[Music] thank you call and thanks for having me again Actually um this is my third time presenting for you in a row So if we keep this up I'll be visiting the Marion more often than Bruce Springsteen um I think So um so thanks uh thanks for the opportunity again Um and Jer I really enjoyed your presentation by the way and it'll touch on or it did touch on some of the the developments and the areas that we we're seeing with this rise of agentic AI Um and just to understand how your business is evolving to to take that on board was really fascinating So hopefully what I present here will will complement that a little bit And I I had this ever so slightly cheeky um title of why four o'clock is the new five o'clock because and again I think this echoes some of the points that that you made Jared This is not about AI replacing people um in their jobs It's about using AI to make people's jobs um easier um and to make them more productive and to make them more um empowered And hopefully what we'll be able to do is make people's work a little bit more enjoyable in that sense as well So you know over the last um I guess well since November 2022 when chat GPT hit the streets we've charted this remarkable rise of of AI and you know at first it was a mechanism whereby you could ask a model some questions and you could get some answers Um and I think when I presented last year I talked very much about how you could use AI as a companion to um make some innovation type decisions in your business a little bit easier How you could use the multimodal aspects how you could show it images how you could show it video how it could listen to audio and how it could make all sorts of inferences from that Um and the new phrase dour is agents or agentic AI Um and I'd like to try and demystify that a little bit basically to explain what we mean by agentic AI And I'd also like to turn the temperature down a little bit on this whole concept of agentic AI because um if you listen to many companies Microsoft included um all you hear in in the industry is agentic AI is going to change everything and and it will and I'm utterly convinced of that Um but I think it's very much in line with the time frame that Jared mentioned What we have today in terms of Aentic AI is almost um prehistoric in terms of what we expect to see from Aentic AI within about 12 months So that's the area that that I want to focus on Um and I I will talk about this from a Microsoft perspective simply because not as a as a cheap product pitch simply because it's easy for me to use what we have in capabilities today through something like Microsoft copilot which for all intents and purposes uses the chat GPT models behind the scenes um and to compare that with what we're seeing in terms of agents So when we talk about co-pilot as a productivity assistant today it's about um augmenting humans being your assistant to make you doing your job easier And we have this vision that every employee will have a co-pilot But when we talk about agents it's slightly different Um we're not so much interested on this notion of every individual having an agent but it's a little bit of a broader scenario whereby we envisionage a future where every business will have an agent that will work alongside humans and will augment the work that they're doing So it's um it's very much about on on the left hand side co-pilot where you might ask a question get some information back and you as an individual take action on that information in the agentic world it's about sending an instruction to an agent to say I need this done please go and do it for me and you have this degree of autonomousness that you associate with the agent and you might obviously have some control over it but you give an instruction and the agent agent goes about taking the process and automating it and um and doing a lot of that work for you So um this is what we mean when we talk about agents They're they're autonomous systems that plan reason and act to complete tasks So it's basically like going to a colleague or um a subordinate if you want to look at it that way or a junior employee and saying please do this work for me and come back to me when you've got the results Um and we expose agents at least in the Microsoft ecosystem through C-pilot and I'll show you that in a second We'll expect to see agents built into applications that companies um are developing Um and we will expect to see u agents built into to various websites Um in fact um Gartner produced some research a while ago I think it was a couple of weeks ago and they indicated that you know within the next 12 months they expect to see 30% of all software applications have agentic capabilities built into them So this is this middle box at the moment it's less than 1% So to get back to the theme here that the trajectory of this thing is very very steep indeed And this is what the agents are going to do they will act on your behalf reasonably autonomously Um they will they will reason So they won't just get a question get a piece of data and feed it back to you They'll look at a plurality of different data points and they'll I hesitate to use the word think but they'll they'll reason through that data and they'll come up with some very very significant insights Um and we see that with the models that are taking sort of center stage in the the AI world today We've moved away from the traditional LLMs that OpenAI have like um GPT3 and GPT4 into their newer models which are the reasoning models like 01 and 03 And this is the new AI engine that is parring um these kinds of of models And we have this spectrum of agents as well On the left hand side where we've got simple agents that you can create yourself in Copilot for example with just a few mouse clicks and they're relatively straightforward These are agents where you can you know create an agent that might act as a as a chatbot or a field service assistant and you direct it to some data sources and is there to answer questions on your behalf Right um across to the right hand side where we have much more autonomous um system that will you'll give them an instruction and they'll go off they'll do a lot of work and they'll come back with a very reasoned output Now of the four boxes here in terms of this spectrum we're very much on the left hand side of this spectrum at the moment The agents and I'm going to show you some today that are available um uh within Copilot They're the more simplistic end of the agents but you don't have to be a rocket scientist to work out that you know if you look at the the trajectory of what has happened since November um 2022 the pace with which this is improving is is just huge Um and I'm very confident that in 12 months time we'll see very very sophisticated uh agents So these are the types of things I mean again this this reflects the spectrum relatively straightforward agents that act as chat bots that you as a user can create and share with with other um employees in your organization or with customers over to more sophisticated agents where you can say um I need you to go and test this whole model I'm I'm intrigued and Jared Jared I'd love to work a little bit more with you to understand where you're going to go in terms of um using an agent for example or instructing an agent to say here's a problem that I want to research design a questionnaire for me identify the right demographics to go target um get get the uh get the data from that and then make sense of it I presume that's the direction you're heading in and I'm going to be fascinated to see how you tie that in with the the human side of it um as well So um oh just before I go on to this one other thing I would say and maybe I had it briefly on a on a previous slide um and Connor talked about hybrid work um we see a new definition of hybrid work over the next 12 to 18 months and it's not the the distinction between people working at home and people working in the office It's the distinction between people doing jobs and tasks and agents doing jobs and tasks And it's going to be fascinating to explore the relationships um between individuals as they manage a team of agents versus managing a team of people Um and what we're really focused on is the extent to which these agents can um improve output um while reducing burnout Right And I think that's what we're really interested in that you know how can we get agents to do the work the the kind of heavyduty drudgery type work that that humans don't particularly want to do and indeed can't do it very well How can we offload that to uh to agents So if you use Microsoft Copilot this might be familiar to you It's the kind of standard co-pilot screen Um and I'm sure you won't see it at the back of the room but over here on the right hand side there's a little drop-own option that says create a co-pilot agent And literally creating one of those basic left-hand side agents today is as simple as saying create me an agent Give it a name Um tell it where it can get some information Type in you can go to this SharePoint site for example Um off the agent um goes and you'll see when you go to configure it then it'll show you the information that's available to it So here's some some channels on that SharePoint site and you can um you can enable them and suddenly you've you've actually created an agent and this is an agent that you can direct you get a URL for it you can direct anyone in your organization or outside your organization to this um to this URL and it's effectively an agent with which anyone can communicate and the agent is I suppose a bit of a chatbot but it's an LLM AI powered chatbot that is answering questions based on specific data within your organization So you can create very basic and simple agents very very quickly Um and what we expect to see is a huge increase in a marketplace for agents So today they're already agents thirdparty agents from Adobe I'll just pick a few SAP workday So you can just imagine the way that these agents will um interact with individuals So you you'll go to the agent and you'll say please give me information pertaining to this and the agent will go off and it will get everything and it will make sense of it and it will come back and present that that information to you Um so really when we think about agents you're thinking about u what I like to describe here is intelligence on top right so you know gone almost are the days whereby you have to go to expensive consultants or third party entities to say I need some research done sorry Jared I need some research done on a particular topic can you go off and give me everything I need to know about that now Jared is a was a brilliant example of why AI will never replace the authenticity of of humanity as he gave his presentation And I think humans will always be able to identify insights or or present information in a very nuanced very human consumable way But a lot of the donkey work that is time consuming and expensive in terms of cost and monetary value we expect to see agents doing that in the future Um and like I said earlier we expect to see these human agent teams changing the whole nature of the organization So this is what we mean by the new hybrid It's humans and agents working together So I'd like to pick up on um on just a few agents I'll go through them pretty quickly Um these are all agents that are available today in Microsoft Copilot Um I'm not going to go through every single agent but um I am going to pick out a few that I think are are quite interesting and fascinating So the first one um you might be familiar with if you use Teams and Copilot in Teams is the facilitator agent Um when you turn on facilitator within Teams it acts as a chair for your meeting Even if you don't have a human um performing that function of a chair in the meeting it will look at the emails that were associated with the meeting invite um a draft agenda It will track what's being discussed in the meeting It will produce real time notes um in the facilitator panel as the meeting is proceeding It'll track against what was in the agenda It'll indicate we're running out of time Um you had five agenda topics you've only focused on three Here are some open questions that are in place and it will allow the individuals in the meeting in real time to edit those meeting notes as the uh as as the meeting proceeds and it'll summarize it So you can do some of this a little bit more manually today at the end of a meeting with co-pilot you can say summarize what was said and produce the actions but this is an agent acting on behalf of you and your team in real time um with a meeting and it'll integrate within meeting rooms as well Um it's worth pointing out the interpreter agent Um I think I think this one is just fascinating and there's a bit of audio on this So this is where um in teams now you have an agent that's doing effectively real time speech translation for participants in a meeting Um it is a bit Americated during a team's meeting at any time Once you've turned on interpreter you can choose from nine different languages for real-time interpretation [Music] You have the option of simulating your voice or you can choose a preset voice option Hi thank you everyone for joining Today's session is a quick today I'll be covering a few key updates that impact how we collaborate and where to find I'll just leave it at that I think his eyebrows are great for speaking Korean in that Um actually it always strikes me Um you can expect to see this become better and better over the next number of months Um you you you saw you can have a simulated voice or you can have it in your own accent So that's going to improve significantly over time And I expect you'll probably even see the video stream manipulated so you get better lip syncing and stuff like that So it's just incredible um how this is going to uh how this is going to develop Um Jared this might be of interest to you This is the researcher agent Um I've used this uh quite a lot uh recently but basically you can go to the researcher agent again it's available in copilot today and you can say I need you to research this topic for me Um and you don't even have to direct it to sources right You just say this is the challenge I have These are the insights I'm looking for researcher will go off and this is a little bit sped up but it is an actual screen grab but in the space of about probably three or four minutes it will have assessed hundreds if not thousands of um of resources both within your organization and outside your organization to come up with very reasoned outputs Um I actually used this with my daughter for her end of first year law school exams just a couple of weeks ago Right Um so we didn't cheat just to be clear on that right But what she did was um and the way her exams work she was given a problem and then you'd got like six hours to write an essay and submit it So she would take the problem she would write the essay And what I did at the end of this was I took the problem and I took her essay and I put it into researcher and I said "Could you look at all the areas in this essay where we could move this from a 2-1 to a first?" and it came back with like four or five areas to say this is a good point but have you considered this have you considered that Um so again it's not doing the work for someone It's just helping you do your work a little bit better Um I won't tell you what university she's at just in case uh in case you you say anything about that Um that's pretty much all I wanted to talk about here right Um just you know there is another agent that is a is a partner to the researcher agent called the analyst agent I I do have a video but I decided not to show it but it can basically take all sorts of tabulated data um information that you'd see in an Excel spreadsheet and you go off and you say you know look at all of this information and tell me about um customer trends or what I might expect to see if if this were to change in the future and it'll go off and it'll create visualizations and you can interrogate it To me at least it is what the original Excel co-pilot should have been a year and a half ago Um but we're starting to see it now So you've got researcher agents that's focused on textual information and the analyst agent that's focused on um um more numeric information So um all of these agents are available today in the co-pilot store the the the internal ones And you also start to see some of those third pilot um agents integrated in there as well And with that I think I'm pretty much on time Okay Thank you [Music]