EP 39: Innovative Applications of Microsoft Copilot and ChatGPT

Unveiling the Multifaceted Applications of AI

In this enlightening episode of Tech UNMUTED, George Schoenstein and Santi Cuellar dive into the diverse world of AI-enabled platforms, revealing how they revolutionize daily organizational tasks. From software development, where AI assists in coding and debugging, to content creation, customer support, and even controversial fields like healthcare and legal, they explore AI's transformative role.

The discussion extends to education, marketing, HR, financial services, and manufacturing, highlighting AI's potential to enhance efficiency and decision-making.

This episode is a must-listen for anyone curious about the practical applications of AI in various sectors and its role in shaping our future work landscape.

Watch & Listen

Tech UNMUTED is on YouTube
Catch up with new episodes or hear from our archive. Explore and subscribe!


Transcript for this Episode:

INTRODUCTION VOICEOVER: This is Tech UNMUTED. The podcast of modern collaboration – where we tell the stories of how collaboration tools enable businesses to be more efficient and connected. With your hosts, George Schoenstein and Santi Cuellar. Welcome to Tech UNMUTED.

GEORGE: Welcome to today's episode of Tech UNMUTED. I want to remind everybody, please subscribe, like, give us some comments. We'd love to hear from you.

SANTI: Absolutely.

GEORGE: Today we're going to take a look at some of the potential use cases around AI-enabled platforms or platforms that you would use to automate some of the things that you do within your organization every day. Think of these as things you could do in Copilot or on ChatGPT. We're going to sort of bounce back and forth on a bunch of different use cases. Some of these you're going to be aware of. Some of them clearly exist today. Others exist in some form and will evolve over time. I'm going to throw the first one to Santi.

SANTI: Okay, so we're playing ping pong. I got it.

GEORGE: Yes.

SANTI: All right. The first one I think of as far as the use cases, I think is, well, the first use case we had, I think it's software development. That's what comes to mind immediately. For those of you who don't know, Copilot and actually ChatGPT, but generative AI, in general, started in the software development realm, GitHub. For those of you who are developers, you know GitHub. Copilot was initially there, and what it did, it helped assist developers in writing code. Not only did it help in writing code, sometimes it wrote better code than what the human would do.

Most importantly, it helped a lot with debugging. Debugging was, you write code and something doesn't quite go right. You could have the Copilot review your code and debug it for you. My first use case I think about when I think about Copilot and ChatGPT is absolutely software development and how it revolutionized code writing. That's the first one. I'll ping-pong it back to you. What do you think about when you think about use case?

GEORGE: Yes, I agree. I'm not a big coder, but one of the first things I used it for was to try to write some really simple JavaScript that I couldn't figure out. It did, at the time, a decent job of it. Second one that we hit on a lot is content creation. Think of this, we'll hit a little bit later on more of a marketing context to this one, but this is everything you do in your daily life, your daily job. From writing a sympathy card to somebody all the way through editing a resume, maybe changing the context of a document.

SANTI: Sure.

GEORGE: You can do this in a lot of ways and create informal content, very formal content. You can do it in voice of specific people or organizations. It's really helpful to get it crafted in a way that may be cleaner than what you did. You do need to use the instructions really well in this because if you don't, you will get language that you may not normally use. By language, I mean just the approach and the tone of the writing. You need to be very deliberate about what you do. This clear use case, it's, this was one of the early use cases with ChatGPT in particular, and it will probably be one of the bigger use cases as we go forward. I'll flip it back to you.

SANTI: Okay. Let me think about this. Actually, this is a big one. I see another use case as AI or ChatGPT-Copilot type of generative AI, impacting customer support. They use it today. You're able to generate automated responses. Whether it's Copilot or ChatGPT, depending on what kind of support platform or what kind of question, or what data it is you're trying to use, it can handle literally routine customer inquiries. Imagine having a virtual agent asking an AI for some customer-related support and getting an automated response. I also see it as, in the customer support world, there's something called a ticket.

Basically, it's, a customer has an issue, so a ticket gets generated, and then somebody is going to work that ticket to closure. Triaging which ticket sometimes needs to come first, could be overwhelming depending on how large of a support organization you have. I could see AI doing that for you, AI literally doing that triage and assisting with how to categorize the different types of customer requests coming in, support requests, and how to route them too. That's immediately my other use case just in the customer support realm. What can you think of next?

GEORGE: I'll throw one out. This is probably the most controversial use case that's out there today, which is around healthcare.

SANTI: Okay.

GEORGE: There's a lot of risk. Does it diagnose as well as a physician would? Does it miss things? Does it find things that a physician or a doctor wouldn't find? All those are true, potentially. There are some basic things that we're already starting to see emerge and I think will emerge more. Assisting with documentation, potentially taking notes, ensuring that there is quick follow-up on writing scripts and those kinds of things for the patient, monitoring all of that. Has the patient filled it? If it's a critical med that a patient needs and they don't fill the medication, the doctor or somebody else on the staff can then reach out and say, "Hey, this is critical to your healthcare. You need to fill this medication." Some tangent elements of it are around broader research both assisting in the research process, right?

SANTI: Yes.

GEORGE: Also analyzing existing research that's out there. I'm doing a doctoral program right now, not in healthcare, but I'm doing it in business. I review massive amounts of research on a weekly basis to formulate opinions and come up with my opinion based on what's out there in the existing research, and then bring it to the finish line in some kind of a finished document. Having some assistance upfront to be able to identify better and work through the data will be significant going forward. I can tell you most of the search that exists today, and there's multiple databases of academic research, they are standard search terms that we've been using for the last 20 or 25 years, and they are really ineffective at searching.

SANTI: That's a good point.

GEORGE: I'll flip it back to you to see what you got.

SANTI: Okay. I guess I have a slightly controversial one too. At least it became controversial early on. Anyway, I think about the education field. I know that there was a lot of early talks about penalizing students for using generative AI to create their work or to improve on their work or whatever the case may be. I know you and I were like, "What are you talking about?" It's like using a calculator, right?

GEORGE: Yes.

SANTI: Embrace it. Don't push back on it. Okay, fine, but I'll tell you some other ways that the education industry or sector could potentially use as maybe with tutoring. Listen, every teacher or professor would love to have all the free time possible to tutor students. I know they enjoy that because they like to see their kids excel and their students excel, but they can't be everywhere. Maybe it's a matter of setting up some generative large language models with very specific instructions to help tutor a specific student. It could be personalized to that particular student.

You set up, you walk away, the student walks through the process. As the student asks questions, they're getting answered, they're being tutored. I see that as a huge value. I don't think that should be controversial, right?

GEORGE: Yes.

SANTI: Then also, I see it as a help for the teacher. You said it earlier, from a content creation perspective. Why couldn't a teacher or a professor use generative AI to create their content that they're going to then transfer that knowledge over to somebody else? I see education as a use case. You come up with the next one, let's see what you come up with.

GEORGE: Let me hit on that ed one a little bit further even. There's some interesting stuff in that tutoring piece where it could help in note-taking.

SANTI: Of course, yes.

GEORGE: It could help in then, taking notes in class as an assistant, looking at the student when they're doing their homework, and then could interact with them around the areas where they see a weakness in their homework. Maybe even look into the school system and see exam scores and understand there's a weakness in the exam score and suggest an area where they might want to focus more than others.

SANTI: George, here's one. The student gets the exam scores back, feeds it into the AI, and prompts the AI, "Help me understand what I got wrong," and so now there's a teaching moment from the AI.

GEORGE: Let me hit on another one, and again, sort of sat on the first one that's controversial. This one has a little bit of controversy as well, potentially, but it's, how do you use it within the legal profession?

SANTI: Oh, yes.

GEORGE: There have been a couple of examples where people would ask ChatGPT for a recommendation on something for a case they were doing. It made up cases, and they presented those cases to the court that didn't really exist.

SANTI: [laughs] Wow.

GEORGE: That's a technology issue, right?

SANTI: I had not heard that.

GEORGE: It's what they are already grounding on, and it's the wrong data, right?

SANTI: Yes. Wow.

GEORGE: Think similar to the academic research that I spoke about earlier, there are so many legal rulings that can be used in court cases and other things related to the legal profession that it's very difficult to research and understand everything that's out there and be able to put it all together in a single place and really think through all of it. This is a huge, huge opportunity, and then there's an opportunity as well on consolidating or organizing and analyzing documents. A lot of times, you have massive amounts of data that come in on legal cases.

You have transactions, potentially, within businesses with mergers and acquisitions where you usually have a data room where information is kept to be able to cleanly look across that in a better search way. Again, back to my earlier example, most search within those kind of things either doesn't exist or if it does, it's the search from 20 or 25 years ago. It's not insightful searching. It is keyword searching that is not always helpful. If you want to comment on that, you can comment, and then I'll let you throw another one out.

SANTI: I was just thinking about, sometimes people just need a simple contract. Maybe the AI prompts some questions. You input the answer and address a promissory note.

GEORGE: Go for purpose, right?

SANTI: Yes, go for purpose. Right. I need a promissory note, or I need an NDA, or whatever the case. Rather than pull an attorney out of the field for that, maybe you just have the AI do it because that's boilerplate language for the most part. Anyway, I'll tell you what. I'll pick up the most fun of the use cases because we live it all the time every day in our lives, and that's marketing. I'll just speak about what we do. We as a marketing team, George and I, we're on the same marketing team. We generate content, and we generate images. In fact, I'll tell you, sometimes the content that we deliver to you on this podcast, we generate it through AI.

We'll come up with a narrative with all the research that we found or the topics or the notes from meetings or whatever that we went to, and then we ask the AI, "Hey, can you clean this up and summarize it in a way that we can then consume the data to be able to deliver on a podcast?" It'll do that for us. That's, we can use it for advertisement. We can do it for creating blogs so we can generate some traffic. We can use it for email campaigns. If you want to alter the message or if you want to change the tone of an email to a customer or to a prospect, it'll do that for you.

I even use it for market research. I think we said this very early on in our podcast how, from the Power Platform, using Power Automate, we have built-in workflow automation where it goes out and researches specific RSS feeds and based on our areas of interest, finds articles and research and findings and supplies that to us in Teams. I don't have to go out and do this research. I have the AI do it for me, and then I go and pick and choose which ones I want to zoom in on. George and I could probably give you 100 examples how we're using AI from a marketing team. Marketing, I think, is one of those areas that this generative AI, large language models, all this stuff really shines. That's my next use case.

GEORGE: I think that one of the key things that stands out to me with what we do from a marketing standpoint is we've been able to get to a very consistent output from an image standpoint. We use a lot of AI-generated imagery on the website and in other media that we use. The other piece is getting content that gets more consistent when you have multiple people involved in the development process. The AI helps to bring-

SANTI: Good point.

GEORGE: -that tone together in a way that's more consistent and also aids in ensuring we've integrated keywords, for example, into content that we might use on the web or in other places that we want to drive SEO from that. I'll hit on another area, and again, there could be some controversy here. All kinds of use cases in HR from monitoring an employee from a well-being standpoint.

SANTI: Oh, yes.

GEORGE: Are they working too many hours and those kind of things. All the way through, even if we just look at a more narrow example in the hiring process, to be able to understand resumes as they come in, in the context of who's been hired before and who's been successful. That success could be driven by multiple things. It could be actual performance reviews or other information. Then taking that information and matching it up against work styles and other things that can be measured through upfront surveying of an employee in that interview process, right?

SANTI: Yes. My light bulb just went off, George. Ready? Let's build a custom Copilot that uses a depository of resumes that has been submitted to our HR department as its knowledge source. Then you feed it the job description and the skill sets that you're looking for and have the AI bubble to the top, your top five candidates that are in your resume bucket.

GEORGE: Then layer on top, by the way, actual direct analytics with the person to understand business working traits and work styles and those kind of things in a more independent way and be able to come up with an employee that's a better fit for the norms of that organization.

Then even carry that through, how do you onboard them and do you have a modified AI-supported onboarding process that treats people differently based on what's been identified as their learning styles, as well as obvious the needs of the position that they're driving towards?

SANTI: Yes, pretty cool stuff, man. My gears are turning.

GEORGE: Yes.

SANTI: I guess another one, and sometimes we start off with the word controversial because people create the controversy. In the financial industry, why not? Financial services? Listen, I'm not saying that it has to go out and negotiate financial terms. I can see how you can use AI for data analysis on financial outcomes. It'll do that. It actually, I think, will do that very well so long as it has the data. It all comes back to the data you're feeding it. If the data is good data, your output is going to be good. If you have dirty data and it's not optimal, then you may have even AI hallucination from bad data.

For the most part, the financial industry is very good at keeping clean data. It's almost like they have to. If you feed good data and you ground this AI on it and you ask for a data analysis, man, these bots are amazing at analyzing data, and they will give you really good outputs. You know what? It goes back to even communicating with customers. If you have a banker who's trying to craft an email about the next HELOC he wants to close, and he wants to come across as somewhat professional, they can just type up their version of the email and have the AI polish it up.

Now it goes on, and it looks good. It reads well and it's clean. I could see, I could think of so many things that the AI can do in that industry in general. To me, data analytics in the financial services arena, yes, AI's rock for that. That's one thing they do very well is analyze data.

GEORGE: Yes. To be able to even drive them, we've talked in other podcasts about autonomous AI to be able to drive stuff even in a high-touch environment. If you think of wealth management, there's a lot more person-to-person. interaction between the financial advisor and the client, but if you think back to the beginning of COVID as an example, there was a huge drop in the market. Everybody got a bit freaked out about it, like, "What do I do? Should I be selling all my stocks? Should I not be selling all my stocks?" I had somebody pass along a market report from one of the bigger wealth managers that really broke down multiple things that had happened over the last 80 to 100 years and where there were critical impacts on the market that caused a significant drop.

The real outcome of it was the market will recover, the market tends to always recover. That may not be the case in the future, but in the past, the market has always recovered. Be able to proactively have AI communicating as that drop is taking place and maybe with some level of input but to allow interactivity then back from the client because I suspect many wealth managers the day things really started to get bad with COVID and markets, you remember dropped dramatically, thousands of points over a really short period of time, to be able to have an active dialogue with people to be able to get them more comfortable with what's happening within the balance of that client relationship.

This is where some of the controversy comes in is in, what risk does that create for the financial advisor and their firm in allowing AI to do that advisory work? Are they really fully working on behalf of the client, and are they able to prove that? How were those decisions made? Those kind of things.

SANTI: Yes. I think they can get there, and I think the ability to imagine having an AI that's customized to that specific client, and it only sees their data, their investment portfolio, it can't see anything else. That one bot is so focused on one particular customer and their data and their trends that I think they can get to a point where it could probably give pretty solid financial advice. I think where it gets tricky is when you have a bot that's seeing everybody else's data and trying to make assumptions now because welfare management is very personalized and it's very unique to that individual and to the choices they make. I think they can pull this off but it has to be grounded on that particular client alone. That's just me thinking out loud.

GEORGE: On the data, but the other thing that is a gap in that space is, has the advisor really asked every question that they need to ask?

SANTI: Yes. Good point.

GEORGE: Do they understand everything about the person's background?

SANTI: Good point.

GEORGE: Are there other ways within transactions that some of that might be visible? They hopefully wouldn't have missed this, but do they realize that they have a student starting in college? Do they see within the transaction, college transactions? Then payments, for example, to a university, do they then, in the next meeting with the client say, "I wasn't aware you had a student in school, I can see within your transactions there's been payments to Villanova University," as an example, "can we talk about that and what do you think those withdrawals will look like over time?

Let's talk about the impact now." Again, simplistic answer that hopefully someone wouldn't have missed in the process of onboarding that client. There's probably more nascent ones where even in the transaction level detail, things are occurring like you bought a new car. "Oh, I saw you bought a new car. We hadn't spoken about that. It's drawn down this account." Let me hit on one final one. There are elements of this that are in place. There's been automation in this field, clearly, that's advanced quite significantly over time, but it's around the manufacturing space.

SANTI: Oh, yes.

GEORGE: It's, there's elements of quality control within there. There are manufacturing-like things that will emerge. You will fully see a transformation within the restaurant space as an example where you could almost go to 100% automated environment in a McDonald's, from everything through preparation of food, through starting with the ordering, the preparation, the payment, the delivery of the food, the quality control checks within that. That could all happen in an automated way and give you the most consistent outcome you could likely have in the production.

Are the fries the same at McDonald's, at every single McDonald's every single time? No, they're not. Do I personally notice a difference in the fries? Sometimes I do. It's things like you can tell when the fries have been sitting there for 20 minutes versus when they just came out of the fryer. Then there's elements behind that of, are you really looking at all the maintenance requirements on everything within the manufacturing environment and proactively going in when there are minor variations that can be seen in data that are coming back and understanding over time some variations are fine and they can be left for a period of time?

SANTI: Sure.

GEORGE: Others are going to quickly go to a critical issue.

SANTI: Correct.

GEORGE: Again, if you bring it back to if you term a restaurant as a manufacturing environment, does it start to see a variation in the quality of the fries? Does the oil need to be changed? You can even build in, from a CX standpoint and a service standpoint, immediate feedback. "How were your fries today, Santi? Were they a 10 out of 10 or were they a 7 out of 10?" [laughter] If they're a 7 out of 10, we're going to ask you why they were a 7 because last time they were a 10.

SANTI: Right. What changed?

GEORGE: Maybe spotting something in that process to create more consistency in the outcome. I realize people in some of these kind of environments, there's an interactive component. You don't want a sit-down restaurant to be a robot coming to your seat in those cases, right?

SANTI: No.

GEORGE: Some people might like that. You do want some level of interactivity, customization, those kind of things.

SANTI: Consistency.

GEORGE: In higher volume environments, you don't.

SANTI: You want consistency, right?

GEORGE: Yes.

SANTI: If you have a favorite place, it's a favorite place probably because their food is very consistent. In fact, recently we went to one of our favorite places, and my wife made the comment, "I think they changed something about the salad. It doesn't taste the same." Then when she asked the question, yes, they changed the dressing. She goes, "Why would they do that when it was so good?" Humans can detect change, especially when it comes to food. Yes, to your point, if something is not consistent, if something is off, your best immediate feedback is your customer who's a regular because they'll be like, "Something ain't the same about this."

That triggers something else in the background. I agree. That's pretty cool stuff. Folks, listen, we just came up with 10 potential use cases for generative AI. Some of these you might be able to relate to. Some of these you may not relate to. Hopefully, it gives you a sense of how AI can be applied in your daily work life. Hopefully, also helps to calm down some of the fears of how AI is going to replace everybody's jobs. As you can see, a lot of the examples we gave is really about making the human better at their job. Anyway, folks, with that said, please take this time to subscribe on your favorite podcast platform.

That could be Spotify, that could be Apple Podcast, or that could be on YouTube. Now's the time to go ahead and hit subscribe and click that little bell so you get an alert, a notification when we have a new episode ready for you. Until next time, folks, remember this, stay connected, stay curious, or is it stay curious and stay connected?

CLOSING VOICEOVER: Visit www.fusionconnect.com/techunmuted for show notes and more episodes. Thanks for listening.


Episode Credits:

Produced by: Fusion Connect

Tech ROUNDUP

Expert insights, exclusive content, and the latest updates on Microsoft products and services - direct to your inbox. Subscribe to Tech ROUNDUP!

Tech UNMUTED, the podcast of modern collaboration, where we tell the stories of how collaboration tools enable businesses to be more efficient and connected. Humans have collaborated since the beginning of time – we’re wired to work together to solve complex problems, brainstorm novel solutions and build a connected community. On Tech UNMUTED, we’ll cover the latest industry trends and dive into real-world examples of how technology is inspiring businesses and communities to be more efficient and connected. Tune in to learn how today's table-stakes technologies are fostering a collaborative culture, serving as the anchor for exceptional customer service.

Get show notes, transcripts, and other details at www.fusionconnect.com/techUNMUTED. Tech UNMUTED is a production of Fusion Connect, LLC.