AI: From Basketball to Business - Part 2

Welcome to the ECI Pulse, where we break down the latest trends shaping the business world. In this special two-part series, Jeff Schmidt, CEO of ECI, and Rich Eitri, Chief Innovation Officer, dive into the fascinating world of AI and its impact on business efficiency, workforce adaptation, and the balance between automation and human expertise.

In this episode, Jeff and Rich dig into the real-world challenges of bringing AI into the enterprise—focusing on data security, governance, and practical deployment. They discuss how to enable AI with proper guardrails, how to train teams effectively, and why acceptable use policies are essential. 

Join us as we uncover the tools empowering businesses to innovate without extensive IT resources and discuss the future of AI in everyday business applications. Whether you're an executive seeking a strategic edge or simply curious about where this technology is headed, this episode offers valuable insights and engaging anecdotes that will keep you hooked. Tune in to discover how AI is giving us back time, driving agility, and transforming the way we work.

Transcript

Hello and welcome to ECI Pulse, where we explore the latest trends in business. I'm Jeff Schmidt, CEO of ECI. Alongside me is Rich Itri, our Chief Innovation Officer. As we know, AI is about scaling business, streamlining processes, and unlocking deeper insights. You might say time and speed. Join us as we dive into how AI reshapes industries from mergers and acquisitions to investment strategies, and the importance that governance, security, and infrastructure play in AI deployment. We'll also discuss tools like robotic process automation and low-code platforms that empower businesses to innovate. The future of AI is here, creating new opportunities for growth. Whether you're an executive seeking a strategic edge or simply curious about this technology, I ask you to join us as we uncover the answers.

So, you know, obviously we have a long pedigree and history at ECI of foundationally building things with security and controls and then for the client basis based on the nature of our business and the vertical that we serve. So, here's one that now starts to come into view, right? Which is you have this powerful tool, whether it be, whether you're using your own large language application or using a co-pilot, how important is security? Which is maybe redundant or rhetorical to the conversation here because most of our clients would say the same thing. I was like, this is really important. And then secondarily is where do I start, right? Because the idea here, like it just brings back and I keep hearing BYOD in my head, right? Bring your own device and go like, you're never going to be able to bring your own device. We have BlackBerrys, right? So I'll go back to the beginning of this conversation, right? On what you're going to use, right? Like get that toy out of here. Nobody's going to have an iPhone, right? And if you remember in the US, right? Is AT&T had a lockup on the iPhone for two years, right? So BlackBerry had a stay of execution for two years until people started shifting to AT&T because they wanted the fun, right? They wanted gains. And I remember sitting on a meeting and people are like, this is just never going to happen. Like nobody's ever allowed these things in. And I'm like, too late. It's too late, right?

And so when he started thinking about this, right? Is it's not just at work, right? It's your data can be, somebody at home can be using and just giving information verbally to chat GPT or rock or whatever else, right? So like, how do you deal with this? And you have this magnificently powerful search tool that can find anything inside your environment. And then you have the other side, which is I turned on co-pilot and it does nothing, right? So maybe just structurally is like, how do I start to secure my data? How do I make sure that something like co-pilot, which is packaged into my tool sets is ready to run and then I'm actually going to get results out of it that matter. You already gave a couple of use cases just in notes from team meetings, follow, time back, PowerPoints. And then third, which is when you get to the regulators of ensuring that you're actually set and ready to go with governance, risk and compliance.

Look, I think it starts off with the basics, right? When you're implementing any new system, you follow a process and that process shouldn't change for AI, but yet you see people trying to skip the queue a bit because they're so eager to get to the end product, which is the tool. But we've been really using this opportunity and it is resonating with clients around securing your data. So how many people are actually using it? So how many people, big or small, have really done a deep dive around data security, like who could access what? You do some basic permission reviews and if you're SOPS regulated, you have different requirements there, but generally speaking, people haven't done deep reviews around data security. Like think how many times someone starts and says, hey, like they're going to work for Rich, just clone Rich's permissions. But Rich has been here for a long time. Rich might have access to a bunch of stuff. Now this person has access to those things and they probably don't even know it, right? Unless they go browse around or try and map a drive or go to SharePoint and look at every folder that they could try and access.

So what we try and tell clients is like, look, let's start to know the who, what, where, when, why around data. Who could access what? How are they accessing it? All of those things. And now use tools like Microsoft has inherited most of them through 365 licensing called Microsoft Purview, where you could secure that data in a little more of a robust way, a dynamic way with things like auto labeling. And you'd be surprised as we go through this review, how many things we find. Like clients are always shocked at like, whoa, I didn't realize so-and-so had access to this folder still, right? And, you know, it's good to do anyway, right? Cause that's very good hygiene. So starting there around what data you're going to be leveraging, who can access it, you know, what permissions you should have, laying that foundation.

You know, the other big piece is acceptable use. There are so many open source tools out there now. You really need to direct your users to the right place to use the right tool because, you know, users are like water. I always used to say this, right? They always find the path of least resistance all the time. They always do. So if you leave something open that you don't want them to do, like they're going to find it and they're going to use it, right? Even if it requires them to type it in on their phone, right? Not like teenagers. Yeah, like, like T, right? Exactly. They're going to find it. So define what's acceptable and give them a way to use it.

You know, like part of the genesis around Ella, I was thinking about how we're going to use it internally. And then I also talked with a lot of former peers of mine who are CTOs at some large firms. And they were like, Rich, I'm just going to block it. And I'm like, you know, as well as me, the iPhone is a really good example. I'm like, you said that you can't use iPhones out of the enterprise, that people just started figuring out how to use OWA and other stuff on the phone. Like the answer is not no. The answer is yes, but, right? And like, I think it's, you know, figuring that path out on how they could use these tools in a secure compliant way is the better answer than saying no. And, but defining what that path is and what tools are acceptable and what's not, I think is really the next phase of it.

And then last is really enabling them and putting the training in place. You know, that is, you know, again, like you're not going to roll out Salesforce and not train your users on how to use it, but yeah, you're just going to get on copilot or chat GPT or whatever tool you decide to use and say, here, go at it. Like, no, that's not going to work either because you hear me with some examples here, right? There's a right way to ask questions in a wrong way. You know, one of the biggest complaints, you know, a couple of the guys on the team who do the training are like, Rich, we get a complaint all the time around comparing documents. And I'm like, have we just created a template and send it to clients? Because when you just ask copilot to compare a document, it's going to do that. It's going to look at everything and it's going to compare it. And so, you know, the first word we say, hey, these are the 10 things I'm looking for in the document that I want to compare. And can you put it in a table simply? Then you're like, wow, that was really helpful. Like, it actually showed me the 10 things I cared about in a way that I could now analyze it versus some big, like, verbose, you know, answer. So just basic training around that then allows people to use the tools effectively, right? So to me, like that enablement piece that they're at training.

So you hit on purview, which is classifying the data and ensuring that it has the right security controls, but also at the same time, you start to get auditability that you should be doing anyways in the environment. And then hitting on, I think, what's really important, right, is we do think a lot of times is just turning on a tool. It's like, just turn on the tool and people learn how to use it. And we think of Microsoft as it should be really simple and really easy. Maybe, maybe not. But we do a lot of times, like mail is mail, word is word, Excel is Excel, but you do need to train people on how to go do things, right, is the skillset that goes along with it.

The last part of that then is, is from a governance risk and compliance perspective, if I'm a chief operating officer, a CFO, a risk manager, you know, et cetera, how should I be thinking about governance risk compliance then when it comes to AI? I think, you know, the first piece is around acceptable use, but making it acceptable to your organization. You know, everyone goes and uses AI to go write your acceptable use policy for AI. And that's good. Like it probably will cover like a decent number of items. In general, like it needs to be tailored to your organization because look, the first thing that people look for when they're coming in to do assessments on those things are here's your policy. And are you actually like abiding by it, right? So I think that like making sure the acceptable use policy is tailored to your business and how are you going to leverage that technology? I think the data piece is really important, having a documented what you did to ensure your data is secure, what you're doing on an ongoing basis to make sure you're keeping that data secure as you're using those tools.

Jeff, you touched on surveillance, which I think is a big part of it, whether we're doing, you know, whether it's embedded in ELLA or whether we implementing Purview and Copilot for clients we turn on the surveillance and configure the surveillance component. And that's really important for two reasons. I think, look, the regulators are going to demand it at some point, right? Because you're using it in everyday business. The answer, you know, if they ask you a question on why you included something in an investor letter or how you got to a particular analysis on something, you're going to have to be able to prove how you got there, right? So you want to be able to see the questions asked, the answers, and the underlying artifact at the time you asked the question, right?

But the other piece is it also allows you to understand how people are using it in the organization and where you need to like focus efforts to get better use out of it. So like when we do training, we turn it on and then we go back when we do a second round of training and see are people using it the way that we would have expected, right? And we sit with the client and walk through that so they get an understanding of, hey, here's how people are using in the organization. We probably need to do better training in these areas because they're not using it there, right? So I think that to me is key.

And then the last piece is, this isn't like set it and forget it. Like configuring these tools allow you to put governance on an ongoing basis over the platform, right? And being able to show that, hey, like we're enabling AI in the organization, but we have controls around it and we're measuring. I know in Ella, surveillance is part of the recording and the keyword searches, et cetera.

Is in Copilot, is surveillance something that you turn on? Do you define what surveillance does and what it looks at? And then are there privacy concerns that start to go around that from a company perspective? And obviously that would fall into your acceptable use policies that you have inside the business and explaining to the teams, no different than you do for all your other services that you have that way you can and can't do.

But is that like a box you check with? No, I mean, I wish it was as easy as a box. It's pretty easy to configure, but you have to go in and build the rules and configure everything correctly in PureView. So you're recording everything appropriately. The retention policy around it is very flexible and you can build some canned reports around reporting on how people are using it. Microsoft has some really good ROI reports that you can enable as well.

And then you could also permission others in the organization to review it, right? So it's not something that everyone has access to. You have to create the right permissions and conditional access, but you can give the CCO a view, you can give the COO a view so they can log in at any time. You can schedule like reports to run weekly so you can see how people are using it and what are the common things that people are searching for in the company? What are the common questions and use cases? But it's very configurable. I think Microsoft did a pretty good job with it because I think they realized that people are going to want to see how it's being used. And like, look, over time, the regulators are going to allow some PureView over it itself, so.

So as we're winding down here towards the end of our time, there's a couple of pieces here that are probably relevant or maybe just the last question as we go through this, right? Is Copilot Studio, AI agents, I think about how hard it is to adopt something like Power BI in the world. Like you have to become, I don't want to say an expert, but it's learning how to use formulas in Excel to Power BI. And now you have these Copilot agents that are, so just agents in general, but Copilot agents, you have agents for AI that can give you recipes almost out of the box, or you can start to create as reusable recipes where before, you know, it takes a while to start to do the rinse repeat in the business.

How do agents help us and help clients move faster? And I hate to use the term Copilot for Copilot, but I mean, essentially you're getting training wheels in a lot of cases to get you moving with some basic structures.

Well, look, agents are really good at performing specific tasks. And the cool thing about agents is you could chain them together. So if you have an agent that's really good at reading financial statements, and you have an agent that's really good at reading, you know, broker reports, but has a hard time reading the charts in a broker report for the financial data, you could chain those two agents together, right? And you can tell the agent that's reading the report that when you find financial data, go talk to this guy and have this guy go read this, right? And, you know, that might sound hard to do. You could set that workflow up in minutes. And similar to what you mentioned, Jeff, you could ask AI how to build itself, right? So, you know, Power Automate even has a co-pilot built into it where you just type it like where you're trying to do it, it shells out the workflow for you. And you can tell, well, I don't really need FTP, it's gonna be an API call. Okay, you'll go replace it, put the box in for like an API call. And if you have to write custom Python code, you don't need to be a developer, right? You just go in and ask and say, I'm trying to read this Excel file. It'll tell you like what libraries you need to import, you can go get them, it'll go import them for you. It gives you the code, you paste it in. Like, you know, it's very self-service, but you don't even need to do anything with like code. Most of this stuff is already like pre-canned. But yeah, I mean, I think it's even going like beyond like little code, no code, it's just AI code, right?

So I mean, I play around with AI Foundry myself for my own like Azure instance. And like, you can go down a rabbit hole because you can build a lot of stuff like very, very quickly, you know? It's amazing, right? Because I was using Excel and I needed a formula and just thought to the side, I went into Copilot. I'm like, I need a formula, we'll go do these things. And it popped back in and it's like, here's your formula. Go put it in these cells and just drag it all the way down. And I'm like, okay, great. I don't have to be smart at learning the different values sets that's sitting inside Excel, it just told me, right?

Into your flight, you know, creating, enabling some of the suite of tools that are out there today, being able to put a helper to be able to accelerate what you're doing is, you know, does it make you smarter to know how to code it or does it make you smarter to know like the formula and what you're trying to get out of the data? Which, you know, when we talk about this, right, is, you know, maybe summarizing today's conversation, right, is if this is not coming, it's here, right? Number one, it's accelerating faster than ever because it's permeating all different types and points within our life. In fact, I actually watched two AI agents talking to one another on a customer service call who went into computer speak. So if anybody wants to go look at that, you can probably just put, you know, two AI agent bots start talking to one another in a different language. It was insane, right? It was almost like when people used to like, hey, Siri, and then like, it would, then that would go back to Google or to whatever, they get them to talk in chain to one another, but it was just, it was crazy to think that a whole different language came out of this thing to how we use it at business to accelerate what we do.

So there's definitively use cases today on time and speed, right? You can get time back and speed to be able to produce a PowerPoint presentation. You gave a couple of examples of how you've done it, being able to craft that Word document and let that be the basis for creating the PowerPoint presentation for you, or even just getting you out and getting you into flow. Like I'm thinking about writing something for this and here's what it is. Helps me when I do our board reports pretty quick.

Then we talked a little bit about, you know, the, you know, it's not just business people, it's everybody, but then you get into large language applications, large language models. It's really interesting, the acceleration, right? And then you get into governance risk compliance. And what I think you described today is really interesting is this isn't a fear this, this is a embrace this and really start to understand it and draw the smart people in your organization together and look for external ideas and collaborate.

We're in a really interesting and unique industry, right? People share a lot of data and information going back and forth. Most people know each other from some other firm, someplace that they used to be in. So it is a great place in the alt environments, right? Where we can share this information openly about how we're going to be using these and what we do. And even for us in this modern service provider environment, right? Is we do share a lot of information going back and forth with our peers, I mean, not necessarily our special recipes, but at least having a conversation about where people are going and what they're doing. I mean, it's not a secret, right?

No, so I think a really good example of how easy all of this is, Jeff, I had a call with a client and it was a couple of weeks ago. And one of their junior analysts was asking me some basic questions around like how to use like Copilot. And I ran through with her like, look, you get broker reports every day and email. She's like, yeah, I do. And I walked through like what you could do. You can go get those reports, compare them. You could do all that automated fashion before you even get in the office in the morning. That call was being recorded, so they were recorded. recorded the call. She took the recording of the call, put it into Copilot and said, how do I build this? Copilot laid out how to build it. She went in, we enabled, apparently, we went in and enabled Power Automate for her. She took that, pasted in the Power Automate, Power Automate basically built the whole thing for her. Now she has a workflow that basically does everything that I had described in a simple use case, by just recording the call, pasting it in, then doing like, bringing it through, right? And, you know, it may not be the best, but it kind of works, right?

Well, what's interesting is like, there was a conversation, and maybe last point here is when Andreesen was on, on Lex Friedman talking about AI, right? And one of the things that he said was the most interesting and probably most popular use case is code enhancement. Right? So it doesn't have to be first time through, but you can take that and continue to iterate on it, right? And it's a learning model. So you can get better at what you have, right? Like, hey, I like this, but I'd like it to be faster, quicker, and when it's done, can you clean it up and get rid of stuff that doesn't need to be there?

I was really useful in my early days of trying to code, is I found people who coded way better than me, and I'd create really crappy code, and then somebody would come back in and actually fix it and make it work the way it was supposed to, or they would just make it work, right? So it's interesting, right? I think there's a lot of use cases here that come out of this, and the AI vault that's being built for our client base that has the opportunity to give them access to things that they may use, like you just gave an example where we're helping curate this stuff, are interesting models that allow people to get in and start moving faster and have their, not even a POC, right? It's really just an accelerator to a use case to say, does this work, right?

So I think it's interesting, and I'm looking forward to the next one that we get to do here. So I wanna thank you for your time today and working with us, and for those that have joined us to listen to this, thank you as well. This is the ECI Pulse. I hope you enjoyed it. Your feedback is welcomed, and have a wonderful day and a week ahead. Thank you.