Latest / You've Been Heard / 463- AI Will Fill the Gap w/Erich Gazaui

Transcript

463- AI Will Fill the Gap w/Erich Gazaui · You've Been Heard

• 0:00
  Speaker A: What is missing today, I believe, and what will start to really present itself in less than desirable ways is the idea that the skills necessary to interact with these systems seem to be only typing and writing. But the truth is there's a. There's more. There's more to it. But these interfaces are so simple and so powerful that you can very quickly write or type into one of them. And what you get back is very, very strong. So it causes the belief that all the necessary skills are present. Right when I think that there are a few that are absent.
• 0:41
  Speaker B: Eric Gazawi, everybody. Welcome to the podcast. You've been heard. Thank you. Let's give me the. Give me the breakdown where you work right now. How'd you get started in this whole thing? Maybe you've got, like, a really, really awesome story of how you got started in it. If not, just tell me what your first computer was. I. Something like that. Give me the. Give me the intro.
• 1:01
  Speaker A: Sure, sure. First computer was an Apple ii. It was the first one that was, like, commercially available. My dad was really into technology. He had video recording machines. We had Betamax. We, you know, VCR and vhs.
• 1:15
  Speaker B: Oh, did you sound like you have a very. We have similar fathers. I had. I had a Beta. We had a Betamax and a vhf,
• 1:21
  Speaker A: and he rolled everywhere with this big box. He was a doctor, so it was.
• 1:25
  Speaker B: My dad's a doctor. This is getting real weird. This is.
• 1:28
  Speaker A: This is starting to get a little weird.
• 1:30
  Speaker B: And he golfed on Wednesdays and Fridays.
• 1:32
  Speaker A: He just wanted. He wanted to see what it was all about. So we got one. And I didn't know what to do with it, obviously. I was like, I don't even know. Maybe nine, ten. I don't remember. Pretty young. So then I went to my mom, put me in a computer class. It was good for her to learn how to. To program on the. On the Apple, which was basic. So we did cool little fun things and made little images, and I learned a bunch of stuff. Went to, actually, a sleepaway camp for a few nights, a computer camp, if you can believe that.
• 2:00
  Speaker B: That's really cool.
• 2:01
  Speaker A: And we watched the Last Starfighter because at the time it was a big deal. The Last Starfighter was the first movie where all the space scenes were on a computer. An Amiga, which was a pretty big deal because in contrast to Star wars, where it was blue screen and all of what ILM did. So that's kind of like my starting point. Fast forward a few years, and my father's office was upgrading their computer systems. Their, like, Word Processor and things. And so he brought home one of the machines that he had upgraded from. And at the time it was, it was running DOS, of course, and it was a 286, which is. Heck, yeah, yeah, back in the day, man, that's, that's how we described everything. 286, 386, 46. And then Pentium came and we, you know, we started calling it that. So that was, that's a long time ago.
• 2:45
  Speaker B: So that's a real deal. I wished I was jealous of my friends that had a Pentium. My first computer was a 386. Well, my first real, I guess you could say Desktop was a 386.
• 2:56
  Speaker A: Right. Mine was a 286. And I started learning how to do a little bit in dos. Not a lot, frankly. I learned how to start Word Perfect and start a few computer games, one of which was where in the World is Carmen? San Diego, to give you an idea. Long time ago. And then autoexec bat.
• 3:14
  Speaker B: Move some memory around.
• 3:15
  Speaker A: Yeah, well, so that's the story, see. So I go to college, met a bunch of folks also in, you know, interested in computers or some of them were actually at school for, for computers. I was pre med. I wanted to be just like my dad.
• 3:27
  Speaker B: Dropped out of that, huh.
• 3:29
  Speaker A: Well, that, you know, my dad was saying like, hey, you know, medicine's gonna be a weird thing by.
• 3:34
  Speaker B: I was pre med and dropped out of pre med. I'm just saying there's.
• 3:36
  Speaker A: There's more. No, I did too. And what did it was. Well, my dad was very supportive, but also very cautious. Cautionary tale for medicine. I said, you know, no way, man. I'm going to be just like you. I'm going to go. And I go there and first quarter, second quarter, I think maybe third quarter, took organic chemistry as part of the curriculum. And then I called my dad and I was like, okay, you're right, this is not going to work. I'm out. I'm done.
• 3:59
  Speaker B: Mine was inorganic chemistry. I was a chemistry major at first.
• 4:02
  Speaker A: Oh, wow, okay, so you probably got through okay. I did not. And.
• 4:07
  Speaker B: And I was like, nope.
• 4:09
  Speaker A: Well, what changed it for me was, was that. And my dad. But also I wanted to be in business. I wanted to learn what it meant to be in business, to start a business or to be part of a business. At the time, my school did not have a business graduate level program. So therefore the closest I could do undergraduate was economics. And that's what I graduated. I'm really glad I did. I can tell you stories about that. It's very, very strong by going through those exercises. And in my freshman year in the dorms, in the off time, we wanted to play video games. So my friends had shown me a couple things. Like, I forgot the different games that were available back then.
• 4:42
  Speaker B: LAN party, maybe.
• 4:44
  Speaker A: Well, this is pre LAN party, man. Like we're talking. This is earlier than that. What did. It was Privateer. So it was part of a universe that I didn't know too much about. Wing commanders, the universe. But this cool game where you can be a privateer and you can fly around. But in order to do that and to use a joystick, we had to access extended memory between 640k and 1024. And which megabyte, by the way? Right, Those are. Those are kb, Right.
• 5:13
  Speaker B: Tiny, tiny Open World Space Trading and Combat Simulator released by Origin Systems in September of 1993. That's.
• 5:22
  Speaker A: That's when it all happened. And that, that game.
• 5:25
  Speaker B: This is a pretty sick game actually,
• 5:27
  Speaker A: because I wanted to play and. And I needed to call a friend every time to play the game. And I had to call and say, hey, how do I write that batch script again?
• 5:34
  Speaker B: He's like, cd, backslash, autoexec, bat, dot, whatever.
• 5:38
  Speaker A: Eventually I. One day I wanted to play. It was like after finals or something or after a test. And I called my friend. He wasn't available. And so I just started hammering on the thing until I figured it out. And long story short, that's how I got into the field a year or two years later, same guy and another guy. We got together, started our first company while we were still in college. And our first customer was the university. We, the professor in the political science department wanted to write a book, and he wanted to gather a bunch of data. And he was like, look at all these students walking around. Look at all this data I can just be gathering. And so he turns to his lead student, who was my friend, and said, I think I need something like software or something. I don't know what it means, but that's what I think. I mean, so we're like, we got you. We're gonna go, we're gonna start a company, we're gonna write the software, we're gonna sell it to you. So we did all that, wrote the software. It's a what amassed be a survey tool, right? Wrote the thing, got in front of him, hand him the floppy disks, and he looks at us like, cool, so what do I do with this? And we were like, well, you know, you get a computer, it'll auto install just Put the disk in there. He's like, I don't, I don't know any of that. Why don't you guys go and handle all that?
• 6:43
  Speaker B: So we ran MSP from the school,
• 6:45
  Speaker A: went and bought a bunch of computers and actually built him a mini computer lab in his, in his extended office. We set up a bunch of computers, connected them all, wired the, networked them together. I think it was Novell we were using at the time and you know, accomplish the task. So the rest, as I say, is history from there. That actual company we kind of repurposed as a security firm years later and ran that for 15 years until I went to the next shop. So it was, it was great. It was a great learning experience, growth experience, a lot of opportunities, a lot of trial and error, if you will. But that's where, that's what, how I got started.
• 7:21
  Speaker B: Amazing. You worked at Beachbody?
• 7:24
  Speaker A: I did, yes.
• 7:25
  Speaker B: P90X. P90X. I just think you never forget that. Everyone knows BeachBody. Everyone remembers P90X videos.
• 7:33
  Speaker A: I was getting recruited out of TrueCar because at the time Beachbody was contemplating going public and TrueCar had just done it. And I had a pretty big role. I was a senior director there. So I was relatively senior but not executive. And so I had a hand in it and I had a really big hand in the security side, the governance and compliance, like I think controls. So the person, the CIO at Beachbody had heard of that and heard of me and there's a couple of folks there at Beachbody who had worked at TrueCar. So he said, hey, why don't you come over? I interviewed and looked around and said, yeah, this seems good. It was a really exciting opportunity. I wanted to just put another notch in my belt, you know, like another offering. Let's go, let's go. More publics. And then I got there and I knew a little bit about what I was getting into, but I didn't really know until I got there. I was like, wow, this is like a full on exercise company. And they really promoted inside the building and they had a three story building,
• 8:28
  Speaker B: Tony Horton or something. What was that guy? Was it Tony Horton was the guy?
• 8:31
  Speaker A: Yes, I think that's his name. And they did all the shooting, push ups and stuff.
• 8:35
  Speaker B: Did you see him?
• 8:35
  Speaker A: Well, they did all the shooting in the building and they had a gym on every floor. They had a half kitchen, they gave you coffee and Shakeology which was their product. That's all they gave. As a contrast to the more startup world of the dot com era, and it was just a contrast. Like there was gyms and so you'd come in, dress regularly, end up in a off in office or in a meeting, fine, whatever. End up in your, in your workout clothes, all sweaty like you get accolades. People are like, oh yeah, you were, you were in the gym. That's, that's awesome. That's, that's the kind of environment I was, I was walking into and didn't really know.
• 9:10
  Speaker B: It's not that bad though.
• 9:11
  Speaker A: It's kind of cool. No, it's a good thing. But when I got there, first day, going to orientation and we're, you know, filling out the paperwork and like you'd normally do and then I don't know, like an hour in, he goes, all right, let's all stand up. And you know, I thought he was going to say is stretch your legs and kind of, you know, take a deep breath or whatever. He's like, all right, we're going to get up, we're going to do this exercise.
• 9:31
  Speaker B: 50 burpees.
• 9:32
  Speaker A: I thought, what, what have I found myself doing right now?
• 9:36
  Speaker B: This is actually.
• 9:37
  Speaker A: Yeah, that was, that was an interesting.
• 9:39
  Speaker B: A lawsuit, like fat shaming people and stuff.
• 9:43
  Speaker A: It was definitely a different.
• 9:45
  Speaker B: If you take steroids. Good.
• 9:48
  Speaker A: Exactly.
• 9:48
  Speaker B: Don't tell anyone that live.
• 9:50
  Speaker A: It's a prerequisite.
• 9:52
  Speaker B: This is, that's.
• 9:54
  Speaker A: Wow.
• 9:55
  Speaker B: That's, that's also a, another great story. All right, so you've seen, so you've seen a lot. And you actually kind of had your own, I don't know, entrepreneurial spirit, which I think is very, very important to the world of translating it into business speak. And I kind of want to ask this question again live that we asked to a lot of people, which is after all the years of, you know, autoexec back and making wing commander. I know it wasn't wing commander, it was the other game work and everything like this. After the late nights, the firefights, you've probably touched an, I don't know, an IBM AS400 at some point, maybe seen one ERP upgrades, all this stuff. Now we're into this world of AI. It was, it was nice, it was nice back in the day to be appreciated as that person that keeps all the blinky lights on. And when everything fails, everyone needs you. Sure. But it's gotten old over time. When it is beloved to fix everything, but when everything's working like a well oiled Tesla, we become a cost center. So it's nice to have a seat at the executive roundtable, but how much nicer is it to be heard. What kind of thoughts went through your head? Because we ask that a lot and I just want to know how that's changed throughout your career because you seem to have an entrepreneurial spirit. So I imagine you're good at translating tech speak into dollars and cents.
• 11:23
  Speaker A: I am. And translating techspeak into business process understanding of you mentioned cost center. So think about cost structure and resource allocation and ultimately from a budgeting or planning point of view. So a couple things went through my head. I think it's still an issue for IT leaders, specifically CIOs. I think it's changed dramatically in the last few years with AI being very front of front of mind, top of conversation, et cetera. And now people are starting to realize, wow, there's this role or this person that really understands the systems internal to the company and how information is used and governed and maintained and that has I think put a larger voice on a CIO's table and I think that's great. But what I think you're referring to, how do you break out of the oh when everything's fine like a well oiled machine that you're largely forgotten. You said you're a cost center. That's true. But really what it is and it's the role. Right. Like the folks on my team, really great folks. I'll tell them we all know like we signed up for this. We know that's the reality. When everything's working really well, then you're largely ignored. And I think there's some understanding and acceptance of that. But where I take it is twofold. There's a couple different ways to think about it. One is governance in a wider sense. So security is under me, IT compliance, obviously the rest of the system based teams, IT enterprise apps, stuff like that. When you have governance in a space that can give you an edge. At my current company that definitely applies. We have very sensitive information so keeping it safe is important and definitely a requirement to get the types of contracts that we get. The customers are health plans, very large think like blue shield, et cetera. So it's, it's a requirement and it means it's an important part of the process.
• 13:17
  Speaker B: Is papa the. Is that my. That's correct. Am I pronouncing that correctly? Is that how you should be pronouncing that? Are you in home health care, basically? Am I understanding that correctly or it's.
• 13:26
  Speaker A: That's their space. Health care is a little bit further than what we do. It's companionship. So it's very much Visit, centric, friendly type of engagement. So feed the cat, make lunch, read the mail, read a book, play chess. Very popular activities. Actually give them a ride. Give the member a ride. Because remember, these are primarily older folks. They may or may not have a car, may or may not be comfortable driving. So that's a.
• 13:51
  Speaker B: Now my father's. My father just passed away a month and a half ago.
• 13:57
  Speaker A: Oh, I'm sorry to hear that.
• 13:58
  Speaker B: He was 92.
• 13:59
  Speaker A: Oh, wow.
• 14:00
  Speaker B: Neurologist. And. But no, we had both him and, and my mother I took care of until they passed away. But we had in home healthcare for all of them. And it wasn't, it wasn't like a nurse that wasn't needed because they like, you know, fully functioning but you know, with dementia. So I think a lot of that, A lot of the time spent was. Yeah. Playing chess, going for a ride around the beach, keeping people occupied, asking about their memories, that type of thing. I think it's important.
• 14:28
  Speaker A: It is, it's very important. It's what, it's what drew me to the company to begin with, which I can tell. Tell you the story in a minute, but. Right. Sizing governance. But to align with company, need to align with company size, industry positioning. But other examples could include like for example, when I was at ZipRecruiter, becoming a competitive advantage. So like that whole industry, ZipRecruiter created a marketplace and you know, matching job seekers to employers, people who are posting for jobs.
• 14:59
  Speaker B: Yeah.
• 14:59
  Speaker A: And the safety of that marketplace was something that ZipRecruiter took extremely seriously and that largely landed on a combination of my team, my role, my department, and obviously engineering when it comes to the building of the systems. But the vetting of the job posters, the customers, the paying customers was actually really high on our list of importance. And that differentiated us from our competitors. It created a sense of confidence, both external as well as internal, that when we read about these terrible scams that were taking place and still are, unfortunately in job boards where Zupacruiter's name was never in any of those articles, at least when I was there, I haven't seen. But I don't think that they're suffering from that. And that's largely due to the work that my team did and the work we did in concert with the legal team building. The approach of what it means to be compliant is beyond just data privacy or even security, perimeter security, internal security or otherwise. It becomes about a measure, a posture, if you will. And if you do that correctly, it can be. And for Zip, it was a Competitive advantage. So that means that now the conversation has actually turned a little bit and having the conversation with executive team, which I did, or the audit committee which I was on or the board of directors talking about why investments in these areas are actually driving revenue rather than just a cost center. The other piece of it which is a little less traditional for a cio, but I do procurement company wide. All types of purchases at every level. When we had in office at Zip, we had janitorial all the way through AWS and everything in between came ultimately through my desk. And that gave me incredible visibility into the company. A very wide angle lens, probably wider than any other executive because I got to see so much. And that really helped construct some of the realities I'm talking about. It gave me a lot of visibility into the priority of various different departments. The spend. How you're going to end up measuring risk because you're going to measure it on the likelihood of whatever versus the impact. What kind of opportunity cost is there? What kind of cost dollar cost is there. And that is another element. Those are the two really that I brought to bear and I continue to bring to make it at least believable
• 17:17
  Speaker B: that my group demonstrating value in the. I mean I think there's many areas now where it's demonstrating value. We.
• 17:24
  Speaker A: Oh yeah, just got off our. I want to go a step further to your question like how do you stay top of mind without becoming the cost center? And the answer is demonstrate that there's opportunity for competitive advantage. So that's revenue generation and then separately at least on the procurement side which is part of governance. So that fits in the same kind of ecosystem that when you have the opportunity to have that. That viewpoint then and for me I always promise, like I promise the cfo, I come to the company if you're going to give me procurement, I'll. I'll save enough money in the first couple of months. Pay my. Pay my whole year, my whole first year. In the first year I'll save enough to pay for the whole department. So now we have. It's a, it's a gamble obviously, but I'm very good at that. So we get to a pretty good place. We get pretty close. All of a sudden it's less about a cost center, it's more about well what kind of dollars can we. Can we assign to that department to Eric's group that would further these elements to make it more likely to be a revenue generation competitive advantage or Right. Sizing governance approach. When you have visibility, those elements change the view of a person or a department. And I've enjoyed success in that area.
• 18:31
  Speaker B: And I think there's going to be more opportunities as well coming up in the future. Because the IT leader is the one person that sees every aspect of the company. They see every department, they know what's going on in every department. So that person, if they have the right entrepreneurial mindset and have the business understanding and can learn to, I guess, sell and understand marketing from a, from a, like a business growth perspective, EBITDA and bottom line growth and all that, that person has the largest opportunity to, to add value to the organization. We just got off our fourth roundtable since we opened up the community, which is fair, just did our fourth roundtable and this one was on AI use cases. On what AI use cases people have actually delivered on. And one was, I mean one, one IT director and I won't mention it till I get his, till I get his permission because it wasn't a secret enclosed group was able to deliver an, an additional half million dollars to the bottom line in three different particular areas. And one was, interesting what you said recruiting because you had HR sifting through resumes and all this stuff. So they built one AI engine to be able completely increase that, to completely deliver on that process. Then there was preventative maintenance. You know, it was able to deliver valuable data on preventative maintenance not being done on, you know, tons of trucks and vehicles that are out in the field that we're not getting, you know, that did not have, based on the data, you know, correct preventative maintenance being done over a certain amount of time or something. So they're able to decrease preventative maintenance or unneeded costs by like 47% there, which was like another $250,000. And I'm trying to remember what the third one was. There was a third one and there was one elimination of an employee. There was one complete elimination of a job. And I was kind of joking. Like, see, AI is eliminating jobs, but so did the typewriter, right? So did the, so did the sewing machine. But the point was is it did that. IT leadership did that. IT leadership delivered to the bottom line helped recruit better, faster, cheaper, smarter, helped eliminate unnecessary costs in a very cost, you know, in a very margin tight industry. I think it was oil and gas and. And then there was a third thing where we like just eliminated a completely unnecessary employee that was just doing, you know, some kind of like whatever work it was. So there's a lot of opportunity on the horizon for it if they have the right mindset. Which most of them do, they're very creative people. If they can just get the business piece down, where do you think is that? What's the biggest opportunity for IT leaders as far as gap of knowledge and scalability from their own personal career?
• 21:22
  Speaker A: I think today one of the things I talk about in either cast or I have a white paper online is there's two facets of this, but you're going to hit a ceiling. The question is, which one do you want to buy? What about the internal one or the vendor external ceiling? One way or the other, you'll hit it. You might hit both. There's not necessarily a correct answer for any one company. It would be case by case. But understanding that that will come and planning for that accordingly or appropriately. When a person, from an individual career point of view, thinks about how do you expand, how do you take the next step? The reason I brought that up is to say what is missing today, I believe, and what will start to really present itself in less than desirable ways is the idea that the skills necessary to interact with these systems seem to be only typing and writing. But the truth is there's a. There's more, there's more to it. But these interfaces are so simple and so powerful that you can very quickly write or type into one of them. And what you get back is very, very strong. So it causes the belief that all the necessary skills are present. When I think that there are a few that are absent. And I would say thinking about intent, the finish line success criteria, definition of success itself, I think is largely challenged. And that's not new necessarily, but it is becoming more prevalent because of the speed at which these systems can operate much quicker than human responses. And when you leave a gap in a prompt or in a design, AI is going to fill that gap very quickly. And it'll do so so quickly, so well, so professional and constructed, that it's going to be very difficult to determine which part of what is coming out was actually generated from a content point of view versus something you gave it. And it's helping you organize or it's helping you like a thought partner, it's going to meld very quickly into your own process, into your own content. It'll be largely indistinguishable. And if the gap itself was unintentional, which likely it would have been that you won't even know to look for the gap. You won't know to spot what was filled. And I think the longer term reality of that, as that sets in, what ends up Happening is the creation of the output, whether it's a system design or actual system, like coding or something of this nature, will largely mirror what you were expecting, but the underlying process. And there the output of the system could be very different than what you expect. And that will likely not show itself until perhaps it's too late, until it's very expensive to go change that. It's easy to miss that gigantic celestial body that we call the moon. It's huge. So the idea that you could miss something like that big seems impossible. But if you took a laser pointer and you fired it at the moon, I think from here you only have to be off by like some fraction of a millimeter. You can't even see it, and then you'll miss it. That's very, very, very small. So that's what I see happening. So from a CIO or a lead of it. Understand intent and understand what the company believes, because that is where AI is going to find itself. Either it's going to connect to one or more knowledge bases. Most companies have more than one, but they all think they only have one, or they have a statement of authority or a source of truth, so to speak. Any gaps in those same thing. AI is going to fill them, and they're going to fill them with things it thinks is the right thing to fill it with, that it matches all the rest of the content or it matches the theme. But the reality is those are. Those are gaps, perhaps even intentional gaps. That belief system.
• 25:01
  Speaker B: You just helped me. I think you just helped me add a very, very important question to the podcast lineup of questions. Should I. I'm gonna. I'm gonna ask you it, dad.
• 25:13
  Speaker A: Shoot.
• 25:14
  Speaker B: Okay. It is. What is your most important framework, Commandments, so to speak, that cannot be broken, that will break a company and take it off track?
• 25:25
  Speaker A: It's what we're talking about because it's
• 25:26
  Speaker B: essentially what you're kind of talking about. Like, if we do not have frameworks, if we do not have ways that we make business decisions moving forward with technology surrounding it, then one shall lose data points in their prompts and forget things and hallucinate and be off in some other, you know. Well, we'll be one of the. That statistic of a company that's wasted so much money trying to do an AI rollout.
• 25:53
  Speaker A: Sure. And it is that. That's what I am talking about. There's different ways. I refer to it in the paper I wrote an intent engine.
• 26:00
  Speaker B: Okay. There's intent engine. Break it down, Break it down. I need. We need an infographic for this episode. What is the Intent Engine? What are the five steps of the Intent Engine, or the three, or whatever it is? Let's go.
• 26:12
  Speaker A: I'll tell you a funny story. It's really short, but when I was at ZipRecruiter, a couple of people I worked with at the executive team, great folks, very, very strong. One of them turns to me because as the head of it, I also worked very closely with the facilities group and said, hey, make sure that marketing is included in the event that we're doing. I said, sure, no problem. So I went to the facilities group and said, hey, let's make sure there's enough seats so we can get the marketing team in here. And the facilities person says, ok, who is that, though? Like, what are the people involved? I was like, yeah, that's a good question. There's different components. And they go back to the person asking who was a very senior person. And I said, hey, so who's supposed to be on that list? And he said, oh man, you know what? Why don't you go ask the head of marketing? So I did. I went to the head of marketing.
• 26:57
  Speaker B: Who's that?
• 26:59
  Speaker A: Well, head of marketing had a very similar response, like, okay, so we've got this. We got like the SEO team, the pr. We got these different things. Did he want all those or not? So that's a comically simple but very, very powerful example of what I'm talking about. A company needs to know what it thinks, what it believes. That example is fairly, you know, innocuous almost, but it leads to the construction of everything. Think even simplistic things like a distribution list, marketing.com. who goes in the list? Who controls who goes in the list? Who maintains the list? You can automate that. We automated those things. But even that automation requires rules. So who's going to write the rules? Where do the rules come from? This is what I'm talking about. This is what an intent engine really is designed to curate or maintain or ultimately become that like a knowledge base, like a central source of truth or central knowledge. It's what the company believes relative to its mission, to its goals, to its processes, all of those things. We think of them. And humans are really able to adapt and frankly, quickly to changes, systems less so. But with the speed at which AI systems can operate, those adaptations will come quickly and invisibly. And they'll start, like I said, they'll fill the gaps. So next thing you know, you've got a system that believes something that may be in addition to what the company already believes, or maybe even a little bit different than what the company believes. When is that going to expose itself? That's the real question. People use the phrase, I use it to when do you see your system drifting? Well, this is what I'm talking about.
• 28:35
  Speaker B: Yeah, it's actually, it's actually a pretty mind blowing concept. So if you were to prompt your company, right. And because this comes around to AI training, which is a whole nother topic and I think I feel like we need like a really, really good AI training. We need like a know before for AI training. I'm sure there's, I mean there's ton out there. I mean you could find out all day long. But when you were, if you were to teach AI in your company across all departments, even if we just picked marketing alone, how would they set up their projects and prompts? Would they put in the mission statement, the vision statement? Would they say, we are a company that's about this and our number one challenge right now is this. And every time that we speak and like how would we know that we're not forgetting key core concepts and things like that? And is it that's ever becoming a piece that most IT people would say, I'm not responsible for that. But you kind of are because of your intent. Content machine or your intent.
• 29:35
  Speaker A: In my opinion, we are. I take great care and responsibility in that exact function. And sometimes that requires deliberation, maybe even arguments to help people understand what I'm actually referring to. Once they see it as you just saw and you have this kind of moment like, wow, that's a pretty big deal, they get it and then they give me the reins to go do whatever it was that I was explaining was necessary to drive that. So we had brought in several different tools. Most of the tools that we already had, all the big ones, the names, you know, were driving things like, oh, here's our AI, here's our chatbot AI, here's our search AI, here's this and this. We were really restrictive on what we would enable and in some cases, if the system in question doesn't allow disabling that component piece, we'd either replace the system or downgrade the tier so that we wouldn't expose our staff to those tools yet. Because the tools, the more mature ones especially have function to be able to point the knowledge function like the concept of knowledge inside that system at another system. And that meant that we could really drive towards the idea of a central knowledge store, which was then where you'd have the basic, like the tactical, just raw content, but you'd also have the process around how we set goals, what are the goals, what are the processes and the goals, what are the teams, what are the names? Literally everything that gives.
• 30:55
  Speaker B: I would love to know what that framework is because that's, I mean, it's actually really. I'm thinking about doing it for us right now because we've got, you know, reams of podcast transcripts. We've got the, you know, what's the mission of the podcast to begin with? What's the whole, you know, what's. Then you, Then you move on to brand, branded brand concept. And then you move into the messaging behind it. Then what's most important to the listeners. And, and having that all come from a central single source of truth. But then being. And then like having how things reflect back or point to different systems. Is there, there's a construction to that. And that could be, I mean, that could be a. It's like it, it's literally like your background. It's maps.
• 31:38
  Speaker A: That's interesting.
• 31:39
  Speaker B: And is there like, is there a way to map things? Is there a philosophy to map things?
• 31:44
  Speaker A: I mean, there is, you know, and I've done it. I, I can. I've done it more than once, obviously very recently, but definitely could give you a few pointers. The way to think about it is there. There's two. Two elements. It's a start and a finish. Almost the start. What we're talking about here is looking to be deliberate about statements or beliefs, but equally deliberate that there are no gaps or as few gaps as possible, and when to try to instruct the engines, the models, to identify gaps rather than fill them. That is probably the biggest challenge that anyone would face. These systems move so quickly and are so trained, right, pre trained to drive responses that ultimately fill gaps. Not intentionally, just by matter, of course. That is the starting point. The other side of this, the equally, maybe even more important is the finishing point. So you can call it anything, the success criteria, the definition of done. There's lots of different ways these are referred to, but they're often lacking. And because of the, again, the speed of these systems and the speed of the response, especially the initial response, which is so well constructed, so professional looking, so aligned with what the person is thinking, that they start to believe that the system knows what the finish line is. Because the system must know. Because I asked for, you know, I asked for a couple different questions or a couple different points, and what I got back was this amazing amount of stuff. Then the, the false sense of security sets in, false sense of confidence. And what ends up happening is the person, the user starts to believe that the system knows and then it'll know when the finish line will be reached. That's a, that's a flawed assumption. It's a big mistake. Systems do not automatically know the finish line. And if you think about this, let's say you have a couple of conditions that you want the system to correct, whether that's through automation.
• 33:34
  Speaker B: Give me an example. Let's use an actual use case or an example here.
• 33:39
  Speaker A: Think about, you know, mean time to resolution on, on tickets. Whether that's. Let's, let's take a little bit more of a serious version of a ticket like a, for a DevOps team or otherwise, production engineering team. And you feed the engine a bunch of information, the history of the tickets for the last 12 months. There's all different ways you can do this. And the system starts learning. Here are the types of tickets, here are the commonality, the frequency, the resolution, what worked. And it starts creating for you a playbook that works really well actually. And it's very professional and it's clean. If you got to the place you wanted to get, but did not necessarily tell the system that you got there, it's gonna continue to refine the processes. Now, going into this exercise, the conditions that were at place were considered either less ideal or in need of.
• 34:29
  Speaker B: Can I stop you there for just one second? Are you saying that it went too far? Like if you didn't give it the definition of done?
• 34:35
  Speaker A: Yes, but. But yes, I am. That's exactly what I'm saying. But listen, conditions here, solve. If you don't tell the finish line is here, the system's just going to keep going. But what it's going to do, it's going to refine from current state. Those conditions are the ones you wanted. So now it's going to refine from there. What do you think, what do you think is going to happen? You think it's going to get better and better?
• 34:55
  Speaker B: I don't know.
• 34:56
  Speaker A: I will tell you what will likely happen. The refinement of current state conditions will lead to undesirable outcomes. They think of like a, like a needle that goes all the way around, right? It's like at 100% and then it just flips back to zero reason that that can happen.
• 35:10
  Speaker B: Is it what it starts, like hallucinating or some weird something?
• 35:14
  Speaker A: Well, it's optimizing with the instructions it was given, but it's optimizing For a current set of states that was reached through the initial optimization. And that current set is what the company wanted. We're happy with our mean time, we're happy with the visibility, we're happy with this, we're happy with that. If it's refining, especially if it's an automated sense and it's building playbooks or those playbooks are either automatically implemented or used by people to then implement those Playbooks are going to be a refined set based on current state. And that refined set is taking you away from current state. And frankly, it's a gamble, it's a roll of the dice. Does that actually get you in a better state or does it change from where you are today? And that change can be a degraded state. And that is probably the biggest risk that can be faced from a system point of any system, especially an automated, fast moving AI system, that it does not understand success criteria, it doesn't understand definition of done. So it never stops.
• 36:08
  Speaker B: It reminds me of like just overriding a disc thousands of times is kind of what I was thinking of. Or that old like sketch game where you put the pen in and the thing rolls around and it makes like a star. Like what was that thing called back in the day? And once you get to the end,
• 36:21
  Speaker A: I know exactly what you're talking about.
• 36:22
  Speaker B: I don't remember when you get to the end, it just goes over it again. It gets darker and darker and darker. But. But I think the idea in the future is that no, it will go Terminator 2 on us and it will start learning on its own and it will start knowing more than we could know and iterating well.
• 36:39
  Speaker A: And I think it'll be doing it. So invoking those movies and the first one, second one too, but the first one is one of my favorites, is a great example of what could theoretically go wrong. But in the movie, because of drama and the intent of the movie, it was shown as evil because of the way the story went and the, you know, the sentience of the. What was it? Skynet was what it was called. And then establishing that humans were the problem. And then it went after the humans. What's going to happen here is there's a lot of talk, especially the last couple days with whatever the statement that
• 37:13
  Speaker B: was made about my wife sends me all the, all the, all the, like the Instagram clips. It's like AI is going to end the world in two years, 10% chance
• 37:21
  Speaker A: it kills us all and all those terrible things. But the point is the systems, despite those types of statements and also despite how real these systems behave and they feel and look a whole lot like a living being, they aren't. They're going to do what they thought you wanted to begin with and that's going to keep going. And that's what I meant by success criteria relative to current state. If you wanted to get the meantime, let's say I'm going to use that example reduced. You get reduced to a place that's not only meeting your expectations and your needs, but also is balanced with everything else that's happening in the company. Right. We don't all live in silos, despite sometimes that ends up happening in companies. There are interactions with people, with systems, there are priorities, there are trade offs that are made all the time. If you continue to optimize for one. Again, I'm going to use a very simple example meantime, or something of that nature. Optimizing for that at the cost of other things. And if the system has the authority and autonomy to make those decisions or to make action to be implemented and then to build its own little agent and then go deploy that thing, it'll continue to optimize for that statistic at the cost of other things. And even if you could say and point directly at that and say, oh great, you know that meantime went down even further but the building's on fire. Like who cares if we shaved another minute off meantime.
• 38:41
  Speaker B: If that's the point. Thank you. That's what I was. You said it without me wanting to have to kind of say it. Which is the age old saying of. It's like we solved the wrong problem with the best technology.
• 38:56
  Speaker A: And that will happen with AI in general. I was referring to that when talking about the intent engine. Same idea or the moon, the laser pointer to the moon. Same idea.
• 39:04
  Speaker B: Yeah.
• 39:04
  Speaker A: How long is that even realize what's that like?
• 39:07
  Speaker B: Why are we even going to the moon? Who cares? You know how much money that costs? I think we could solve a lot of other problems than trying to go to the moon right now.
• 39:14
  Speaker A: I think there's truth to that. But the point is all these NASA
• 39:18
  Speaker B: guys are gonna get mad at me now. What are you talking about?
• 39:21
  Speaker A: You can get further down, you can build this beautiful system, wonderful, incredibly well structured, architected data layer everything. But how long is it gonna take for you to realize, wow, I actually built a beautifully wrong system. It isn't what I.
• 39:34
  Speaker B: The data was wrong.
• 39:36
  Speaker A: The data was wrong. But there's all sorts of realities that come out of that. And that's the kind of thing that I'm trying to guard against took a long time. I had to spend political capital with my CEO and president a year and a half ago when, you know, because everybody was like, get AI. Bring AI AI AI. And I said, look, I can do it.
• 39:54
  Speaker B: What did they want for? What did they want AI for?
• 39:56
  Speaker A: Well, they want what every leader wants. The.
• 39:59
  Speaker B: I'm just curious, like did they have a sensible reason? I'm just curious, like, did they? Were they like sensible?
• 40:03
  Speaker A: Sensible, yes, in the sense that, you know, increased efficiency or potential gains in areas are always desirable and that's what they wanted too. The belief, I would call it misconception that you can just sprinkle some AI on top and next thing you know you're driving faster, everything's better. That's. That's a mistake. That's a flaw. And most leaders in my space know that that's a flaw. Being able to convince the rest of the execs or the CEO of that is a different story. But I was able to do that and I was able to be given the time to build this appropriately. So in the other systems I mentioned all bringing them in, rationalizing and all the different knowledge bases, but then bringing in an ITSM or really just an sm. And I say that because every department, almost every department right now is in the system building playbooks and automating tasks both for themselves and from a request point of view. So something comes into revops now it's going into the playbook that was constructed. So that reality took a while to build. But what we have now is profound.
• 41:01
  Speaker B: So what does that look like? What do you mean when you said everybody's doing this now and building playbooks and everything? What do you mean by that? Everybody?
• 41:07
  Speaker A: Sure. So one of the things I'll say to groups, my groups are all largely service based groups, so would be like a people team. But there are other groups, marketing, finance, legal, probably spans both procurement, which I own all of those things. Even if they're not a solely service bureau, they have a service component. Whether that's somebody coming ask for a report or they the group in question is going out to all the other groups. So my message to everybody once we got this thing stood up and sanitized and connected to all our production systems and well built and guardrails. Everything else I told everybody, like think about, I hold my hands about, think about the interactions that you have both ways. When you're asking and you're being asked, think. Can that be built into an automated playbook? We call it Playbook because The system we use calls it a playbook. Okay. The playbook has an approach that's different than a more traditional itsm, which is deterministic, right? It's logic gate driven. If then, you know, yes means this, no means that, go down here. But at the end, if it. If the system solves the problem, fantastic. But if it doesn't, it's pretty much escalate to ticket to human. But in this system, it's not deterministic, it's probabilistic, it's. It's AI driven, Buzzy term. But it is designed to understand the request at a deeper level and respond to the requester with the necessary tools and. Or information that will give them what they need. That couldn't be as simple as a link to the knowledge base, or it can be as complex as actions taken. I'll give you an example. It's my favorite example because it's. It was like the first when we finally got the thing stood up. Oh, I don't know. This is probably December last year, end of last year.
• 42:51
  Speaker B: Okay.
• 42:51
  Speaker A: And CEO slacks as usual, and we're a very chat first company slacks. Either me or the head of my head of it. Hey, I need to do X, Y and Z. So this time he says, hey, I need to.
• 43:02
  Speaker B: So now we've got an interaction. We have an interaction going back and forth.
• 43:06
  Speaker A: And it's funny. Quick side note on companies that sell these systems, and this goes back to like, even, you know, before AI was top of mind. Hey, you know, if you put a system like this in and it's got ticket deflection, that's the point, right? So Phil comes in and says, hey, where do I find the PTO policy or where do I find this? And then the system responds, great. Ticket deflection, no ticket. And what they'll say is, hey, look, what's the worst that'll happen? And the worst is, oh, you'll get a ticket. You're no worse off than you are today. True statement, but actually I think it's even more important than that. Even better than that, which is now, in addition to quote unquote, no, no worse off because you get a ticket anyway. Now you've got all those interactions captured. You've got everything that didn't work or wasn't answered, and you've got the time inside a single channel. By contrast, if I asked my head of it, hey, you know, let's talk resources, let's plan for budget next year. Let's talk about whatever we need head systems. Otherwise, how Is he gonna figure out what his capacity is? If everything's in a chat across hundreds of different people, thousands of different channels, it would take him days, weeks to try to figure out where all that time was going. So now, with a system in place, even a basic system solves or at least addresses ticket deflection, as well as tracking of all of the interactions prior to the ticket being created in the first place. Fantastic. But now there's one step more. The step more is the system can take further action and it can do it based on the probability, just like any of the responses you're going to get from any of the main models that it's deciding or basing what it thinks you're asking on all the access it has, including to knowledge, but to systems. So the example we gave back in December, CEO says, hey, I want to take my old laptop. It's really old, it's crappy, and we wouldn't reissue it anyway, so why don't. Well, I just want to take it for personal use. And we said, okay, cool. Message the system rather than messaging me or my head. And he did. And he said, hey, I want to use my old laptop for personal hit control.
• 45:04
  Speaker B: Alt delete.
• 45:05
  Speaker A: Well, what it did instead was it said, first let the system, all by itself, no human interaction, reached out to our MDM and figured out which one is the older machine. So I got the serial number. Then it reached out to our identity. Our IAM identity and access system initiated MFA to him to prove his identity. Then it reached out to the knowledge base to get him the link to do the work. Then it reached out to our key store, grabbed the encryption key matching the serial number, sent all that to him in a bundle, and he was able to do all that work on his own. So previously that would have been at least one. The, the using of that machine.
• 45:47
  Speaker B: Yeah, how did that even get approved through the system is what I'm curious about.
• 45:50
  Speaker A: Yeah, so there's the, the, the combination. The jump between MDM and the identity system will check. That's what it was proving. It was proving he is who he says he is. And he's also in an approval chain that would allow him to approve such a request. All of that is. Is contagious, which is what you mapped
• 46:04
  Speaker B: out back to the beginning, full circle back to the, you know, the, the frameworks. And this is. Well, he, he's the guy that gets the email from info at. So he's right. You know what I mean? So he's in charge of SEO and marketing, who we don't use.
• 46:18
  Speaker A: So that, that example is the one I like to use because it's very powerful. It really tells the story. But it's also table stakes for us. Now that was, that's the entry point. What we do now is 10 times more than that across every kind of department you could think of. So legal procurement, the people team, everything from the leave process to the corrective action process to you name it, they're all now built as playbooks. So a request comes in, it runs through the playbook, they build a bunch of stuff. Depending on the process, they'll get the documents ready, they'll form files.
• 46:49
  Speaker B: Is basically SOWs or SOS. This is basically scope of work type of things. Playbook scope of work. But we're taking it like a step further.
• 46:56
  Speaker A: A playbook is closer to like a step by step. Whereas SOW is generally used to describe the work that will be done and as a, as a set of deliverables. Whereas the playbook is more like, okay, so based on these deliverables, this is step one, step two, step three, the playbook. And the reason why it's interesting that the company in question used the term playbook is that it is actually different than a deterministic playbook, a more traditional, like the big names that we all know that build all of their routing engines, even if they're using smarts of some kind. This now gives the ability to have a true pre trained answer to the person asking, I need this, I want this, how do I do this? Where do I find this? Those can be straightforward answers, but they can also be actions taken on behalf of the user. Now if you get really fancy, which we've started to get a little bit ago, that system can now start purposely impersonating with, again with purpose, the user to give them access to things like, hey, what's my PTO balance? So then it'll say, hey, can I put in a, can I find out my PTO balance? Can I put in a request for pto? The system can help you do all those things. It'll use your credentials. It'll literally pass that authentication through to you to say I approve for you to go get that access. And then it'll start taking actions based on your ask. That's probably a lot of.
• 48:20
  Speaker B: If there's a bunch of CISOs listening to this today, they may be, they might be cringing right now.
• 48:25
  Speaker A: I cringe a little bit too sometimes. But these are user based, well sanitized environments that we understand and know what's taking place and every one of those things has a gate. You can get if you want, you can get more automated, you can get more autonomous and let things start doing on their own. But at least in my world, I would prefer it a little bit slower, a little bit more supervisory before we get down that road.
• 48:48
  Speaker B: If I start taking notes, it's really, really good. So I've been taking notes throughout this thing. I'm taking notes because there's like key things that I don't want to forget and just some really good ideas this conversation has sparked. I've got one last question to ask you, but just, just so you know, kind of the things that I've been writing down because you mentioned that you did a lot of procurement and we do a lot of procurement as part of our like triple threat supporting leadership. We do it for free. Um, and one of the things that we've been thinking of doing lately is measuring ticket requests for external vendors across all different marketplaces. Because a lot of times people are asking, well which vendor is better than the others? And there is no right fit always for somebody as like you're unique group of end users might be different to others. It's kind of like why do I use Verizon or T Mobile in some cases? And it could just be those maps that are behind you. And, but. And the idea came up from wanting to create vendor scorecards and kind of more of a living breathing vendor scorecard that would have the current pulse of actual vendors. Because a Gartner Magic Quadrant snapshot doesn't really do much. It's a, first of all, it's a snapshot in time is how they were doing at that certain particular time. It may be a pay to play model. I'm not going to say that it's is, but it is. And it may not be an accurate picture of vendors and what people go through, but if we had a live breathing data from third party actual CTOs and CIOs of well, I got this many tickets through Comcast and I got this, I've had to do this many through AT&T and there was this many follow ups on the exact same ticket. How cool would that be? From like a. That would be just a real rating type of vendor scorecard stuff. It wouldn't just be G2 four and a half stars and you know, it wouldn't be anything like that. So that was one of the things that came up. And then because one of the things that we started doing was well, can you just look at all the emails across all of my consultants and can we just look up how many numbers of requests were made on a certain rfp? And, and then it's like how many man hours were involved, how many meetings were involved, how many, you know, you know, changes to a contract or terms and conditions. And now let's just rate everybody's MSA across against everybody else's msa. And now it's great. Everybody's, you know, you can really get, you could just do a lot more than you used to be able to do. Kind of like you said back in the day, it was in a chat stream and, or it was in a big slack group or something like this. And it's kind of like, well, we know it's there, but in a matter of 15 minutes to 30 minutes we can actually correlate all this data into some meaningful stuff. Did I understand, was that, am I understanding this correctly from your chat streams and stuff like that and grabbing important information?
• 51:26
  Speaker A: I think there's companies out there that are doing this slash, attempting to do this. Depending on the company, they get more focused on, you know, pricing or satisfaction. Like you mentioned, it would be interesting to see repeating elements from a ticket point of view and how we have
• 51:41
  Speaker B: the ability to do it just because we have a niche group of mid market guys, mid market CTOs and CIOs. So I think it would be interesting to do it in the mid market space because they're the ones that are ignored the most. It's like if I'm not big enough in enterprise, I'm not going to get the special treatment that I get from everybody else. Probably to ask you to share your best thing as an IT leader, whatever that is, that thing that just makes your life simpler, better, faster. What is your best thing? You've got to get in. You're going to get into whatever the Bill Gates, Warren Buffett, Elon Musk, I don't know, mastermind group. And the requirement to get into that group is you need to share your absolute best thing. I don't know if you'd even want to be a part of that group. I'm just trying to make it a big deal. I think I. You need to share your absolute best thing. What is it?
• 52:29
  Speaker A: Oh, it's a good team. There's no replacement for that.
• 52:34
  Speaker B: So it's how do we build a good team? And we need a whole nother 5 episodes for how to do that.
• 52:41
  Speaker A: How to build a good team.
• 52:42
  Speaker B: Maybe what's your one best thing for picking a good team member or coaching a good team member or bringing up a good team member, Maybe it's allowing them to coach up to you, maybe it's open door policy. I don't know what it is. What's the best thing then surrounding. Now we've got another question for the show. I love you for this. What is your best thing for building? What's the most important thing when it comes to building a good team?
• 53:05
  Speaker A: Well, you have to have the vision, but you have to believe that the people who are gonna build that vision or further it and enact it, maintain it, those are your, those are your team members. And when I say hierarchy is a reality we face. But when I work with my team, especially my leads or we do our security forum, which is a wider audience, more people can come in. I only have one rule in those meetings. Everybody has a voice and it's equal. You leave the title at the door. And what makes a good team or how you build a good team is you have to believe what you say. So the things I say, I believe. But you have to have reciprocation and in a sense it needs to be someone who understands what you meant and then takes, I think the word would be advantage of or at least follows that lead. They actually do contribute, they actually build or be part of building the vision rather than waiting for instruction. And that's how you know you have good leaders on your team, you have good team members and at the end of the day we really do work together as a functioning unit. And once you get good at that and you have fluidity, man, the speed at which you can operate and the excellence of the output far outpaces anything you'll get from a top down, instruction driven, project granting, step giving type reality. So the one thing is the team, how do you build a good team? It's giving them the space and the belief that they're the ones who are going to do it. And my job is support them, clear hurdles, provide resources, help with trade offs, but really it's about them.
• 54:38
  Speaker B: I can't disagree with you saying giving everyone a voice so that they're heard. Because this show is called you've been hurt.
• 54:45
  Speaker A: That's, that's, that's fair. I forgot that.
• 54:48
  Speaker B: So we're just gonna end with that. So you know, it's like Eric Gazawi, you've been heard, appreciate it.