Kirstin Burke:
Hello and welcome to DataEndure’s July TECH Talk. We are going to be discussing today how to unlock AI productivity with Microsoft Copilot. And interestingly enough we had a lot of registrations today, a lot more than usual. And I think it’s because we see AI everywhere, right? Every boardroom, every headline, every product. And I think a lot of people are feeling overwhelmed by it. And you’re not alone.
The customers we talk to, prospects we talk to, are asking, what’s your AI strategy? And I think people feel like it’s so big, they’re trying to figure out what to do with it. In fact, a recent survey showed that 70% of business leaders are exploring AI and only 15% of it have put it to work. I’m joined by Shahin Pirooz, our CTO here at DataEndure, and he’s going to help us unpack what the big deal is about Microsoft Copilot. He and his team have been in the trenches really taking Copilot apart and looking at all the different Copilots. And we really today want to talk about what we can do right now.
So we want to cut through the hype, we want to cut through the big shiny objects and we really want to talk about what we can do, what organizations can do right now to really drive impact across their organization. So, Shahin, welcome. Glad to have you here as always. And I think the best place to start is talk to us about Microsoft Copilot, why use it? And what’s the big deal?
Shahin Pirooz:
Yeah, I think the best way to break this down is if we do that first, to start with the introductions, and then give a couple use cases about how this–
Kirstin Burke:
100%
Shahin Pirooz:
And then I’d love to wrap this up with two things. One of them is some best practices around how to get started. But the other component I think is critical here is Microsoft’s kind of confused, in great Microsoft form, the consumer base with all of their marketing changes and notifications. When we say Copilot while it’s merging into a unified thing, it’s still a bunch of different Copilots. I think a little bit, hopefully what you’ll walk away is a little bit of demystification of what is the Copilots and how do they become a Copilot. And so that’s kind of what I think would be helpful if I was listening on the other side of this and jumping right into it.
At the root of it, Microsoft has basically created an AI assistant. And they’ve done this across the Microsoft 365 app ecosystem, things like Word, Excel, Outlook, Teams, and PowerPoint. They’ve also built into Windows a Copilot software, which if you’re running Windows, you see Copilot in your taskbar and you can pop it open and talk to it and ask questions. You know, very similar to what Apple’s done with their AI agent but it’s more natively integrated into your Microsoft ecosystem, which is kind of where the big why or what’s so exciting about this comes from is Copilot has access to what you have access to. So now in a corporately secure environment, you have the ability to get this AI assistant to help you boost productivity for your people.
And it comes in like three different forms. One of them is generating content. So you’re going into a meeting and you need some help putting together a presentation. And you want to say, I want a presentation about this meeting we had with this customer. And give me scenarios X, Y and Z and it pops out a sample flow for a presentation that you can talk about. You want to create a case study based on a project you just finished. You can feed it the project documents or point it to where they are in your Microsoft Project management platform or project files or whatever the case may be, and tell it, give me create a case study based on this but genericize it. Don’t call out customer names, and it will create that for you. And the last really more interesting thing, well there’s two components. Aside from generating content, it’s like one of the key things is ability to deliver insights. So you can do things like ask it questions about, you know, what’s our sales pipeline, what are things. And there’s different Copilots for achieving these things, but we’ll get into that a little bit later.
Then finally, where Microsoft is taking this is they have jumped on the agentic AI bandwagon and they are creating the ability to do automated tasks. I want you to go create a new user that we just hired based on this information in this profile, and it can go in, do all the things you need it to. I think In a previous TECH Talk, I warned about the risks of agentic AI and what you need to be careful of. And if a bad actor gets access to an account that has access to create accounts, you now are exposed, but contextually it’s that kind of generating content, collecting insights and being able to do actions for you. There’s some pretty interesting capabilities that are coming out. And as a wrap up today, I think I’d like to share with you where it’s going. But let’s start with that as the context. So Copilot is fundamentally an AI assistant that can help you generate content, collect insights from your business and automate tasks that are boring or repetitive.
Kirstin Burke:
And so from a high level, I think we all listen to the promise, the big promise of AI, all the transformation and disrupting business models and things like that. But when you get down to it, what is the thing that’s going to most benefit anyone’s business right now is how do I save time, how do I save money how do I be more efficient in my operations? And what’s so interesting, let’s say 80 to 85% of the market uses Microsoft. You’ve got the ability to start doing that right now. Doesn’t mean you’re not going to go out and focus on those bigger picture things or much more strategic ventures. But you have the ability across your organization from the things you talked about, right? It’s not just one place, but all of your different users with all of the different use cases across the organization have the opportunity.
I read a quote somewhere that said, people are worried about AI removing humans from the model. And it’s like, well, what if AI removes busy work? And what if AI can help the people that you’re paying the money to, to help them really focus on what they do and make that busy work and the information that they need to do those jobs easier.
Shahin Pirooz:
So, I will echo what you’re saying and double down on it. AI is removing busy work and it is doing it for your competitors. So if you’re not removing busy work in your business, you’re going to fall behind because your people simply will not be able to keep up with the pace that somebody who’s leveraging generative AI to create content, to collect insights and to automate repetitive busy work, they are going to be so much faster to market than you. They’re going to be so much more responsive to their customers than you.
The fundamental promise of this generation of AI is productivity of your staff. And at the end of the day, it’s not about removing people. Like you said, it is about making those people 10x more productive. And if you’re not doing that you’re certainly going to fall behind.
Kirstin Burke:
So that leads us to the next point in our conversation here, which is you kind of gave us a little preview on some things that Copilot can do. But give us one or two examples of where someone has looked at what Copilot can do and went, oh my gosh. Where are the things that make it shine?
Shahin Pirooz:
Yeah, let’s break it down into categories that I talked about before. So content creation is an example. You can take a, rather than spending cycles trying to read a document that is 300 pages long, whatever it is, contract proposal, whatever, you can load that document into copilot and, and I want to stop for a second and back up. Copilot is just like any other generative, generative AI.
So it is powered by OpenAI. Microsoft owns 51% or better of OpenAI. It is OpenAI that is restricted and secured within the Microsoft ecosystem. So rather than putting your content into a public system like OpenAI or Google Cloud or whatever, you’re basically getting your enterprise version of it from Microsoft. That’s really the contextual way to think of Copilot. So it is no different than the other generative AI solutions. It is powered by OpenAI. So if you like ChatGPT, it’s basically Microsoft’s iteration of ChatGPT. But you could do what I’m suggesting in anything except you’re now putting your company’s IP into public resources that now are out in the wild and part of the large language models versus keeping it inside your ecosystem. And that’s, from a security perspective, that’s where we think Copilot has an advantage over the rest of the players in the market.
In the context of content creation, imagine taking that 350 page document, a proposal that one of your reps created for a customer with the technology team and they’re about to go and present this to the customer and you just want to summarize it and see what is it about. And you could take that 350 page document and drop it in and say, please summarize this, give me the executive summary and give me the key takeaways and give me what is the call to action for this thing and what problems do we solve? And within less than a minute you will get a one page report, two max, that gives you the high level executive summary. Here’s what we’re trying to accomplish. Here are the key takeaways from this document. Here are the budgetary numbers that we’re putting, whatever prompt inputs you’re asking for, it’ll give you a result associated with that. So rather than spending hours reading this 350 page document, you just spent less than a minute and then another minute to scan the executive summary.
I can tell you personally that I have had it create executive summaries after I write proposals. And it does way better than I ever did. And I’m pretty fond of my own work. So that’s saying a lot coming from me because I’ve got a big old head and that big old head says I can write a better executive summary than anybody. But clearly not. Because the models that they’ve created for processing data, analyzing data into the intentions are, do a far better job than I do even though I wrote the proposal myself. It’s impressive beyond terms I can even put out there. You can then say create a presentation from this executive summary to present this proposal to the customer rather than having them read a 350 page document and give me an outline. And then I want you to create the presentation for me in Microsoft PowerPoint. You get a link with a Microsoft PowerPoint document, you click on it, open it and you’re able to use it. Then you realize another example is you want it to create content for you. Say I want you to help me create the executive summary for this proposal so that I can put it back in. And it writes it, and it writes it in a tone that isn’t you. Now you can teach it, too. I don’t talk about that. I don’t use those words. I don’t use X, I don’t use semicolons, I don’t use whatever. Please fill in this, get rid of any emojis, those types of things. And please remember with the adjustments I’m giving you for all future content you create for me.
So now you’ve created an assistant that you’re training that you’re teaching how to write content on your behalf. So in the context of data analysis, most people aren’t spending time in other tools other than Excel. Let’s just focus on some Excel. You’re in Excel, you have some massive amounts of data and you’re trying to figure out what to do with it and how to get an output out of Excel. That gives you a chart, gives you a formula, gives you whatever. And all you got to do is say things like recommend a formula for me that will give me a column at the end of this that tells me this. The percentage or the ratio of managed services sales versus enterprise services sales. Generate pivot tables based on the data on the screen so that I can show X, Y or Z. And it’s human English, we’re speaking English to this thing as opposed to X equals Y or Z.
You’re able to have a dialogue similar to what any of us would have had with a developer or a information analyst to say, hey, I want a report that looks like this or that or the other thing. You now can do that with Copilot, right In Excel. You open up the Copilot tab in Excel, it shows up on the right pane. You start typing your dialogue with it and iterating with it saying, no, that’s not exactly what I want. I want it to look more like this and you can iterate content generation. And it gives you the ability to give some real time coaching in terms of how you’re doing it, both you to it, and for you to clarify the tone or the output you’re expecting.
Email assistance. So you want to summarize a thread. You all of a sudden come back to your desk after lunch and there’s 25 emails in a thread that people are bouncing back and forth and arguing about something and you’re like, okay, what the heck’s going on here? I’m not going to read 25 emails and get caught up. I am guilty of I come in and I read the last email and jump in and bark and my staff and everybody gets really annoyed and says, you didn’t read my answer below. I said exactly what you said. Why are you jumping all over everybody? Now all I got to do is say summarize this thread for me and within a few seconds it summarizes the points, who said what, what the counters were, and gives you basically a rundown of okay, now I got what the 25 emails are about. Here’s my response. And if you don’t want to write your response, you can say, give me a response in my tone and it will, based on the emails in your inbox, understand how you write, how you communicate and you can teach it and train it to do that as well.
Meetings and collaboration. You can turn on Copilot in Teams to transcribe meetings, generate summaries and action items that come out of that. It’s as good as any transcribing tool that you’ve seen in the market today. And it also does things like prepping users before the meeting so you can send out briefing materials, you can send out content about what this topic of this meeting is, what the last meeting minutes were. All that kind of thing.
Customer service. You can automate tasks in Dynamics CRM service modules to do things like email summaries like I talked about, to have agents that are creating workflows that do things faster for onboarding, for example, and resolution of problems. So I think those are some of the high level use cases that are easy to wrap your head around. And I’ve just talked about four different Copilots.
Kirstin Burke:
Right, right.
Shahin Pirooz:
So, what Microsoft is really trying to do though is rather than saying the Copilot for Excel and the Copilot for PowerPoint and the Copilot for Dynamics, they’re trying to unify the Copilot user experience and interface and they’re getting closer and closer to it.
Kirstin Burke:
Well, as one of the early guinea pigs on our team, because marketing was a very early adopter of, I’ll just say ChatGPT, were a tremendous use case for using it. Just in terms of scale, in terms of being able to do more, respond to more.
But one thing you mentioned too that is so powerful is it’s kind of like your phone. You’ve got this phone and you’re probably using 10 or 15% of what it could actually do for you. And I think with the Microsoft suite, that’s often the case too. You do what you do repetitively because that’s what you know how to do. And I think with Copilot, I’m not an Excel power user, but using Copilot I am able to expand my skill set. Hey, I know what I want. I know the business outcome I want. I don’t know how to make it do that. And so you’re able to be much more effective that way instead of either spending five hours hacking or trying to find somebody else to do it or whatever. So there are just so many ways where once you really start embedding it into your workflow or your team’s workflow. It’s transformational in terms of how you do your job for sure.
Shahin Pirooz:
And you and I, not to call us out, but you and I are Gen Xers. If you think, audience, if you think you’re Gen Y and Gen Z people are not using generative AI out there, you’re absolutely wrong. We’re in the tech industry, so we’re a little more cutting edge than most. But you absolutely have to assume that they’re using it. So, how best to protect it? Give them tools that fit into your security practices and policies and procedures and that you have the ability to wrap controls around to make sure it’s securely being used and people aren’t making mistakes that expose your financial data, that expose whatever. I think on a previous conversation I shared that I was speaking to a CIO who mentioned one of his financial analysts uploaded the last seven years of financial data into a public generative AI solution to get an analysis of what’s the last seven years look like. And it was good data that came back. But your data is now out there.
Kirstin Burke:
Now everyone has it.
Shahin Pirooz:
And thank you to that individual for training the generative AI to help us all a little better.
Kirstin Burke:
Well, you mentioned a little earlier. It wouldn’t be Microsoft if there weren’t some complexities to it. And I know just in our conversations about Copilot it’s, well, it’s that Copilot, it’s that Copilot. And so I did a little research and apparently Microsoft has over 100 Copilots within that ecosystem. So, if you’re an organization that’s kind of thinking about this and saying, hey, that sounds good, I’d like to capture this opportunity that have today now with the tools I have. How do you think about, like when you’re out there talking to prospects or customers about kind of getting this going, how do you get started?
Shahin Pirooz:
That’s a great question. So first thing is you need to make an assumption that your people are already using generative AI of some sort. So find out who and what they’re using it for, and then you can understand, okay, who are the people who are our cutting edge people that would be the best use cases to use this. So identify the top productivity challenges that you have and how your people have responded to those challenges to become more productive on their own. And then let’s map that to which Copilot or copilots can address that same problem within your secure ecosystem, enterprise ecosystem.
Now, when we’re talking about Copilot, if you’re a Google workspace company, you’re not going to get the benefits out of Copilot that we’re talking about. And you can look at what Google offers in a similar space. But you’re not going to get the benefits from Microsoft Copilot unless you’re a Microsoft Office 365 company, user or subscriber.
So as you think about what are the best first steps to take, the first thing I would look at is the people who were on the cutting edge. For us as a company using generative AI were Kirstin’s team, the marketing team and my team, the architects and engineers that wanted to play with it. And so we started to look at, all right, what are the ways we can address what challenges these folks are trying to solve and is there a Microsoft Copilot that can address and solve that?
Now because we were very early adopters being in this space we all spent a lot of time training our own ChatGPTs or whatever tool we were using. And now we have to figure out how to retrain Microsoft Copilot to do some of those things. And there is a disruption associated with that and understanding how best to solve that and how best to extract data from the generative AI you’re using into the one you’re going to use, Copilot, is a factor. It’s also understand that that spend, that is shadow IT.
So a lot of us get concerned from an IT governance perspective when we don’t know where our users are spending money, where they’re going and putting their credit card into a website and spinning up resources and assets and it just comes in on an expense report, which finance may or may not tag to IT, that is basically out of IT’s control. And when I say control, the only thing we’re really looking for is compliance and security. Are people doing the right things? Are they educated? Do they have awareness? And so when you start going down this path, having a AI use policy is an important factor as a starting point. Identifying what the top productivity challenges are and how best to solve them. Training your users on how to take advantage of it. Because if you make this investment, you want people to take advantage of it by showing them how to do it and having a program which teaches them use cases and scenarios where the coin will drop for them and they can have that aha moment that says oh wow, I can do that.
And for those of us that have been early adopters of this, I remember in the early days of generative AI I would say things like, I’m not sure that this is any better than search. And now I don’t ever go to Google search anymore. So fast forward to why. Because the content is summarized, the content understands the intent of my question better. The content is responsive to how I’ve corrected things before. So it learns me, it understands me, and it improves based on that. So just that level of productivity has increased my own ability to get to answers faster when I’m looking for something.
Are there hallucinations? Yes.
Do you have to use your brain to validate whether something’s real or not? Yes.
This is why AI will not replace people. It’s people plus AI that make AI successful. The last piece, some implementation best practices. You really have to spend time on that hands on training and showcasing so that you can get some quick wins and show benefits to your end users. About here is literally what you can do. So when I ask you for that TPS report, you can go and ask Copilot to generate it for you in Excel.
Kirstin Burke:
I think part of it is just showing people the art of the possible. Because if you haven’t used it, you don’t know the scope. You don’t know the art of a prompt, right? Like you don’t have to be super savvy, but including one or two words, that you may or may not have thought about, can completely change the quality of what you get. And so there is, and it’s not like Google search, there is some training and there just is some headspace that you need to work on for sure. One other thing that we ran into just as an organization, I think others do, too, is making sure your data’s in the right place to get the best outcome. You want to talk about that for a minute?
Shahin Pirooz:
Yeah, it’s funny because we started launching across our executive team, Copilot. I have an acronym I’ve used and it was a mentor of both of ours that taught it to us a long time ago. It’s called LUDI. Learn, Use, Teach and Inspect. So you learn something, then you use it. And then you teach someone how to use it. And then you inspect their use to be sure that they learned it and their teaching it well. And they go on and do LUDI themselves. So it’s a way to disseminate information and skill sets throughout the organization. So we wanted to implement that methodology in context of leveraging generative AI internally in Copilot, in the use case we’re talking about.
Having a conversation with one of our executives, they weren’t getting the results they were expecting. They weren’t finding the information. It wasn’t giving benefits. And there was a couple things that were highlights that we learned walking away from that. When you use Copilot, you have the ability to use it in two modes. In work mode or web mode. And work mode is leveraging content in your Microsoft 365 ecosystem. Web mode is like OpenAI’s ChatGPT. It’s using the web for content. And so when you’re in work mode, it has access to files you have access to.
So if you do not keep any of your documents in OneDrive or SharePoint or Teams or groups, it doesn’t have access to anything but your email. So all of your content really ends up coming from email, which is interesting, but it’s limited in terms of what it could do. In contrast I keep, because we use OneDrive as a mechanism for data backup, we back up OneDrive on the back end with our own SaaS backup solution. And because of that, my team all puts all content in Teams, in SharePoint, in OneDrive, because we know the benefits if we lost our laptop.
When I was demoing and doing the showcases, trying to teach by demonstrating, I generated a service specification document or a proposal from a prompt without really any context. I want to create a proposal for this service to a customer that looks like this. Simple prompt like that. Within minutes I had a 10 to 12 page proposal prepared, and it was a ride along, so the folks in the room were also doing it. And the individual that had not moved their files into OneDrive was saying, I’m not getting anything interesting.
It became very clear and very evident that you’re not going to get the benefits of Copilot if you don’t have access to centralized repositories. So, to Kirstin’s point, having things in a team SharePoint location or team group location, which is ultimately SharePoint, having your own documents in OneDrive, giving people access to content by centralizing it as opposed to keeping it for yourself and emailing it around, all makes it simple to be able to take advantage of it.
One key thing was RFPs. As we’re working through RFPs and generating new RFPs, Copilot’s phenomenal for saying, use this as a template and get our answers from previous RFP responses and respond to this RFP. And it is 90% of the work is done for you. And then you go in and fine tune it and clean it and make sure there’s no hallucinations. But all of that requires access to the RFP archive. All of that requires that the folks who are using Copilot aren’t just simply using it without any content as seeds or information.
Kirstin Burke:
Right. I think we’re getting close to our time, but what you had promised at the beginning and what I want to give you time to do as we wrap up. Where is this going? We see where Microsoft is going. We understand they’re trying to make this Copilot ecosystem a little bit easier as we move forward, but where else is it going? What do you see in the future?
Shahin Pirooz:
So what Microsoft is trying to do today is move towards a centralized Copilot experience, Copilot app, which has a unified experience across all of Office 365 or Microsoft 365, which includes Dynamics Sales, financial services. They’re not there yet. They’re trying to get to that. I used to say, when somebody said, what do you mean there’s more than one Copilot? I used to say there’s hundreds. So it’s good to know there’s about 100. I can take the S off hundreds. But but the reality is that creates a lot of confusion because people start asking, which Copilot do I use?
From an IT individual, you’re going to start asking, what license do I give a user? And every license I give is that an extra? How does this all work?
So the unified experience is very good. The ability to leverage copilot and collaboration workspaces. So you use a Word document online in the online Word version and you can turn on Copilot and you can be collaborating with someone in a document and having Copilot generating content and evaluating the conversation as well, the changes that are happening live. That’s a phenomenal use case for how to improve productivity for individuals trying to build something together.
There’s Copilot labs that are coming out that are basically giving you the ability to experiment and do prompt evaluation or testing and also generate AI workflows, and improve those workflows without impacting production and being able to do it outside of the production ecosystem.
There’s some interesting cool new things that are coming from Copilot, and some of the roadmap for Copilot right now is in the space of voice commands, deeper analytics integration, so being able to look at data. If you’re not familiar with what Microsoft calls Dataverse, Dataverse is the backend data warehouse that Microsoft has for your Microsoft 365 ecosystem. Your Dynamics is there. Your file ecosystem for SharePoint is there, some metadata from that. Metadata from email is there. So there’s so much in Dataverse and that’s why Copilot has access to content, and Dataverse has role based security access, based on who you are and what your roles are to the data in the environment. There is an ability to interact and create agentic solutions and analytics solutions with Copilot labs directly into Dataverse.
So you’re not having to interact from a user perspective but from a corporate perspective. You can now start creating AI integrated models that you can expose as applications to your team to be able to do financial analysis, sales analysis, pipeline analysis, so on and so forth, support analysis, all of the things that you need to run your business. And then Microsoft is putting some energy into industry or sector specific AI models. So those models would be available on the back end where you say I’m going to sell to the healthcare industry. Talk to me about what are the key things they’re worried about or leveraging those internally to say we are a healthcare organization and HIPAA is a concern for us. Tell me what we need to do to protect ourselves from exposing HIPAA data. That kind of thing, or PII if you will. So that’s some of the futures that are coming.
And, ultimately today it’s a complex ecosystem. It helps to have a Sherpa to navigate it, and we happen to have got some bruises from climbing that mountain. So happy to help you avoid those bruises. And, I’ve always said start small and focus, and identify impactful areas that we can help you, and we will guide you through how to identify those impactful areas and then scale it out from there. Microsoft is evolving the next generation of Copilot. The current generation is leveraging a previous version of OpenAI’s platform. It’s using 4 instead of 4.0. And one of the things that’s coming out is OpenAI because of some guidance from the industry is getting rid of all the different versions of ChatGPT. They’re basically creating one version. So it’s going to be 5. When 5comes out it’s just going to be 5. It’s not going to be a 5.0, 5I, 5F. And so it’ll make it easier for partners like Microsoft to integrate the newest next gen models into their ecosystem as well. So I think there’s some great opportunities here.
It’s tying back to the comment you made earlier. Generative AI doesn’t have a brain, it’s a inference engine. It’s trying to infer what somebody’s asking based on natural language processing and it’s trying to, based on that inference, respond with content that it goes and finds. It’s not creating content. It’s pulling content together and based on its inference engine responding in a way that it thinks is addressing your question, your prompt. So it can’t replace humans because it doesn’t understand sight, sound smell, taste, it doesn’t understand the ecosystem, it doesn’t understand physics. Those are the things that today you have to have a human. It doesn’t understand emotion. We know that if we send an email in a specific way we’re going to create an explosion. And maybe we want to, maybe we don’t, but we get to make that decision. And so it is human plus AI. It is not AI replacing humans.
Kirstin Burke:
Well, when I think back on what we were hoping to deliver today was a very realistic perspective and data points on how you can start having some immediate benefits from AI, whether it be around productivity, efficiency, scale, using tools that a majority of you already have in your ecosystem. Shahin, I say this every time I learn from you, every time we talk. I hope that our audience has, too.
Like Shahin mentioned a few minutes ago, we are happy to help. We’ve walked this path. We are navigating with other customers, with other organizations. And it doesn’t have to be too complex, but there are key decisions that you make. There are key actions that you take at the beginning that really affect the way that this is implemented, that this rolls out and that it delivers the value that you want. So we would love to to be a part of that guidance to help you all get these gains and these benefits. So, Shahin, thank you for your time as always, and we will see you all next month.