoint slide based on some images I give you” — how can I shift gears into something a bit more strategic?

Some of the new features rolling out from Microsoft right now: Copilot Cowork — I’ll just cover this one. Copilot Cowork is the newest feature from Microsoft. It was launched about six weeks ago. Sarah mentioned, and somebody also asked a question, around what’s the preference between using Copilot or Claude — or is it company dependent? Something you’re likely to start seeing across multiple vendors — and Microsoft has already started to do — is bringing in the models from Claude/Anthropic into Copilot. You’re starting to have the ability to have multi-model experiences through Copilot.

What Cowork does is take it beyond “I need you to give me back an answer” — this is evolving to “I want you to help me get stuff done,” either planning for things and looking at the context of your diary and what you’re doing. It knows how and who you work with using something called Work IQ, which I’ll cover very shortly. This is probably one of my favourite features in Copilot right now. It’s been released in the last 12 months, and it’s going to move to general availability probably in the next quarter. If you get the opportunity to work with your organisation to have access to Copilot Cowork, this is where you move beyond phase one into more of a phase two — “Can you help me get stuff done? I’m going to delegate a task to you. I need you to create a set of reports based on some information you know about me, and information you can find about me.” Then you can say, “Now, can you do this every week, give it to me, put it into Teams, send me a notification once you’ve done it every week.” And you can start to automate some processes using the same instructions every time.

That’s really the key difference between an agent and chat like you’ve experienced — an agent has a memory. It has the ability to know who you are, how you work. And Microsoft has something called Work IQ. The reason I’m telling you this is, it’s something you’re likely to start hearing — these phrases like “What is Work IQ? Why should I care? What does it mean to me?” You may have heard of the Microsoft Graph about 10 years ago when Microsoft started to talk about that. But Work IQ is the new thing that sits underneath Copilot. It’s the bit that knows who you are, where you sit in the organisation. It’s memory, but it can also understand why you’re looking for something, and the results it gives you back are far more accurate. The pain you had with Copilot perhaps 12 months ago — “ah, that’s not quite what I’m looking for” — you’ll have seen the models changing.

Even yesterday, they moved to ChatGPT 5.5 — that was announced yesterday — and Claude is now becoming free as part of Copilot chat within Microsoft 365 and Copilot. You’re starting to see this multi-model approach, which is giving far better data and far better outputs from the information you give it.

Finally, one thing I heard this week about Work IQ — think about your sat nav. Work IQ is the thing that tells you where the traffic is. It tells you what route you should take, what matters right now, and which route makes sense. Whereas the old way of working using the Microsoft Graph — that’s just the map. It’s static. It knows where the streets and buildings are. But Work IQ is the bit that helps you navigate around to get the smart work you need done right now, give you the answers you need. I wanted to give you a little glimpse into the future of what it looks like, and a couple of new things coming out that are likely to land with you in the next quarter.

Just before we go into final Q&A in the last five minutes — to help you think about moving from simple human interactions and conversations into more “getting work done.” This slide is meant to represent an evolution. But one of the things, Sarah, that you mentioned earlier about reskilling people for growth — and you’ve experienced this yourself in the last week — how important it is that you’re able to take an idea or concept from “I think we should have an agent, this is where we should use it, I do this task on a regular basis” to upskilling people to have that change of mindset. As one of the things you said in your early slides — it’s not just a skills and capability change, it’s a change of how I work and how I do things. Sarah, do you want to comment on that quickly, before we close out and I look at some of the Q&A — about the importance of upskilling people for future growth?

Sarah Meurer

Yes, absolutely. We’ve been on a bit of a journey with our development and training over the last 18 months — really starting at some of the basic levels of what is AI, why would you use it, how to prompt. The sort of task-based activity you can use it for, right through to really in-depth training now — and it’s a journey you’ve got to take people on. There’s a really interesting debate — and I don’t have the answer — about what does it mean for young graduates, or people entering the workplace, and how are we evolving?

I was talking to the Institute of Internal Communications last week about their training offering and how it has to evolve to think about the way we operate and deliver internal communications going forward. It is definitely a change of mindset in terms of thinking agent and human. It’s a change of mindset about being curious, constantly curious, having a learning and development outlook, not being scared to fail, take risks, try new things — which comes very naturally to me, but it doesn’t always come naturally to everyone. It’s just how we support that next generation and the skills gap they’re going to have. I’ve got an 18-year-old daughter and a 16-year-old daughter who are not allowed to use AI at school at all, so they’re going to come out of school not really having the experience of the tools in the way I’m being encouraged to within the workplace. That’s also a really interesting dynamic. We’re sort of finding our way, Alex.

Alex Graves

Si, any other comments from you before I quickly look at the Q&A?

Simon Andrew

No, Sarah summed it up nicely. In the interest of time, mate, feel free to crack on.

Alex Graves

Cool, I’m going to stop sharing that so we can see this now. We’ve addressed a lot of the questions in there. If I scroll down — the preference between Claude and Copilot is company dependent. Sarah, did you jump in and answer that one?

Sarah Meurer

I did. We currently have Microsoft 365 Copilot, integrated into our Microsoft tools — people use it, for example, to scan email or work with SharePoint. We’ve got ChatGPT Enterprise, which we use much more for things like content creation and knowledge retrieval. It complements company data and analytics. That’s really cool. Claude, which has now been introduced, takes it to another level again — it’s really thoughtful analysis. It’s doing the thinking, the doing, and the crunching. I’ve seen a real quality step-change using Claude versus what I’ve been getting from other tools.

It’s actually not about providing any one tool — it’s better to provide a lot of different tools, because people go on the journey with them and use them for different things. It’s all about adoption. The best thing you can do — the thing we’re definitely doing — is just putting all the tools out there, providing all the training, making sure leaders are inspiring their teams and leading by example, and really encouraging people to self-learn and self-trial.

Simon Andrew

Off the back of that, I totally agree with what you’re saying. There’s that element of play. People just need to pick it up and try stuff, get it wrong to start with, but just work out what works for you. The more we can give people the chance to play and learn through doing, the better.

Alex Graves

I think that’s a really nice place to draw it to a close, with 30 seconds left. Thank you both, Sarah and Si, for attending today and supporting with the conversation. We’ll be able to share the recording, and you can share it amongst your team as well. Any questions that have just come in that I haven’t been able to address, I’ll make sure we follow up after the session and make them visible so everybody can see. Thank you very much everybody, and have a great sunny afternoon.

Sarah Meurer

Thanks, everyone.

Simon Andrew

Thank you, bye.

Alex Graves

Cheers, guys. Thank you. Thanks, bye.

Simon was an early adopter of AI and has been using it ever since. He’s helped numerous IC teams think about how it can support them in their day-to-day roles, and is a firm believer we should bring the AI conversation out into the open, addressing anxieties, learning together, and taking more ownership as businesses. Recently Simon has been building agents to simulate employee responses.