51: Using AI for Boosting Workplace Productivity with Jimmy Vaughan

This podcast episode is sponsored by chiefaiofficer.com offering training and certification through the International association of Chief AI Officers. Interested in a new career or leveling up your value in the marketplace? Chiefaiofficer.com can help. The following is an interview with Jimmy Vaughn. Jimmy is a specialist at Microsoft with a wealth of knowledge in generative AI, workplace technology and bridging the gap between cutting edge innovation and everyday business needs. His unique perspective combines technical expertise with practical approaches to help businesses harness the power of AI effectively. In this episode, we discuss how Microsoft's copilot transforms productivity by addressing common workflow challenges, the emerging role of AI agents in delivering focused solutions, and the critical need for data governance in the AI era. Additionally, Jimmy shares insights into fostering intentional AI adoption within organizations and preparing for the future of work with AI driven strategies. And let's begin. Welcome to Using AI at Work. I'm your host, Chris Daigle. Each week we'll be learning how today's business owners, entrepreneurs and ambitious professionals are getting more done with smart use of tomorrow's tech. Let's get started. All right, ladies and gentlemen, today I'm pleased to have Jimmy Vaughn dialing in from New Jersey, a specialist at Microsoft. And we're going to be talking about all kinds of fun things related to generative AI in your role, your business, your team. So, Jimmy, before we get cranking, why don't you let everybody kind of know a little bit about what you're doing in the AI space right now at Microsoft?

Well, obviously with Microsoft, you can't have a session where you don't mention copilot. So we'll just start with that sort of in the AI side, I'm looking at a lot of the workplace things that happen and a lot of different ways that people can, number one, use AI, but also understand where it can actually be a little bit of help for you. Satya Nadella actually spoke a couple weeks ago at Microsoft Ignite and he said we're in middle innings of AI and I, I actually disagree with him. I think we're still very, very early on and I think there's a lot of adoption and a lot of things that, that still have to work themselves out. So if, if I, if I put it in a perspective of a history side, think of the Internet in the early 90s, you know, and think, think of where we were. I mean, Windows 95 wasn't even out yet. You know, Mosaic browser was, was the head of things and, and there were a lot of different things inside of the Internet that people were sort of questioning, hey, where will this end up? How will we actually use this? What will be there? And I, I think that's where we are with AI right now. Now we are in the, you know, 2024, you know, approaching 2025 and the speed of things I think will rapidly, rapidly increase. So, so while he thinks that we're in middle innings, I still say, I think we're in the early part of this whole solution.

What were some of the takeaways? Because we had a lot of our members of the chief AI officer community at least watching the Ignite sessions. Was there anything that stood out to you or that you think people should know about that was revealed and just explain what Ignite is for those who aren't Microsoft.

Ignite is Microsoft Super Bowl. It's where we come together. We say, hey, here's what we're doing and we make all these announcements and we go out and we make sure that everybody knows where we're heading for the next year. So we put, we put partners, we put customers, we put everybody into a notification style broadcast where we say, this is what we're doing and we hope that you're along for the ride for all of us. And the biggest thing for me was the agents being able to, number one, build your own agents and number two, just have all of these different places. Before Copilot was the word and it was copilot, copilot, Copilot. And it was this, this grand umbrella of everything that exists. You have copilot and PowerPoint, you copilot and Word you have, and you still have all of those, but those are now sort of individual agents that are working together and can be under that Copilot branding. So the, the agent side is really what's, what's exciting for me. It becomes less of a large, large language model and becomes more of a focused. And I want to, I don't want to say a small language model, but a focused model approach to, to how AI is actually rolled out with, with Copilot.

This is crazy. The, the last episode that I recorded agents were the topic 2025 is going to be where a year where at all strata of business they start hearing this AI agent, AI agent. So our, our chief AI officer community in particular is because we've, we've had some dialogue with Microsoft. We're excited about opportunities to support their users when it comes to upskilling and that sort of thing. So for at least the shout out to the chief AI officer community as well, as the rest of the listeners, kind of help me understand the Microsoft's intention when it comes to generative AI in that copilot agentic day to day ecosystem.

Yeah, it's funny because, you know, I've been using Copilot, what was Copilot? Even early on we didn't have the name for it and things like that. Very, very early and getting some of that experience, you know, when ChatGPT came out, that was the most, one of the most exciting times that we had and showing what it could actually do and then bringing it to a business perspective and saying, hey, with the data that you use on a day to day basis, let's see if we can work some of these models in and make sure that we're getting the results that we get and being able to show business leaders, hey, here's where it is. And Chris, right before this, I think you and I were even talking and I was, I was, I was sort of a little bit critical about some of the ways that things are being rolled out. I see people demo these moonshot ideas and these huge, let's go look at this and let's go do this. And I think what, what, you know, we as a community really have to do is sort of take a step back and say, number one, do you need that moonshot? And if the answer is no, then okay, it's great, but why, why do I need it? But let's really focus on where, where things like this can help. And one of the examples that I give to so many people is, have you ever thought about it? Hey, Chris just sent me a PowerPoint. It was maybe a couple of weeks ago and everything else, and you comb through your email and you comb through your team's messages and you combed and he went through all these different places and then you're 20 minutes in and you're like, wait, what was I working?

I got a meeting now?

Yeah, because. Yeah, because you just lost that focus.

Yeah.

What if I had something that was, that was sort of a natural language where I could say to it, can you tell me the last three PowerPoint items through, through email, through teams, through all of this where there might have been a PowerPoint from Chris and bam, I've got the link, I got the document, I got everything. And now I'm still focused on what I was doing and I think that's where we have to look and we have to say this is the idea that we can, we can have now because it's not reliant on me. To do everything and all these little things, I mean even go back to, to, to four or five years ago. Think about when you used to put in the email name and you had to remember the exact address because there was no auto fill. There was no. All of these little things. I mean, this is, this is again the idea of AI working its place in. And, and those moonshots are great. They're, they're, they're spectacular. They're great for a demo. They're unbelievable. And, and, and I love them. But at the same time I think we got to dial back and say, okay, I, I know where I can use this. I know, I know where I'm seeing.

Value at this, this is interesting. So, so what I'm seeing here. Cause, cause we kind of talked about this before we started recording. There's cracks and those cracks are distraction, cognitive load, you know, wasted time searching for stuff. And AI, generative AI is kind of coming in and it's filling in those cracks. Like I didn't even think about the auto. Autofill feature on the email. Like otherwise was, it was another thing that I had to hunt for, that I had to go and look up interesting perspective and that and then, and look for those listeners who are, well, I thought it was going to be, you know, the moonshot listen in the meantime. Except that it's filling in those cracks of productivity.

So yeah, 100%. And I even take it a step further with some of the, some of the partners and customers that I speak to and I say, all right, number one, you know, everybody's going to talk about how you increase productivity. Well, I find it very difficult to actually measure productivity. I think a lot of other people do. And, and Microsoft is creating dashboards to show you different things that you can do and everything else. But let's say you're not buying into this. And okay, you, you can't actually measure productivity. But what I can do is I can show you those cracks or those places that productivity is lost. And if we, we start to look and we start to fill in those things. A colleague of mine, Mark Hodge, is a global black belt here at, here at Microsoft. He coined the term blind spot for AI. And this is, this is something that, that, like it was, we, you know, we, you know, we had a late meeting that day and he mentioned blind spot. And then we went on a 45 minute conversation.

Yeah, yeah.

About blind spots. So Chris, think about some of these blind spots that might exist in a workplace. Right. We know that we don't work the Same way that we did six years ago when we all went remote. But if we're going back to the offices of six years ago, what are we missing now in our work life? What are we actually not taking advantage of? And if we even think about Microsoft conference rooms, Microsoft Teams rooms are specifically built and bringing AI into that workspace as our customers are actually investing in AI and copilot, they might not think that this is tied to the conference room, but it really is. Because when I get to a shared space, what happens if everybody in that shared space is identified as Speaker 1, Conference Room 1234.

Right.

Think about how AI generalizes and goes through that transcript. Well, we're. If we're talking about budget, right. What, what somebody might get in a summary from an email from a, from an AI agent is Conference Room 1234 approved the budget and Conference Room 1234 disapproved of the budget. This is now a blind spot that you weren't thinking about in your AI deployment. Now when we do this, what happens? I lost productivity with my users who didn't attend that meeting because they not they don't, they can't trust this summary. So what do they have to do? Well, they have to go back and watch this hour long meeting to really absorb what was being talked about. Everything else, I lost 15 minutes of their time because that AI summary didn't work.

50 minutes of many people's time.

Correct. So again, don't think of it as oh, I have to measure increased productivity, find out where you're losing productivity and let's see where we can help that. And I think that's really, that's where a lot of even Microsoft sellers just don't hit the nail on the head for, for how we're doing things. Again, that moonshot is great but the little things we really have to start to look at and we can, we can find solutions for that today.

This is great man. Because we routinely interact from, from our company. The company behind this podcast is chief aiofficer.com right. We train and certify individuals to be those AI expert non technical experts in their business. And I always want to show the magic. We call it the candy. Right. When you, you used AI like oh my gosh, is that really you riding an elephant? You know, whatever.

Right?

Yeah, I mean we use it in business context but this is an interesting way of looking at things to where one of the things you mentioned, I don't know how to measure productivity, but I know how to measure loss or lack of productivity. That's a fantastic way to look at it. When you think about that as a user who may be like, well, I thought AI was going to do all this stuff. For me, it starts with no longer losing that productivity, at least. And it's 30 seconds here, it's five minutes there. That stacks into five hours, a week or more, I think, as Boston Consulting Group recently released. Like, it's not the event.

Correct.

It's the process that's happening and kind of making the boat go faster. You know, like, I love it.

And I. And I think that, yeah, I think that really goes back to even the thought that I. That I shared with you of, do we work the same way that we did six years ago? You know, I'm an engineer. I go into a conference room, and I'm the first one that picks up the whiteboard pen and starts drafting on the whiteboard. Right. So, Chris, let's take a perfect behavioral example of what, what happens inside of those meetings. You and I go into this conference room and we start writing on the whiteboard, Right. And we've got this whole grand idea for a new AI scheme that we're doing. Maybe we can train this model. We can do this. Right after the meeting, you and I both leave. And what do we have as the collateral there? Well, we probably took a picture of the whiteboard.

Yeah, yeah.

And we probably left there with action items of what we were going to do and everything else. Right. But when I got, you know, two weeks later, when I look at that picture, I forget now what we were talking about, what that thing was. Right. So what if I put something like a Microsoft Teams room inside of that room and now I hit a button and now it transcribes and it knows that it's Chris and it knows that it's Jimmy automatically from the, from the sound of our voice because we registered our voice in. Now that video conferencing solution now becomes multipurposed inside of the room itself. And now we not only have that digital photo of the. Of the whiteboard, we also have what we talked about. So I can go back and I can query what were we talking about when we were doing this, or I can even go back and look at the transcript. And now my mind goes right back to where it was versus what was it six years ago? Well, six years ago, I call you Kristen. I go, I just don't remember what we talked about. I lost 20 minutes of my time because we had to reinvent that whole thing again.

Yep. Or may have both of us Go, yeah, I don't have it. I don't know. Right. And that whole. It can flow. Yeah, exactly. This is a great way of looking at things, Jimmy. So this, this experience that you're talking about now for, for us in our lexicon, we call that thinking in AI, when somebody kind of, they go from using it, like the Google search right into, oh, before I hit that key, can AI help me do this? Right. Like, they start to make that transition. And that's what you're talking about here. And it's what, what, what are maybe some tips? Because that is the difference between, hey, my boss got me co pilot. Yeah, I know how to push a button or two, but I'm not, I'm not. I haven't created the behavior of deferring to the Gen AI tool.

I think that's even in what Microsoft is seeing here itself. You know, I, I do, you know, you and I were talking. I do a lot of prompting and I do a lot of things, and I think what, what, what you have is this idea that, hey, I have to talk to this entity as it is, and I have to get the results that I want out of it. Yeah, well, even, even you and I having conversations, that's sometimes difficult to say. Yeah, what do I want the outcome to be? You know, Google gives me the outcome if I just say, you know, what is this? And everything else. But I don't always get the same thing. So what you see in Microsoft is a lot of us who have been writing prompts. I sort of know how to, how to lean in and how to, how to run that, that, that agent down and say, all right, this is where I want to get to. And it's probably a process, but what we're seeing is we need to actually give people the prompt itself. Give people, this is something that's effective and go through it because they, you know, and I think you're even seeing this in a lot of demos that you see. What you're starting to see even in demos is people are, or sort of doing those moonshots with videos instead of actually entering the prompt themself, because when you demo it on day one, you may get different results than day two.

Yeah.

You know, and it's an interesting thing to do an AI demo, because if you don't have that video. You know, I think back to a couple of the. Jarrett Spittaro was actually doing a demo very early on. We had just released Copilot. And he goes in and he says, he says, write me an. Write me A Python script about the U.S. census Bureau or something like that and everything else. And Copilot comes back and goes, well, I don't know how to do PI. I don't know how to do Python. And he goes, wait a minute, you do? Because when he did it the first time, it actually knew how to write Python. But. And he sort of. He sort of made a joke about it, that this is really like managing an employee.

Yeah.

An employee is going to come back and say, no, no, I can't do it. But when you press it, you may actually get those results. So what you see in a lot of the demos that people are doing to show these. These really creative ideas and what we can do and where things are going in the future, you see them using videos now so that they know the outcomes and they don't have to figure out, all right, well, what if something happens?

That makes a lot of sense. Yeah. Because I've been. We do live training all the time, and you're like, wait a minute, it's supposed to do this. It's not happening.

And I love. I love the. Wait a minute. Yeah, because all of us have been there in live demos, and when they don't, you know, when they don't go the way that you want, you're. You're left there struggling, and you're left there going through things. I. I had a very early. It was. It was before Copilot was released. I actually got an early, early access to it. And it was funny because for the first time in a very long time, I had taken two weeks off and shut off my PC, shut off my phone, shut off everything. So that, all right, I'm going away for two weeks. I'm just gone. When I got back, I had Copilot, and I haven't been able to reproduce this scenario legitimately. It's. It's been there, but. But it really worked for me. And this is another thing where we can sort of help understand the productivity of it. I went through and I said to Copilot inside of Outlook, I said, hey, can you do me a favor? Look at all the unread emails. And I did it through a series of prompts. Look at all the unread emails. Anything with an unsubscribe put into this folder, this unsubscribe folder. Why did I do that? Well, with an unsubscribe, it's probably a distribution list. It's probably not of importance. It's probably not there. So when I came back after two weeks, I had 2,000 emails, let's say. Right. That lowered it down to probably a thousand. Right. So now that's half of my workload because I don't have to really go through those now. Do, do I have to look at them? I probably do eventually, but I'll catch up in time.

Yeah.

Then I said to Copilot, I said, can you pull out any emails where I mentioned? And that now comes down to seven, ten emails, right?

Yeah.

And then I said, can you summarize these 7 or 10 emails? So where it had been a 2000 email return to work and we all, we all go through this, the anxiety of returning to work and knowing that you have all this, it was no longer a two to three day journey for me. It was probably 10 minutes and I was, I considered myself caught up because I assumed that everybody got my out of office. Those that did mention me probably needed an answer. Like Copilot, it summarized and everything else, but it summarized everything beautifully, showed me the last email that I had to do. So I haven't had, I haven't been able to recreate that 100% yet. Every time, every time I go in and I say, you know, the unsubscribe it, it's sort of, it gives me a little bit of challenges, but it worked for me the first time. And that's where I would think everything needs to be at this point.

It's moving there for sure. Right. Even if it's not 100% consistent. And that's something that we have. People, maybe they've got a tool stacked. They might be using ChatGPT for something, they might be using this tool for that, they might be using Superhuman for their email or whatever. But what's missing is that ability to be able to intelligent like, like even Superhuman. I use it, it saves me time. It's. It screens some stuff and it has an AI feature, but I can't engage as robustly with my email or give it instructions the way that you just described. So that's the difference between something that's built into your office suite as compared to, oh, here's a tool, here's a tool, here's a tool. Doesn't mean you're not getting value, but that integration like that, that's a real value point for me.

And I think it's important to know you just mentioned, I mean, you know, people will go to different solutions. And that's one of the things that I think brings value into Copilot is this idea of bring your own AI to work is a Challenge for. For all of your people, all of your IT decision makers that are listening, that's a challenge. Because when you have users that are taking things outside and going to search and they will do this.

Yep.

That's where it. It sort of becomes, all right, how are we managing this? What are we doing? Are we giving users the tools that they need? Because if. If they don't get it there, I'm going to go searching for something that I do. I mean, I have. I have an account of ChatGPT. I use ChatGPT.

Yeah.

I use all these things. Do I. Do I do it for Microsoft confidential information? Absolutely not. Because I know that I have it inside of. Inside of the Microsoft tool, and I can be confident that that information is secure inside of there. But am I doing it on a personal side? I'm mean, I used Chat GPT the other night. I said. I said, here's. Here's all the things that I have in my cupboard. What should I make?

Yeah, yeah, yeah.

And, you know, those sort of things, you know, and you might not think about that as you're going through and going, all right, I got chicken, I got scallops, I have scallions, I got onions and everything else, but I can actually go to that and I can say, hey, make a. Make me, you know, find me a.

Recipe that works with this.

And you just go out so. So little things. And. And Chris, maybe that's some way that we can get a little bit of adoption. Just when you have a. I can't think about this. Yeah, go ask. Just go ask and see. See what you get. You might not like. And that's been one of the challenges that I think a lot of people do is they're promised this moonshot and then when their expectation is not met of that moonshot. Yeah, they go away. But I guarantee you, if you just even look inside of your cupboard and say, here's all the ingredients that I have. Can you get me a recipe? It will find something.

Tell it your keto, tell it your. Whatever. Diet.

Exactly. Yeah, exactly. I, you know, I actually use it. The other day I said. I said, can you make me a workout routine? I have these. This gym equipment and I want to. My. Yeah. My focus is to lose weight.

Yeah. Yep. Said, put down the beer, Jimmy. That's what I told you.

Exactly Right. Stop drinking so much.

Yeah. Um, so you're making a very compelling. Prior to this conversation, I saw copilot as an ant. Microsoft's competitor, or answer to that. That category. But now I'M seeing there's a much more compelling narrative here, which is the integration of not only the integration, but also certainly the data privacy. I mean, I'm sure OpenAI would say, oh, and they even have the toggle switch and stuff. But you still have to wonder. But if you. I've already got my stuff in the Microsoft environment. Like, they've already got it anyway, so I should feel comfortable using. I don't have to go outside. It's native.

That's part of, that's part of it. You know, when, when, when we were, when they announced Copilot and when we, when Microsoft initially said, hey, Copilot is there, every single customer says, I want copilot. So, sure, I would stand up inside of a room and I would, I would say, all right, raise your hand if you think you're ready for copilot. And everybody's hand would go, right, I'm ready.

Yeah.

And I said, not one of you are ready for copilot. And you know why? Their data governance was not there. Nobody thought about it. Nobody did anything. I'm a tech guy, right? So I may get a SharePoint site and I may lop off the end of it and see what I can find.

Right.

And everything else. But you can do things inside of, inside of AI. There's no. I hope it doesn't find this. It's going to find it if it has access to that data. So people can, can go and query and say, who are the top 10 people, top 10 salaries at the company right now. Right. And if somebody has that in an Excel spreadsheet somewhere and it's. And I've got access to it, it finds it. So, so there, there is this data governance that, that has to be done. And Microsoft right now is in this Security first initiative. And the reason why we're in this, they call it sfi. Everything is an acronym in Microsoft. But yeah, Security First Initiative. The reason why security is so important is because as AI starts to reach out and starts to learn things, it's going to be able to return information. So one of the initiatives that we have at Microsoft right now is when we start a new document, when we start a Word, when we start PowerPoint, when we start anything like that, the confident, the, the label at the top is automatically set to confidential. You have to actually go and undefine that as confidential if you want to share it with somebody outside of the company. And the reason why we're doing that again is in the age of AI, where AI is going to reason over tons and tons of data. We want to make sure that the data that we have is actually classified and qualified as exactly what it is. And who has access to that is of the utmost importance. Now, so as, as we were going through for the last couple years, we were finding. Yeah, a lot of people never even classified their documents. Never thought of this as a thing because, oh, well, I have, I have these ous, and you know, I, I separate everything based on the OUS. Well, what if you've got a SharePoint site that actually crosses OUS and, and doesn't do this and, and what if you've got somebody who's, who's saving locally on their machine? Sure. An Excel spreadsheet. And all of a sudden, hey, that's, that's now here. And it comes up in a search that I have and everything else. And then I share that search with somebody else. Now they have access to that data. So, so there's all this. It's like that string on the sweater that you pull and.

Yeah.

And by the time that you're done, you realize the sweater's all gone.

Yeah.

You have to shore up your data first. So, so when I, when I'm going to customers, that really is one of the, one of the first things that you need to do. Instead of handing out copilot licenses, make sure, whoever your data governance official is, make sure that they're securing their data. Because as much as bring your own AI is a security risk, even internally, understanding who has access to what is even more important.

If you're enjoying this episode and want to learn more about how to start using AI at work, we've made it easy for you. For just $1, you can have full access to the Chief AI Officer community, which will give you additional training, custom software, daily Training calls on AI tools using AI automations, getting more from your ChatGPT sessions, and the business of being an AI consultant, simply go to chiefaiofficer.com insiders to accelerate your AI journey. Now back to the episode. You know, that's an interesting concept because when I hear data, I think data pools and like, like the more technical side, like statistical data science, right? Yeah, but in this context, the data is. It's that document, it's that email, it's that attachment, it's that. And one of the things that I kind of went, oh, like at first I was like, oh, man, this makes so much sense because I'm already like, okay, I'll give you the paradigm here would be my concern. Or, or an individual's concern. Big Brother reading my data or big data, like, like knowing everything that I'm typing. And that's why people have these fears and concerns about uploading proprietary information into to chat GPT or CLAUDE or whatever is that, oh, are they going to be using this to train the model? First off, how big is your spreadsheet? I mean, the model's got 34 billion parameters.

Those same people walk around with this.

Yes, yes, right.

There is an assumption of, of okay, I am, I am okay with this because of the convenience that it provides and that, that needs to be understood. While I am a huge privacy person that, that says, okay, we don't want to overstep these boundaries. We don't want to make sure that it's there. We all walk around, like I said, with this mobile device that's listening to us all the time. We all, I mean I have, I have, I have Google and I have Amazon inside of my house listening and it just broke up. But I have all of these devices again for the ease of, of my life. I'm willing to give up a little bit of that to, to go through. Are they locked down? Yeah, I, I, I do because I read the user agreements. I, I make sure that this, this is secure here and it's not bleeding over through there and so on and so forth. Um, but in a work environment, I, I think it's up to the, the employer to say this is where we need to live and, and, and, and we need to be on the edge of this or at least looking into adopting this. Because if I'm not doing it, I know that my competition is a hundred percent and, and that AI gives that competition an unbelievable edge because as, as things move faster and faster inside of this world, it's harder for, you know, a single person to keep up with everything that lives there.

You know, and that's a, interesting point. Businesses see their competition doing it, they think they're not as FOMO because they're like, oh, we'll just start using it and catch up. Yeah, you can start using it and benefit from it. But I don't know, if you give somebody, your competition a head start with this stuff, like there's a learning curve, there's trial and error, like you won't be able to really catch up to like get at parody again.

So and again, Chris, back to the early days of the Internet. There's a reason pets.com isn't around anymore. There's a reason why all of these other things that you thought, oh, it will never go away. Think of Kodak, think of all these companies that went away. There's a definite edge that will come with having implemented something like this. And I'm not saying you have to jump all in and you have to go through it, but you have to start looking at it. Because as, as, as people, you know, as I look at even television, right. I come from an AV background. Everything else, I look at television. Right. I need 100 inch TV inside of my, inside of my bedroom. Why? It's the biggest that I can get. It's, it's the best there. Right. But if you talk to somebody of a younger generation, Chris, they don't even want a TV in their house. They can sue all of their, their entertainment on personal devices y and not in a group setting. So, so when you, when you look at behaviors, when you look at the way that, that people are doing it, you've got to, you've got to actually adopt to some of these things and understand where your employees are going to actually benefit from it. And like I said, I mean that Moonshot is great. I love those demos. I love watching them and everything else. But I think that there is the everyday society and you and I were talking about it, right. How do you, how do I not pay for this in the first day that I'm using it? But again, it's an intentional. Right. Now let me go and let me look at this and you've got to change it. And people of my generation are not doing that. I know that it's very hard for somebody to change their workflow.

Sure.

And it's very hard for somebody to even adopt new ways of working and things like that. I get it, I understand it. But at this point we are, in my opinion, early days and it's gotta be an intentional way of work.

Well, let's talk about that. Obviously we've got people on all different kind of. Some are in the Microsoft ecosystem, some are G Suite, some or whatever. I guess really those are kind of the only. Well, I mean.

You got Apple Intelligence now.

Yeah.

You, you have a lot of AI that's coming into the space right now. You might not even know it, but you know, like Apple Intelligence. All right, I'm going to reply to a text message. All right, write it for me.

I'm ready to start using that. So where I was going to go with that question was that okay for us? We are cloud based. We're, you know, we use the G suite. We've got a lot of our chief AI officers that Are, you know, have long careers in the Microsoft ecosystem for anybody. What would you suggest would be the ways. Because listen man, I, I've, I've talked to, it doesn't happen as much anymore. But even in 2024, I would talk to people and say, they'd say oh, I tried AI, it didn't work.

Yeah.

And I, I, like I, that's changing. People are understanding more. But how do you get the employees that are in the companies that are like I tried it, it didn't work. I'm not going to mess with my co pilot instance, my Gemini, whatever. How do you get them to what would you suggest that a leader or AI enthusiast in the company does to get them to say no? Just, just here's, here's what happens. Do it this way.

So there's gotta be a champion inside of the company. There's gotta be somebody that says it. If it's the AI officer, okay, it's the AI officer and he has to figure out what does he actually think will benefit his quote unquote users the most?

Yeah.

One of the things that we did here as, as a small team inside of Microsoft, you know, my, my, the, the team that I work with every single day, we actually put calendar events on Fridays to, and Mondays to go ahead and use AI and, and it's, you know, we shared prompts, we shared different things and one of the prompts that stands out in my head of, of those early days that we did was something to the fact, and I don't remember exactly how it was phrased, but it's something to the fact of on a Friday during that calendar event, you ran the, the, the prompt and it asked copilot, look at my emails, my team's meetings, my team's chats and my channels. And did I accomplish this week everything that I aimed out to do?

Right?

And, and is that aligned with my manager's direction? Right. So when you think about that, all right now it becomes, it becomes even more important for leadership to have clear direction on what are your goals and what is happening so that there's no misdirection anymore. Here are the goals that you're going to have. Let's, let's have a Friday where we, we sort of look and we say did I reach those goals or did I, did I reach those? And on a Monday put a calendar event and say what are the goals that I have for this week and what are the meetings and everything that I'm doing to achieve those goals? And we may be, we may Sort of see and become a little bit more productive in, in a top down management style from managers actually being more prescriptive about how, how an employee and how you work. Now this isn't, this isn't, you know, it's sort of oh, your manager is going to look at you and make sure that you're doing this. This isn't over overly management and everything else. These are overarching goals that we want to achieve as a team and am I working towards those goals or am I actually straying a little bit and going out? So when you, when you think about this, you can use AI to sort of make sure that your, your team from the top down are aligned on what they want to do. I mean we can all talk about hey, I'm more productive at home or I'm more productive in the office or more productive here. But again, measuring productivity is very hard. But measuring goals should be an aspect of how are we being productive inside of these things.

Okay, this is, this has been really helpful.

I think that translates. I'm not saying that this is the way that this is, this is only for Copilot. I think that translates to every single cloud based AI that is a work based function.

That's the difference. Focusing on individual tools it like yes, you get value but when everything is kind of in that it's all integrated and you can say here are my goals.

Help.

Like you don't have to take anything outside of your work environment. It's in there. And you can say is this on, is this on task? Right. Like am I. That's an interesting.

I hadn't, you know, I'll say this, and I probably shouldn't say this publicly but I'll say it. Do we every, every year we have to do what's called a connect. Right. That's our self review with our manager to, to say this is what we achieved, this is what we're going to go and everything else. And this year they actually encouraged us to use Copilot Nice. As a prompt engineer.

Right?

Yeah, I sort of took it out and I said all right, I'm going to, I'm going to go to the N degree with this. Right. Microsoft has a ton of guidance on how to write your connect, how to actually think about your connect, how to do things. There are SharePoint sites where it would take you a week to read all these SharePoint sites and everything else. Right. So I actually trained a model using all of the data. Right? Yeah. This is the style of writing that I want you to do. This is how I want you to answer questions. This is what I want you to do. Chris, some of those questions that were in my connect I just fed right to the. Right to the co pilot agent that I had built and it output an answer and it was unbelievable because I shared it with some of the leadership that was here and I said here was the response and their reaction was, oh my God, that's the perfectly worded Microsoft answer to that question.

Yeah, yeah.

Because it said a bunch with the acronyms that we use in the style that they wanted you to write.

Yeah.

And it was perfectly phrased, wasn't specific to me, but it was, it was, it was like if I saw that. That is a beautiful, beautifully crafted response to that question.

So this is on, on Copilot in particular has been an enlightening conversation. I want to switch gears into this. One of the highlights of the ignite which was the agents.

Yep.

And for those who are listening, who have heard the term but may not understand exactly what's going on with an, with an agent process. Do you mind explaining for the layman?

Yeah, I think it's really all the research and development here at Microsoft. You know, you can have a large language model and think of that as ChatGPT. I mean, that is a large language model. It's a. And what, what was a fifth grade education on everything in the world. Right.

Yeah.

Now lately they've actually come to a PhD level.

Yeah.

Of everything in the world. Right. And I think you said it when, when, when somebody experiences and they ask that question, the response is very, very aggressive in I'm right and this is the right thing and everything else. But you may know that it's wrong.

Yeah.

So, so you sort of look at it and you go, all right, well this, this is basically coming and saying, yeah, this is right. But I know that it's wrong because it's a large language model. It has to reason over everything that's ever printed on the Internet, which has right and wrong answers on it. So it has to reason over all of this information that lives out there. Well, what if we take an agent and instead of this general knowledge that we have, we sort of say, here's what I want you to be an expert on. And we look at it and we say this is exactly what you need to know and you have to do. Right. And that's been one of the more effective ways that you get right answers because it's not reasoning over everything that's out there, the right and the wrong. It's reasoning over what you say is right. Yep. And, and agents are really there to sort of be that expert. So let's, let's take it down to an everyday solution. All right. I want to be an expert on fitness. Okay. I'm going to feed everything from the, from the, the, the, the way, you know, the personal trainers, their certification units and everything else. And I'm going to do that now when I ask the questions, it's going to be able to answer me just like a personal trainer I hope would. And it may have a PhD level in that specific subject. But if I were to ask that, you know, hey, what kind of food should I eat? It doesn't know.

Right.

So it's not going to be able to give you that. It may try and say, well, but it's not going to be that, that, that, that right answer that you want. So agents are focused on specific things on call centers. It's going to be great for call centers. You'll be able to actually feed it, here's the, you know, FAQs, different things like that, and train it on what it should know and it'll be able to do things. And that's, that's really where agents are going, is that this is, we, we started with this, hey, let's get, let's get AI in this expert in everything. But what we're doing is we're slowly seeing we get better results when we're actually focused on the actual topic and not. Yes, training it as everything.

So I've seen people that kind of game the system inside of the large language models by being very descriptive in their prompt about you are an expert. We call it the Persona pattern of prompting. And that helps. But if it doesn't, if it's in an agentic environment and it doesn't have the ability to even go out and hallucinate at all, if it's not in its box, it's not in its box. That's much more reliable for consistent accuracy, I would imagine on that.

And, and actually one of the ways that I've seen that, that people do the, the large language models and get it down to better answers. A colleague of mine, Michelle Bauman, actually helped with this prompt. He said, ask me questions before you answer and you can have that back and forth conversation so that it can drill down and know what, exactly what you're looking for. You know, instead of just saying, hey, you're an expert on this, ask me questions before you answer to make sure that we're, we're both in Alignment on what, what I'm looking for. And that, that sort of helps people further understand, like you said when they first try it and they say, no, I got, just didn't work for me. You know, you go through it and you know, if you can have that conversation with it and drill down a little bit, it helps, it helps your answers even more.

So now that you've kind of explained the, the benefit of an agent, what are some areas where people should expect to see early adoption and activity of these agents in their, their workflows, their processes, their business?

There's, there's so many verticals that this is going to help. And let's, let's take one for instance. All right, we were talking about going back to the office. What if you were hired in the last six years and you've never been to an office?

Yeah.

Wouldn't it be great for an HR site to have all of its information, everything else and be able to actually answer questions instead of, hey, well we're all going to get on a meeting and then we're going to see what questions you have. And by the time that you get to the office, you, you forgot everything that you have. Wouldn't it be great if I had a mobile app that actually access the HR SharePoint site and everything else and I could ask, how do I book a conference room? Where do I sit? How do I, you know, you know, are, you know, all these different things that are, that are going to come into play. We're now building actually an AI agent for, for teams rooms where we, we're a manager of all the shared spaces and conference rooms can go in and he can query about his teams rooms. How do I set up a custom background? How do I, you know, how do I do this? And it's all sort of fed back to him from, from the data that's there. Not only will it be able to do that, you'll be able to say, are there any rooms that are not utilized? You know, as we do move back to the office, we're looking at the AI powered workplace and we're sort of looking and saying we can do better with real estate no longer. You know, think about it, Chris. When, when, when we would, when we were a company, a small business, right, and we, we started to grow and we said, all right, I have to think about now hiring a new employee. Well, I knew the cost. Number one, they needed a desk. Number two, they needed a monitor. Number three, they needed all of this. But what about this day and age? How do we figure out the cost of an employee because maybe not everybody's in the office every single day. Maybe I have to outfit their, their home office. So what do we get? And, and where are we seeing returns on our investment? And how do I actually qualify this? Because the hundred years of data that we have on this is what it costs for an employee. Probably no longer accurate at this point. And we're starting to look at things. So if I can have a, a, a, a, a solution and, and Microsoft's doing all of this with AI. If I can have a solution that goes in and are these spaces being utilized? Are they, are, am I underutilizing some spaces? Where is my investment and where's my return on investment and should I be investing more in things? We can bubble those kinds of things up using the data that's inside of your systems and actually help people. So smaller agents are going to be for the users and everything else, but even the broader agents of all those agents working together and querying together. You know, one of my, one of, one of my colleagues here actually took data from, from a number of customers and I think when he said it, it was almost 6,000 spreadsheets of data from customers and everything else he just exported because these are the customers that he's, he's, he's, he's been assigned to, for selling to. Right. And he got all this data and everything else and he asked actually Copilot to actually write a Python script to, to merge all 6,000 together. And it did it. And he verified with, with, and this is one of those moonshot things. I'm not expecting people to do this, but when you combine that together and then you combine that with an agent that can query over that data, let's say is, is there a correlation between the number of people that have adopted Copilot and the number of employees that they have? Yeah.

Yeah.

Is there a correlation between the number of teams, licensed users or E5 licenses that they have and teams Rooms that they've adopted? What does the workplace start to look like when you can combine all of this and say, my most successful customers are adopting E5 teams rooms, copilot and everything together versus the ones that are not doing it and looking, looking to say, all right, well, well, let's just do this. They're not, you know, there's not as, as, as advanced and their, their solutions and their agents aren't providing better results because they're not all here. And we've got our data split up from here to here to Here. And that's like I said, if you let users go, they will start to adopt their own AI and bring your own AI will become a thing and then you don't have access to that data for you as an organization to query over. You will not get the right results. And then you're going to come back and say, yeah, this doesn't work, man.

This has been like I, I'm immersed in this for the past two years and I don't want to say I think I know it all, but I don't hear new, I don't make new connections often I regurgitate the things that, that our own discoveries and our own. But I have definitely made some new connections or some new ways of looking at things because, because of the ecosystem environment, because of the, well, it has access to this and this and this and this and this and this and not just going to chat GPT and saying be that utility player for everything. Right?

Yeah. So yeah, it's listen, AI, AI needs data to reason over. And the only way that you get accurate responses to feed it data, I mean it's the old garbage in, garbage out model of data databases that we had too. You know, if you don't have good data, some of your decisions may not be effective and if you're giving users the ability to go do other things and things like that. And I'm not saying you have to be all in on Microsoft, I'm not saying you have to be all in everywhere. But I think, I think we need to think about when we go, when we go places and we give people the opportunity and we, we don't make decisions about this, where's our data going to go and are we going to lose out because somebody else did go all in in one place, whether it's Google, whether it's, you know, Microsoft or other places. Yeah, you have to think about things and you have to, you have to actually be intelligent about how you're deploying all this.

Dude, great, great interview. So any closing remarks for people who are at all levels, all in, all the way to hey, I'm brand new. The moonshot things. Great.

Yeah. The brand new side, you have to, you have to be intentional with it. I just, there's no other way to adopt new technology and then to be intentional about will work into your workflow. Like I said, I think we're early days. I think you're going to start to see things get a little bit easier. You know, if you think back to the early days of the Internet, we were all on, you know, dial up modems and it took five minutes for a page to load and things like that. And you waited for it because you wanted to see that picture, you wanted to see what was there and things like that. I think that's sort of where we are with AI right now, is that, that you've got to be intentionally, you got to want to go to this, to this place and adopt it. I think as, as the younger generation starts to come to the workforce, they're the ones that will adopt it. And I think that's, that's, that's something that they're, they're, they're sort of ingrained in the, the decision makers and people like that. I don't know a way that you're, you're sitting back, right. I watched people in the 90s say this Internet thing is not going to work. I'm watching people in 2024 say this AI thing is not going to work. This is a, A, a generational shift in the way that there's, there's no stopping this. There's, this is going forward. I watch people, you know, the, the, the people with all the movie stuff that, that comes out and everything else go, oh my God, it's going to take over the world and everything like, like that. I think there's a place for responsible AI, there's a place for responsibility to make sure that some of those things can happen. And Chris, I think if we wanted to go into technically, why, why do people think that? Well, is it possible? Yeah. Do we not know? Yeah, we don't know. That's the whole, that's the whole key of this is, is the same way that I can't read your brain, I can't read the brain of what's going on in some of these things.

Yeah.

And then if you look at quantum computing, I mean, Google just released something yesterday with quantum computing that, that solves an equation in five minutes that would take what, 900 trillion years.

Yeah. It's crazy.

So when you look at quantum computing, when you look at AI, when you look at all these things, we've got to have somebody that's looking at it from, from an organizational standpoint. Yeah. And say here as an organization is how we're going to do it. Organize a champion, like I said, even putting a calendar, invite for your team, your small team, five people, and just come together on a Friday or a Monday.

Yep.

And do this, come up with an exercise that it's just an idea. How can we work better? Right. And everybody Brings a prompt or something like that. Go, go research it on the Internet. I share stuff.

Yeah. Yeah.

You know, it's. It's the idea just, just having this intentional thing that we do together I think will help people listen. I know the large majority of people are not technical out there. They don't understand how to. How to do a lot of this stuff. You know, you and I are technical people. We understand the back end of things and everything else. We need that large majority of people that aren't technical to be here. And if it, if it, if it happens gradually, that's the best way for this to happen.

Yeah. Jimmy, great call, bud. Thank you so much for being a guest on the podcast. And I don't know, is there any. If people wanted to pay attention to your chatter, where would they go for that?

Follow me on LinkedIn. I got a big LinkedIn following with a whole bunch of stuff and everything else. I primarily do it for Microsoft Teams rooms and spaces there, I said. But I do a lot of copilot. I do a lot of things that are out there. I share a lot of stuff. So.

Yeah.

So LinkedIn is where you're gonna follow me? Twitter, kind of, but not really.

Same here. Yeah. Awesome, Jimmy, thank you so much. Everybody will see you on the next episode.

Thank you, Chris. Thank you, Chris. And thank you everybody for the time. I appreciate it.

Thanks for tuning in to Using AI at Work. Don't forget to subscribe for more conversations about how to use AI at work. And a special thank you to our sponsor, Chief AI Officer for empowering businesses with AI education and training. Visit their website for a free AI Readiness Assessment and AI Strategy Guide to help you get started using AI at work. That's www.chiefai officer.com. so thanks to our producer, Evan Des, for making this episode possible. Follow us on Twitter at the handle using AIAWORK and visit www.usingaiatwork.com for free resources to help you harness AI in your role.