Agents in 2025: Hype or the Future of Work?

all of the conversations the announcements the media about technology at work right now keeps coming back to this concept here agents agents are going to transform business processes and completely change the way we work but how much of this is hype how much of this is real what do you need to know to get across this what does this mean for your business your career your job right now and what should you do about it I'm here to take you through all of that the best definition I've come across for agents in a workplace cont

text is that they are tailored knowledge workflows we're very much here in the world of process Automation and process transformation but the difference here compared with traditional automation is that we're bringing Ai and knowledge into those workflows so that the agent can actually refer to things and take independent decisions so if you think about a traditional workflow where you might say go out and retrieve a customer order detail or even a traditional chatbot where you can say hey chatbot please find the customer's recent order and it can go out with a connector and bring that

back this next level of Agents means that it can reason it can use its knowledge and the language model to reason over other information sources so at its simplest that example I was just talking about turns into something here where we can say Don't just bring back that customer return but also now check the returns policy and make a decision using that knowledge and reasoning that I've given you to decide whether or not the person can return it this is one simple example but it's a nice one that I think most of us can understand

we've all wanted to return something after all the idea here though is that we can start to rethink what's possible to automate in a business context based on the fact that now we have these AI these large language model capabilities that are able to both understand take in know reason over that knowledge find answers generate answers summarize things back to us and a whole lot more because this world is moving so so quickly that speed is part of why we have such a huge hype cycle because seriously you turn your back on this stuff for

one day sometimes 1 hour and there's something new going on I keep up with this as much as I think it's possible to do while I also have a whole job and a whole life going on and honestly it it's overwhelming to keep up with so here are some key things that are why this is turned into a big hype right now is because the capabilities of what generative AI can do are starting to allow these situations to become real so there are four parts to this the first thing is just that the models are

becoming more capable and more useful all the time we have seen a huge uplift in the last 2 years from where we started to where we are now and there are even other new models coming through now that can just do more they can do more faster the second thing is that with the capability of these models things like this new preview from open AI 01 model and I'm thinking this is brand new but look at the date on this announcement I'm recording this January 2025 so this is old news if you like but this

is a model now that is showing increased ability to do reasoning and calculation type things which we haven't seen before so the capabilities of what generative AI are able to do are essentially just growing and changing all the time the third thing here is what is happening with generative Ai and memory I'm going to let someone much more qualified than me tell you about this I've got a bunch of little Snippets and resources that I'm going to show you here and they're all in the description if you do want to Deep dive further in but

uh let's hear from Mustafa we have prototypes that we've been working on that have near infinite memory and so it just doesn't forget which is truly transformative I mean you talk about inflection points memory is clearly an inflection point because it means that it's worth you investing the time because everything that you say to it you're going to get back in a useful way in the future this is a prototype that he's talking about but is a vision for where this is going and why there's hype the idea that your agent has everything in its

memory makes it far more capable of being able to take on business processes even without these prototypes of infinite memory which is a whole other thing that I'm struggling to get my head around the idea that you can start to build an agent that goes beyond what I said at the start yes let's take this information and do something about it but it actually has a living memory of all of the things that are going on in the organization the same way a person might so we become I think in the future here less dependent

on people who've done the same job for a long time or processes that have to be documented a certain way and then people having to find them and actually feeding all of that organizational knowledge to an AI which is able to retrieve it and do it in a more useful way I think that's where we're going with this the other thing that is really really important to understand and this is the fourth thing here and I've left it till the last because it's the most important is that agents are more than just chatbots so in

my very first example that I started with here and I do have a whole video showing how I built this one out this one is actually what's known as an autonomous agent so we're not just interacting with this by a person coming along and saying Hey I want to return something this agent is actually monitoring and looking for emails that are coming in taking that action understanding what's going on and replying to the customer now of course you can put your human in the loop anywhere you like but even beyond the idea of this sort

of experience we are seeing multimodal capabilities coming through with generative Ai and by multimodal what I mean is that we're no longer just restricted to somebody typing in a text box we are seeing generative AI able to work with vision with audio with speech this is where it starts to get to a Next Level so let's look at three examples of that that are real in the Microsoft platform right now these are all fairly early preview but these are all real demos the first one is this idea of vision and co-pilot vision is one of

the products this is only available to a very very select number of people in sadly I'm not one of those people yet but stay tuned when I am I will show it to you so the idea here is that you are actually able to use co-pilot in your browser and have a conversation with it where it is aware of what's going on on the screen so the examples that they show here and again Link in the description if you want to check this out there are actually sort of things in here where you've got a

shopping site open or a place here where you've got accommodation and you can ask it questions and it's actually genuinely interacting with the content on the screen so it can see your screen there's been quite a lot of things coming through in the Microsoft announcements that show this idea at work as well of hey the co-pilot is actually able to view what's available on the screen and have that conversation so vision is the first one the next one is where we start to get into to the audio and the speech so the big thing here

and this is going to really change what's possible with customer service solutions this is a video here from Microsoft ignite showing an example of an ivr a voice solution so you know when you ring a call center and it goes press one for this press two for that and you have to go through a series of pressing all the numbers this is how this can change so this is again built using co-pilot Studio where you can provide it with knowledge this is core to the concept of what we can do with generative AI at work

and for service provide it with the organizational knowledge and the customer or the human agent can ask questions and get those answers multimodal now means that that can happen in ways other than text so let's take a look at this I will there is a there is a pause in here that I tell you what this this guy who was presenting this at ignite must have just like held his career in his hands for a moment but high drama this one is real let's take a look now I'm going to try to attempt a live

demo by placing a call using my phone and we'll see how the agent responds thank you for calling ignite Financial how may I help you today please choose one of the following options you can press one or say Branch hour press two or say lost card and press three or say apply for a loan what's your best rate available on a mortgage the best mortgage rate available is for the 5/1 adjustable rate mortgage arm at 5.5% APR okay the the drama of the pause there and and also the reality of this technology I did start

this video by saying what's real what's hype this is real but if you were in that real customer service environment right now that delay is too long so like that we're getting there but you need to understand where this is going so that you can start thinking about where this might be taking you so instead of that experience of the agent is constrained to The Choices put in front of it and you have to press the keypad to choose the right one that person just asked a question that wasn't one of those options and got

the answer from the knowledge Source in the same way that you can and we've started to get used to the idea that you can just chat with an agent and it will provide you with an answer now you can voice chat and it will speak the answer back incidentally if you're building this out with co-pilot Studio you could build it once and it will work across those multi modals if you're interested in an actual demo I'm I am going to put something together but let me know in the comments because I need to get this

set up and and and prove it to me and to you and uh and do it for real all right so now we've had a look at vision we've had a look at speech and audio the other thing here is to start to think about agents even in other different ways so this is an example from the Power Platform Community Conference this was in uh September 2024 and this is Ryan Cunningham who is the vice president of intelligent applications talking through the idea of a data entry agent so this is a whole other thing entirely

what we've got here is his traditional Sasa application the business system where he has to do the data entry on the left on the right we have an email from a customer that is unstructured so what we're going to see here is how an agent can reason over interpret understand that unstructured text and turn it into something structured he does get excited about this and I I do love that and luckily with an intelligent app we don't have to I'm going to take the whole text of this email I'm going to let you in on

a little uh fancy programming trick I learned in college it's called copy and paste and check that out we've got a data entry agent working behind the scenes here in power apps boom it's going to take all of that text and fill out every single field in that form for me it's going to reformat the dates it's going to get everything right it's going to pull it out game Cher massive Time Savings all right I people get a bit skeptical when they see this because is a just a contrived demo but again all of these

things are real they're very early but that idea of the generative AI can actually understand what's going on it can interpret that email the content of that email and then do that work for you that is a a real capability so we're going to be seeing more of that I am going to let Ryan wrap up this section for me on multimodal agents with the best quote I've heard on this subject about why agents are more than just chat and why it is important for us to start thinking about this technology in ways other than

chat this is a snippet from an excellent podcast called uh powertime highly recommend that you check out this whole episode and other episodes here but let me just play this snippet of Ryan on the subject of why this is more than just chat there's a lot of stuff that we do today that doesn't need to be fully reactive to a human pressing a button or typing into a box and I think that's the other big fallacy of chat as the the ux modality is you know gosh if we're going to really build agents that are

incredibly capable and can reason over all our information it's like hiring a bunch of new employees and training them on everything your company's ever done and then going and sitting them in the corner and saying only speak when you're spoken to I love this using an agent only for chat is like having employees with all this experience and all the knowledge and saying only speak when spoken to we want these agents to become an active part of the business process so that they are doing things on our behalf the way that you would delegate and

again if I'm going to keep going back to my example here at the start when you are building these agents you're actually providing them with a purpose you're providing them with that here is your job description here is what you're here to do please go ahead and do it and let me know if you need anything so that's much more the idea of agents and the idea of being able to use them in these independent and autonomous ways so essentially where we're at with all of this is that it opens up new ways of thinking

about Automation and new ways of thinking about business processes so so many organizations have got tedious laborious business processes that they haven't been able to do anything about this is where we get the chance to actually start to come back and really properly look end to end with these new capabilities coming and these new ways of thinking what might be possible and not all of that might be possible to today although maybe if you're watching this video in the future it is I'm sure it is so the idea is that we start to rethink we

start to think in different ways about how can we bring intelligence and knowledge into the workflows and therefore what does that mean for them now I did say at the start there was a recent podcast with Sachin Adella CEO of the Hall of Microsoft This Guy's in charge of the whole thing which is really really insightful about some of these things and also stirred up a bit of more hype and more controversy and I'm going to play you that and we'll talk about it in a moment but on this idea of rethinking business processes he

does also talk about how they're using co-pilot and agents in the office of the CEO at Microsoft of course they are using their own technology don't listen to the competitors who say they AR the idea here is that the reinvention of the process as he describes it is instead of saying he has someone in his office who he's going to see a customer someone prepares a brief and then they have to gather all the information and he reads it that that gets flipped around where he can actually go in and say to the co-pilot hey

can you please prepare me a brief with everything that we know about this customer and then there are agents at work there that can retrieve information from different systems bring it all together and that gets dropped in real time into a collaborative tool and everyone adds and works with that so the chance to rethink and to reinvent and to not continue to just do the same things we're doing we we miss the great opportunity of all of this stuff if all we do is the same things a little bit more efficiently or a little bit

faster what we're really looking for is this Innovation how do we rethink things how can we challenge the way that things have done and the AI tools that we're getting here provide us with a huge opportunity to rethink now strap yourselves in this is this is where it gets this is where it gets big in terms of rethinking this is a very very big rethink here he's talking about the idea of agents and what happens to SAS applications business applications like crms and all of those databases that we use at work every day as well

as things like spreadsheets and what happens here I'm going to play this for you and then we're going to have a chat about the SAS applications or Biz apps so let me just speak of our own Dynamics thing the Approach at least we're taking is I think the notion that business applications exist that's probably where they'll all collapse right in the agent era because if you think about it right they are essentially Crow data basis with a bunch of business Logic the business logic is all going to these agents and these agents are going to

be multi- repo crud right so they're not going to discriminate between uh what the back end is they're going to update multiple databases and all the logic will be in uh the AI tier so to speak um and once the AI tier becomes the place where all the logic is then people start replacing the back ends right we people you know that's what you in fact it's interesting as we speak I think we are seeing pretty high rates of wins on Dynamics backends um and the agent use and we are going to go pretty aggressively

and try and collapse it all right whether it's in customer service so there you go they're going to go very aggressively collapsing at all so what he's talking about here is that agents essentially become another this is my interpretation of this agents become another layer in between instead of having individual business applications where you go in and you log into your CRM and you do that data entry you've actually got an agent experience or co-pilot experience that's got multiple agents going out and finding information across multiple systems and working across those systems for you this

is hard to get your head around this is one example the first example I think that illustrates this the best this is an example of a lead qualification agent so this is just one business process but the way that this business process would have run in the past is that the salesperson logs into LinkedIn logs into their CRM perhaps has to look at some kind of ordering system Gathering information or perhaps even administrator is doing that and surfacing up and saying hey here salesperson here are the leads that you need to work on and then

they need to do that qualification and research and then reach out to the person so this is the vision for what I believe he means when he talks about collapsing these applications it doesn't mean that CRM goes away or that any of those applications go away but the user interface layer of them changes so they become these underlying very very important in fact I would say more important databases because we can start to get more use out of them a lot of these systems just don't get used cuz no one wants to do the data

entry and no one wants to go in and look up the things but if they come become these valuable sources of knowledge with a different user experience over the top that starts to use AI in a very Dynamic way this this is what I think it is so let me let me just play this one for you let's maximize this one out and take a quick look here the agent autonomously research SE es leads in your pipeline 24/7 it analyzes multiple data sources and flags the hottest prospects for you with just one click you can

learn why a particular lead is prioritized based on research data synthesized from CRM records thirdparty data and public web sources when you're ready to reach out the agent has already drafted a personalized email a quick review a couple edits and you are ready to hit send and just like that you're ready to move on to the next prioritized lead again I will bring you a real life demo of this one when I get my hands on it this one is coming as almost everything here I've shown you is in early preview but do you see

what I mean about this idea of the interface changes it's not just forms over data anymore you are actually still using that system but the agent is sort of cating and summarizing and suggesting and writing for you that one is sitting inside that application but that same concept except imagine just inside your main co-pilot or inside teams being able to action those kinds of conversations and you've got an agent checking LinkedIn and an agent checking your CRM and doing all of those things so in conclusion is it hype is it real yes both of those

things there is a lot of hype and and I think that puts people off somewhat but I've shown you a bunch of things here that are very real we need to start thinking about things that are not just chat things that use Vision that use voice things where the agent becomes another layer in the experience that the user uses the idea of the real opportunity to transform the business process and think about it in a different way to think about Automation in a different way and I would encourage you to start with that kind of

process of what is something you're doing that is taking up time that you hate spending your time or your money on what are those manual processes it's not dissimilar to how we started with conversations around pal platform to be honest and it might be that you find a use case that's perfect for an agent or maybe you find a use case as just something you could automate without AI I'm actually finding in conversations that I'm having sometimes the ask is actually something that you can achieve right now just with basic automation there's still a lot

of untapped automation out there you do need to start somewhere with this please please don't kind of take the hype and say oh it's all hype it's not real real there are some things here that are real but start with those defined use cases start with a nice small use case something that can help you learn understand and the most important thing if I'm going to leave you with anything is the ability to start to think differently because all of this stuff all of this generative AI at work if nothing else is going to challenge

us to think differently about how we do our jobs and about what's possible and just just start somewhere with that let me know if you've got questions what else you'd like to see I've got my video here if you'd like to learn how to build autonomous agents and I'm going to be spending a lot more time on agents and co-pilot on the channel this year so stay subscribed if you're into that with me as always thanks for watching