← Using AI at Work: AI in the Workplace & Generative AI for Business Leaders
Using AI at Work: AI in the Workplace & Generative AI for Business Leaders
100: Human Plus AI Strategy: Redefining Team Structure in the Age of Automation with Evan J Schwartz
00:57:49
Loading summary
Transcript100 lines
- [00:00]
A
If your goal is to add AI so you can cut headcount for cost, you've already lost and you don't even know it.
- [00:06]
B
What industries do you see? We're really going to notice a bigger impact.
- [00:10]
A
The service oriented industries are going to get hit first. Anywhere there's a service involving a computer, you're going to start seeing the embedding of AI across those services. Anywhere where you've got a set of hands wrenching in and making something happen, that's going to take longer.
- [00:23]
B
Let's talk about that business structure. I'm particularly interested in what that structure of the future looks like.
- [00:28]
A
Almost all corporate structures are going to squeeze down to about three level. I'm just going to have stewards managing orchestrator orchestrators that are managing hundreds of sub agents that are doing the execution plan. The iteration is going to be faster and more efficient and more efficient. This is why I'm urging companies to not sit it out and wait because your competitor who gets this model and gets it working, their ability to iterate will be so fast.
- [00:54]
B
Evan Schwartz is a technology and innovation leader who helps organizations use AI to grow faster, operate smarter and and build a future where people and intelligent systems work together. Welcome to Using AI at Work. I'm your host Chris Daigle. Each week we'll be learning how today's business owners, entrepreneurs and ambitious professionals are getting more done with smart use of tomorrow's tech. Let's get started. Right now, every business leader is asking the same question. What are we going to do about AI? If this is you, chiefaiofficer.com has the answer. We give you a simple path forward where we provide executive and team training so your people know exactly how to safely use generative AI AI in their day to day. We also manage the deployment and implementation to make sure tools actually get adopted and deliver results. And we'll also guide company wide transformation so AI becomes part of your operating system, not just another shiny object. The companies that act now will increase productivity, cut costs and grow faster than their competitors. Those that wait will get left behind. So if you want to make AI work in your business, visit chiefai officer.com and see how we're helping companies of all sizes finally get results from AI.
- [02:05]
A
Hey everybody.
- [02:06]
B
Welcome to another episode of Using AI at Work. I'm Chris Daigle, I'm the host and I am coming to you from Austin, Texas where our organization just sponsored and spoke at a Vistage event for local CEOs here in Austin. And we talk to them all about AI and It's pretty fascinating to see the different pockets of America and where the adoption is happening. As you would imagine, Austin was a pretty tech forward place. With that being said, that's why I'm not in our normal studio. I'm coming at you from a, an Airbnb, but our guest today is Evan Schwartz. And Evan and I had a chance that we've been trying to do this episode for, I don't know, a month or so. And we had a chance to talk, but he was on his way out of the country, heading to Ireland. Galway, Ireland. To facilitate. Yep, to facilitate. Well, I'm going to let Evan talk about it a little bit, but today's conversation is going to cover really a lot of topics that are interesting to me. The impact of AI on education. What are you younger users doing and how are they behaving with these? Because if you're listening to this and you got kids and they haven't really picked up the AI bug yet, this might, there might be some lessons learned from today that will help you communicate with your, your younger ones the importance of starting to leverage AI in their business. And we're just going to talk about AI and AI expertise. So, Evan, before we get started, I want to ask you a question that. And again, I'm borrowing this from Greg Eisenberg with his podcast, but he always asks, what do you want people to walk away from today's episode with?
- [03:40]
A
If. If we can inspire people to realize that we're not victims, we get to choose the future. Life happens because of us, but not to us. I would love that. I preach a person plus AI strategy, and I hear on more than just these two talking heads here today of how doom and gloom and how everyone's going to lose their job and all that stuff. We get to make that decision together. Right? So person plus AI strategy isn't happening to us. And if you're not participating in the outcome, you might be part of the problem. So I'm hoping we can talk about it and I can arm your viewers with what they can do to help drive the outcome of a better world, rather than the apocalyptic world that I think is getting a lot of. Clickbait.
- [04:31]
B
Skynet.
- [04:32]
A
Yeah, yeah, yeah, yeah.
- [04:33]
B
Well, I'll tell you what, Evan, why don't we just start there? So tell me about this thesis of, of people plus AI or so.
- [04:42]
A
Yeah, so it's a person plus AI strategy and it's, it really comes from a lot of background on digitalization adoption. Where do people fit into it? So if you were just to kind of give you a juxtaposition between it. If you take all the tech and AI away and you go way back in time, it would take 200 people to plow an acre of land to grow food. And with the invention of the tractor, you can now do it with two guys. And now with the invention of AI, you could plow multiple acres with a robotic tractor with one guy, right? Yeah, yeah, But I guess the important note is you didn't lose the guy, you just reduced the number of people it took to manage the output of work. That's really the crux of the person plus AI strategy. AI, here's another way to look at it. AI would know everything about your business or everything about business, but know nothing about your business. The way you run. There's a difference between knowledge and context when it comes to AI. So the person plus AI strategy squarely puts stewards in control of an orchestrator agent, sub agents that are running the activities. The execution layer is very different than the strategic, strategic and the tactical layer. Humans still live here in the strategy and in the tactics. And there's a variety of, I guess, arguments and use cases I can give you to back that up. Because that's where the agentic experience or the autonomous enterprise is going. Right? The to replace operations of tomorrow is going to be very different than what we see today.
- [06:14]
B
And, and when you say that it's going to be. It's going to. That's where it's going. You mean that the agentic is going to start getting into the places where it's been exclusively human or what are you suggesting there? Got it. Okay. Will it be explicit and known or will it be. Because if you ask five years ago, three years ago, if you ask somebody, are you using ad? They'd say, oh no, don't touch the stuff. But they use Netflix with a prediction engine, they use Amazon, so they weren't aware of it. Are you suggesting that the synthetic contribution is going to be overt and explicit, or will it be.
- [06:55]
A
It's going to be a mix. I think there's going to be business directed use, explicit implementation, but there's going to be a consumer that is more implicit. For instance, I get up in the morning, I start my day, I go to the bathroom, I get in the shower, I get ready, I eat, I go to my car, my phone goes, you're 30 minutes from work. That's an agent that had watched what I did, pulled in information and realized at this time, every morning, Monday through Friday, I'm most likely going to work. And it popped up a little helpful message telling me that given traffic, you're this far from work, I didn't give it that directive to do that. So AI is going to permeate everything in a way of context awareness. I'm going to start walking into rooms, lights are going to come up, it's going to understand my behavior. Coffee is going to turn on and start making itself. We're going to start seeing it in everything and we're going to just adapt, adopt it naturally over time in our lives and permeate but on top of a very directed measure by early adopters. So it's like anything else. Right. We still, I mean today me and you say we're dialing a phone number. No one's dialed a phone in decades. Right. So the lexicon's still there, it's going to carry forward but it's not going to be adopted the same way at all levels in society.
- [08:10]
B
I would again imagine at some like level this is already happening like you mentioned, I guess with the announcement of the distance to work by Google Maps or whatever. The thing is that it's already happening. What industries do you see where we're really going to notice that? I guess it's going to have a bigger impact.
- [08:31]
A
The service oriented industries are going to get hit first. Anywhere there's a service involving a computer, you're going to start seeing the embedding of AI across those services. Anywhere where you've got a set of hands wrenching in and making something happen, that's going to take longer. Right. So you'll see a lot of your constructions will take a little bit longer to get in. But there may be facets of that business that are automated through the agentic process. The filing, permitting, ordering of supplies all gets automated. But you know the guys with hammers and nails that have to put it all together, that's going to take a lot longer. Yeah, yeah.
- [09:05]
B
So not, not even shifting gears but expanding on that when we spoke the first time for our pre interview as I mentioned, you were heading to go and, and be a, a host or a judge at a, a mentor. A mentor at what was the organization that was hosting.
- [09:24]
A
So it's the IFTP invent for the planet and what it does is it, it goes across and I think the numbers up to 60 plus number of colleges globally and they all get these use cases or challenge statements and they come together for a heads down, you know, shut in workshop and try to use the best technology today to solve those Global problems, right. And each one, it's all streaming live from one university to the next. Each one of the universities have judges, each university has mentors. Jack Saad at ATU out in Galway kind of organized the whole thing. He's brilliant. And we got to see eight teams start Friday night with a problem statement using today's technology, their acuity for the development and solving of problems and coming up with solutions and then present those solutions as practical applications Sunday. So it was a two day shot to go. How close to a reality could you get? And the shock in all was that where it was a service based problem where they could solve it. With today's AI, they went from problem statement to pretty close to a commercially viable product by Sunday afternoon. It was phenomenal to watch it.
- [10:37]
B
So their hackathon vibe, they're building and
- [10:40]
A
not all of them was code. Some of it was designing platforms to, you know, evade floods in certain, you know, different disadvantaged areas of the world that get flooded. What could we do from just simple rising up and elevated farm where they could grow and get it above the. So it wasn't all tech. So they were able to pick their challenge statements and come at it just from an unencumbered creative mind trying to solve the problem.
- [11:07]
B
And the two, you said there were two problems that were the target of this particular competition.
- [11:13]
A
Yeah. So these two particular, where they were able to get all the way almost to what I consider commercially viable was solving world hunger and solving food waste at a global level. And look, you can get down to speed specifics, but globally we waste about 40% of food. Globally we grow more food than we can consume or we do not get it all the way through the supply chain before it gets wasted. And it's embarrassing as a, as a society, as a culture, as species to waste that much food. You know a general statistic that's easy to 140 million eggs United States are produced. 40 million of them never make it to a plate. They go straight to the garbage. That's a lot of food to even have a single person in the United States hungry. Yeah, it's an embarrassment. And the reality of that, where you see the most significant impact in food waste is in the home. It's mine, your house. And if you've got a rotter in your fridge like I do, they call it a crisper. You know what I'm talking about? You'd have a tendency to over buy week to week or whatever your cycle is. And there's food that just doesn't get eaten. That you end up having to waste. There's just tons of waste in it. So the challenge was, how could you. What, what could you do? We didn't even suggest it's a technology problem, what would you do to solve it? And their solution was brilliant. They were able to scan the receipt from the grocery store, just a quick picture of it. AI disassembled that and said, I see all the things you bought based on the sku, looked it up, knew what the product was, put it into a category, guesstimated approximately what its expiration date would be based on its type and brand and all that stuff. And then that would give the system the ability to go, hey, have you eaten your bananas yet? If not, you're about three days away from being able to consume them straight from the banana. Or if they're a couple more days, you could make banana pudding because it's not considered rotten if it's got bruise. Not everyone knows that when you get the bruising on it, banana pudding is a good way to not waste that food, right?
- [13:13]
B
Yeah, yeah.
- [13:14]
A
So they came up with ways to promote that. Give this app away for free for me to take it to maximize my food, make me aware of the food that I'm wasting month to month, that might change, you know, if I'm not going to eat it, why am I buying it? Right? And then if that food item doesn't get bought, that's going to create an upstream ripple to the producers and it produce less and we're going to waste less. Right. All the way through. It was just the whole system that they came up with I thought was brilliant, very commercially viable, capturing all of that user data and then they could take that data and promote it to food producers to say, hey, we want to promote a banana pudding recipe. Do you have products for that? How much would you pay me to put your products, give them a coupon so they can go try your product in the store. So here's a digital coupon, here's a recipe to try your stuff so they can salvage those bananas and consume them rather than let them go to waste or insert any item for any variety of use. It was a brilliant, entirely self contained system with a way to monetize it, commercialize it and create a very positive outcome at the end. I just thought it was brilliant, you
- [14:27]
B
know, and it's interesting because without AI, the friction required to get participation from the user and all that would just be too much. If I could take a picture as I'm leaving the grocery store, bing and Forget about it. And there you go, minders that, you know, items in my cart that I bought a week ago, it's time to eat them. I just, just the awareness alone, I would imagine would have a pretty big
- [14:52]
A
impact on the food waste with recommendations of what to do with it. Because look, I kind of pitched this back to my wife I was so impressed with when I came back and she goes, you know, when I'm shopping I have ideas in my head of why I bought that. But when I'm throwing it out going, oh, I meant to, we were going to have this and I just forgot.
- [15:09]
B
Right, yeah, sure. So this k. This opportunity came out of your work at the college level, is that correct?
- [15:18]
A
So I, because I, I'm also an adjunct professor at Jackson University. I think I got packed, but it came out of AMCS Group. So my role as Chief Innovation officer, we sponsor these. So we sponsor ju. We sponsor ATU because look, we're, it's, we're a technology company so it's within our best interest to be looking for the best and brightest. So I'm telling you, these guys are probably going to get a phone call. But the reality, because that gives us an opportunity to kind of look and see where are things heading, where are they? Right? And we're also seeing the transformation internal because our software, we sit at a global level. We sell software into resource intensive industries to SaaS. And we're seeing the challenge of these resource intensive, salt of the earth style industries, I call them wealth industries, where they're taking raw materials, processing them, getting them into new materials, producing a product and getting it back into the world and then having the reverse logistics problem of okay, I sent this thing out, I now got to get it back, break it out into its parts, sort it again to get it back into the supply chain to the point where it makes money. So we service the waste, the recycle the resource intensive industries like paper businesses, scrap metal businesses, those are very difficult industries because we spend thousands of years getting good at digging something out of a ground or harvesting something out of the ground because that farm or that mine doesn't move. But once I'm a products company and I'm sending it out to the world and it goes to a million stores and logistics has distributed it, getting that reverse logistics problem back in, that's a real problem, and then breaking it into its constituent parts. But if I don't have some kind of efficient system, I've killed any and all margin. It's cheaper to go back and just get Fresh out of the ground and we've got to get better at reusing these materials.
- [17:00]
B
Let me ask, so what are the things I've got kids, college age, all the way down to just getting involved in school. And you know, I did a post about this on LinkedIn the other day about what does the AI relationship need to look like with between your children and the technology. I don't have a lot of answers for it. And we, as we mentioned, we want to start with an abundance frame on this. But when businesses ask me regularly, this is education and business, they ask me like, what does the future look like? Because, you know, my daughters that are in high school are, they're not encouraged to use the tools. And to me I'm like, well, that's, that's like telling a kid you can't use email because it's cheating. You should learn penmanship, right? Well, maybe. I mean, I think it's a disservice and the same thing in, in business when we have companies or, you know, industries even that are playing the wait and see game when it comes to AI. Like, oh, well, we'll let them figure it all out and then we'll get involved. Believe it or not, if you're listening to this, there are plenty of your competitors who are on the sidelines still in 2026. But let's start with, because I want your perspective on both. But let's start with what are you seeing at the college level when it comes to interest, ability and perspective on generative AI for their careers, like their budding careers.
- [18:30]
A
So I just want to touch on first, the university's perception of AI as a cheating tool was a knee jerk reaction very early. Ju because they're a privately funded college was able to wrap their heads around that early. That first year it came out, I think everyone had that knee jerk, oh, this is a cheat. And then they realized, oh, this is another tool. We need to now incorporate this into our curriculum, which they have ATU Galloway doing the same thing. So while maybe your child's university might be still thinking that way, we're seeing those ripples now go out that the colleges are now adding AI as part of their curriculum. They're going to have to. So it's not alone.
- [19:13]
B
Excuse me, how are the kids adopting? Do they see the importance of it? Are they looking at it as a shortcut? Or are they like, oh, I need to learn this new mechanism that will increase my economic viability. How are they?
- [19:26]
A
I think that's an unfair ask of young minds who are trying to get in the world. They're going to college to understand, to form that framing. Colleges need to take ownership at the higher education to put that framing in place. Get rid of your victimness, get rid of the slight that it's a cheat and teach them how to properly use it. So what we're teaching at JU is the stewardship model. You're not going to go into the workforce worrying about how to longhand calculate. I think Elon Musk said this best. You could outperform with a single financial analyst and an Excel worksheet and entire skyscrapers of guys doing it by hand. It's the same thing. But there's no AI. It's just an efficient technology and nobody put a bullet in the head of Excel. I remember the day of my teacher going, no, you need to learn how to do this. You're not going to have a calculator everywhere you go. Yeah, well, we showed them, didn't we? Yeah, we got that thing everywhere we go now. I get almost panic when I leave my house.
- [20:22]
B
Yeah, yeah, right.
- [20:24]
A
So. But it's the same principle again. Here's a new tool if to frame it differently. I read an article and this really frames it well is it was a artistic photographer, the kind that just doesn't take pictures for weddings. He sets things up in a museum. You may have seen it. And he took a selfie of himself and he described this portrait of a stark black and white water dripping off smoke in the back. Just this compelling image. And AI came back and produced it and his jaw dropped his stomach. Val. It was compelling. It was gorgeous. Off of what was a selfie of him in a T shirt.
- [20:57]
B
Yeah.
- [20:57]
A
And he went and had a walk and he's like I need to find something else to do in this world. After about an hour he realized, you know what? I still had to tell it. The lens, the lighting, the smoke effects. I had to describe that lighting impact on the image. I had to describe the color of the car. I did use all the same knowledge. I would. What I didn't have to do, I didn't have to set up a stage. I didn't have to create stage lights. I didn't have to have a guy with filters in front of it. I didn't have to have a smoke producer guy and I didn't have to take a thousand pictures hoping I get the right one. What we've done is we've compressed the iteration cycle but the knowledge is still there. We just conducted an experiment at JU for being able to Teach architecture. And we had juniors, mids and seniors in there to go. Do I need you to pound out 135,000 lines of code to be a good architect, or can you do it with today's modern tools, Claude code and vibe coding? And you can still get there. And it was just an interesting dichotomy between the juniors who didn't get the abstracted concepts of solid architectural principles, the seniors who didn't like the way it was done and still didn't get results out of it. But our middle guys appreciated the black box, knew enough to test the output, knew it was their input that was the problem. And they're the only ones in the class that produced a complete product, end to end, ready to go. So there's just a certain amount of what we're going to be teaching to the next generation isn't the hand doing long division. It's the fundamentals. What is the fundamentals of photography? What is the fundamentals of architecture? What is the fundamentals of, you know, engineering and design? I'm going to have this tool that's going to do all that labor, but for me to know the fundamentals, that allows me to drive tactical execution. So I'm probably going to come into the workforce a steward. I'll have five or six orchestrator agents I'm responsible for. When me and you go into our job today and the world around it, there's a construction site, there's something happened down the street. Context comes at us through our senses naturally. We adapt naturally. And business has relied on the fact that humans can do that. You can have one conversation with your manager and you could pivot on a dime and start running a different thing. But at the context you gave an agent, and if you put it into production and no one's managing it, it's going to run that playbook over and over again, even though the world around it is changing. And it's going to start to have what's called entropy. You're going to get leaching of leakage, and then you're going to find out it's not doing it, it's not scaling well, it's not keeping up with it, because you took the human component out of it. That is meant to look at the metrics. Look, observability. Why is it going off? Oh, that's right. Because it doesn't know that there's a traffic jam or construction here. We had to reroute. Let me give it more context that we're rerouting. Please recalculate and do it. And it brings the agent back into place. The world is going to, you're going to see this. This. Now this is Evan's view. Right, But Evan's view is you're going to have board level SLT stewards. Almost all corporate structures are going to squeeze down to about three levels. You're not going to have the mass sub hierarchy because it's too hard to manage more than seven people. And when you do that, you have to spin off a manager and a group and a department to keep it functioning. I'm just going to have stewards managing orchestrators that are managing hundreds of sub agents that are doing the execution plan. We're watching that execution through steward, keeping everybody in line. And we're reporting that up to make sure that those activities are fulfilling the strategic objectives for slt. And SLT is now driving strategy. The iteration is going to be faster and more efficient. And more efficient. So you see something going wrong, there's a competitive advantage. This is why I'm urging companies to not sit it out and wait. Because you're a competitor who gets this model and gets it working. Their ability to iterate will be so fast. Yeah, imagine that photographer where he's using AI to produce art. And a guy still setting up his stage. That's an all day event, waiting for the pictures that yeah, to get one shot that he has to go into a black room. And this guy's like, I can get you five of those in about five minutes. What else do you need? So my ability to produce at the same level is phenomenal. And it is exactly. What Elon was talking about is that I used to have entire towers full of accountants with those cranking machines working it and now I can get one guy with an Excel worksheet to outperform that entire tower. It's the same thing.
- [25:24]
B
Let's talk about that business structure because I'm, I'm particularly interested in what that structure of the future looks like. Not necessarily at the high level. I think they're pretty secure. But it's the army, it's the bottom of the pyramid and the OR chart that, that is at risk. Let's say. How far off are we from what you're suggesting as that that corporate org chart. And I'll tell you why I'm asking this is because I've referenced it before on the podcast, but there was a study done by mit. I think the, the results came out in December and they, it was partnered with the Bureau of Labor Statistics and they were looking at all of the jobs, or I think it was 900 jobs. And they were looking at how much can today's current technology impact the different roles and that sort of thing. And they said universally about 12 to 14%. But anybody listening to this probably is looking like we're not getting 12 to like 12 to 14% of most of my employees. Work is not being handled by AI. Maybe ad hoc here or there, but it's not. It's the friction of the human. Just because the technology exists doesn't necessarily mean that the human knows how to use it, wants to use it, or will use it. Right, so how far off are we from this dynamic that you were describing of the three tiers of board? Slt. And SLT stands for senior leadership, sorry, Senior Leadership team. And then the steward, which would be the agent orchestrator.
- [26:59]
A
That's right. It's the human that's making sure that that strategy is converted into tactical execution commands, workflow, stack. Right. So imagine what your SLT typically today has a plus one. So here's our strategy. We got to increase ebitda, we got to get free cash flow, we got to do this. What are all the ways we could do that? That gets put into plus one. Plus one builds a balanced scorecard. I'm going to have to touch every customer a little bit more to reduce my churn. Right? So. And I know because I've run these numbers every time I call and I talk to my customer and I feel like they feel like I love them, they're less likely to leave me. But right now, because I got humans making all these phone calls and doing all this stuff, I can only touch a customer so many times a week. But part of that touch is what I call low value necessity activities. I make a call to you, you're my customer, we have a talk, there's action items. Right after that call, what do I do? I have to take that information, write an email, get it to you. Here's what we said, here's what I agreed to do, I'll follow up next week with it, blah, blah, blah. That low value but necessary work reduces the time I can call the next customer. So if I've got AI listening to this, transcribing it, reading it, and I've spent any time training my helper agent, my co worker, to be able to take that transcription, put it in the format that I like, and send it to you so you can see it. And then go ahead and put updates in my calendar for me. Let's do a check in because one of those action items, I got to go back to product and go have a talk. Put that on my calendar for me. Do all of that so that I can spend more time doing this. So I can cut my churn rate in half that fee. That's a tactical balance scorecard activity. You can measure that when you do it reduces churn. So you can put that simple math. What is the total ARR of your company? What's your expected budgeted churn? If you're at 5%, you cut it in half to an F. That could be millions of dollars. Now it comes down to how many of those stewards do you need running those agents and low value to be able to double the number of touches. That's the kind of activity there is a process. So it's not just from where we are today to autonomous. Okay. There is a journey of OCM organizational change management that you're going to have to go through. Some things might be I offer an agentic way, but I still have the old way to do it because I might not. I need a way back to give my company time to adopt the change. Because if you don't know your company's culture, are you an innovator, are you an early adopter, Are you pragmatic, Are you risk averse? If you run at an innovation deployment strategy and you're risk averse, it is high friction everywhere. Right? That's gonna. So that's where you have the weird disparity around timeframes because companies have different cultures, companies run different. So you'll see some companies running slower than others, some domains run a little bit differently because they all have like a collective culture. So it's going to, there's going to be a little bit of peanut butter spreading, but most of it's going to be a rocky road depending on the nature of the company.
- [30:01]
B
So in this three tier framework that you're talking about, who is touching the
- [30:07]
A
client, that's more likely your customer, right? I mean you're steward. No, you're steward.
- [30:13]
B
Okay.
- [30:13]
A
Yeah, yeah. So there's going to be human touch point, but I don't necessarily need. If, if it's an electronic transaction like we were just talking about, I can have my agent do that. As long as it looks like I wrote it, right. I would train it. Here's how I write it. These are the word usage, this is how I talk, here's how I organize my stuff, keep it consistent, send that email for me. But if there's a problem, the whole point is I need to connect with my other humans out there in the world. That's a customer. But if I'm doing an order from a supplier and their SAP back end talks to my order producer and I'm not talking to a human, then I can have my agent talk to their agent and make sure that my inventory runs low. So it's not an all or nothing, one size fits all. There's going to be a mix of this. You will put humans where humans make sense. But as far as the agents go, none of them are fire and forget. You will have a steward above those. That steward might be able to control eight or nine orchestrators. So what used to be an entire department of financial analysts might be one guy running multiple stewards and then that one guy does the human touch. But you got agents doing the execution and producing the output.
- [31:20]
B
Interesting. So and that, and I'm going to dig in on this because I really like the future of what an organization looks like is unknown, but certainly not going to be the same as it looks with the traditional. We've all seen the org chart. It looks like a pyramid. Right. That's going to look a lot different. So this steward role is going to be somebody. They're going to have subject matter expertise on the domain.
- [31:47]
A
They're going to be domain experts over their orchestrator agents. You're absolutely right. They will also be good at understanding measurement metrics. I'm going to have a dashboard. I'm going to be watching and stewarding my orchestrator agents, interrogating them. Why did you do that? What was the problem? I need to improve the context so they're going to be technical enough to know how to keep the agents on rails. The tech works if it has the right context. Everyone keeps talking about it fails at scale. It fails at scale more often that when you put it into scale. There's edge cases that it doesn't understand how to use. That's the person plus AI that if you just fire this thing off and say I'm going to give it, I'm going to unleash it on 2 million customers. And I don't think there's going to be a variance between those 2 million customers. I'm being foolish at that point.
- [32:32]
B
So what are the characteristics of an ideal steward in this lexicon? Obviously domain experience or understanding like oh, that number should not be that high. There must be something wrong with the agent or with the input or whatever. But they're also going to be interacting not just with the humans that are consuming the Deliverable of those agents necessarily on your client side, but also internally interpreting the strategic demand from the board level, the strategic plus tactical from the senior level. What would I study today to be good at that in a year or two?
- [33:17]
A
Yeah. So there's fundamentals I would say that would need to fit and this is being explored. Right. So the domain knowledge is going to come with its own bag of tricks. Right. If I'm in my customer success mode and I own my customer success orchestrators, I need to understand how to do problem solving with my customers. I need to know how to engage and be a good communicator. There's going to be a different profile for that type of steward than someone who's in finance. I need to be able to understand net present value. I need to be able to look at my metrics, I need to be look at my inbound. I should be able to see on my dashboard if we're off or we're trending off and then I need to be able to dig in and find out why we're trending off of whatever the metrics that designate success. It is just someone who is well organized, able to monitor and react to inbound stimulus and be able to then give appropriate clear communication, that's context into your agentic orchestration layer so that they can now drive the sub agents which are just very narrow action. There's not everyone that's going off of a mega agent has gone off. They're chasing a rainbow they'll never get.
- [34:30]
B
Yeah, yeah.
- [34:30]
A
So keep it narrow, keep it short. This agent does this and does it very well.
- [34:34]
B
Okay, so let's talk about this. Normally when I, when you look at an org chart, you assume that the person, the people that are higher up in that org chart towards the top of the page, they've got more experience or they've got like a natural ability at that, whatever that role is and those that fall under them aren't as motivated or they haven't had enough, enough experience to ascend the corporate ladder or whatever. Right. But if we don't have to have that army of, of. I mean everybody wants all A players. Yeah, true players are hard to find and they're expensive and they have choices. So if their environment isn't filling what they're looking for, they can go somewhere else. So in this environment I don't have to worry about losing if my, if my agents are performing at the A player level, I don't have to necessarily worry about them getting poached.
- [35:21]
A
Right.
- [35:21]
B
So that goes out the window, however, is that an entry level. Not saying, not going to say entry level. But what I almost want, what would be today the equivalent of like a department head, to be at the steward level.
- [35:34]
A
Yeah. Because anywhere you have humans, you're still going to want a training and the ability to grow that human into the role that you need them. So it could be. I have a very large financial services and consulting division and I've got my big wealthy customers and I've got my smaller customers. Maybe my entry level stewards start here in my smaller customers and they learn how to play and then grow. So you will see. But it's going to be lateral and very limited up. It's not going to be anywhere to the end tiers where I've got departments on the departments under teams. You're just not going to see that level of hierarchy today. Extremely flat.
- [36:16]
B
So this kind of fits into a question that I've been getting more recently than I ever had before. And that is, hey, in our industry, if we don't have, because you know Dario Amade from Anthropic, I think in May of last year, and he repeated it again earlier this year, was suggesting that 50% of white collar entry level jobs, they're not going to exist. They won't be needed because they fall under that, that steward below the steward
- [36:40]
A
that you were just talking about.
- [36:41]
B
Right.
- [36:41]
A
Like, yeah.
- [36:42]
B
So when I say that to some certain professionally or especially professional services, they say, well, those people from tomorrow are going to be the ones that after 15 years have my job. How do we create that pipeline? And they may be looking at it the wrong way. They may be, but it's a question that I have not been able to answer. But how do we. Because I started at the bottom and now here I am at the, you
- [37:07]
A
know, because you had to learn the tactical skills, right? You don't need to learn that anymore. The agents are handling the tactical skills execution. What you do need to learn is the fundamentals of that domain. You need to know what good looks like and you need to have great communication and orchestration skills. You got to be able to look at a dashboard, see that it's going off rails. Why is it going off rails? That takes almost a triage. Troubleshooting. Yeah. And then you need to know, okay, once I've identified the problem, what's the best course of action to correct? And then let me, let me drive that down into my orchestration. It is new muscles. Right. It's not the old muscles. So you're. But because there's domain expertise. You, you should pick amongst your existing workforce and then start training them on orchestration. How do I do this? So your first question is, what do I tell my kids to start playing with? Now there's Nemobot or Nemo Claw just got released. Clawbot's out there. You can get Docker onto your workstation, start to just automate the simple things of your life. If you have social media, how would I put together a few agents on my desktop to be able to go out, get my feed, allow me to type, and then pull replies back in there? How can I manage my email? Maybe I've got Gmail and a handful of free emails. I want to pull those in and automate parts of my life. You know, my youngest son has a series of agents. He's hungry, he talks to his agent, it goes out, gets all the stuff done, and food shows up by the door so he doesn't have to worry about is it Uber or is it this service? It goes and works out what the best service is and food magically shows up, money shows out of his account, and he's automated that whole thing, you know, I mean, and the weird thing is it's like he has no desire to drive. That's a weird side effect. But I've also learned that if those, those young individuals who don't have that strong desire to get in a car and go somewhere and drive and do some of those, or have learned really good at how to push a button and get just about, about anything delivered to the door. Which tells me they're pretty good at managing agents already. They just need to find a good domain that interests them and then sell and upsell those services into the next generation of corporate world.
- [39:22]
B
That was probably more helpful to me than you may realize because I don't know how to answer that question when it's like, how like law firm. Yeah, we've got somebody who's, you know, spent their 10,000 hours in the courtroom room. How do we train somebody to do that? Well, they have to get their 10,000 hours in the courtroom.
- [39:41]
A
Right.
- [39:41]
B
So unless the whole paradigm shifts and your son that you're talking about here, the desire not to drive. Well, if you wanted to drive, there's full self driving. Yeah, you don't even have to know how to drive. So this, this idea of like the journey that I went through, I got my permit at 13 when you know, like, so they don't have to do that because the technology. So this kind of equates back to that Role of well, how do I get these junior level people to become senior level people in 10 or 15 years? Well, maybe they don't. Maybe the whole model has changed to the point to where they don't need to be senior level. They've got access to all the domain expertise through a model or through an agent and they just need to know what food they want, what outcome they want from the, the legal and how
- [40:33]
A
to tell this thing to go do it. Yes, they don't need to know that there's an Uber Foods or maybe your Publix has a delivery food so you don't need to. That is all abstracted. The AI is going to find the cheapest one to achieve the results you're looking for. You told it what you want to eat, it'll go across the delivery services, it'll request the amount of money or if you're okay with it, it put in the orders, get it done and your next interaction is in fact he doesn't even interact with the guy at the door. He has him knock, put it on the ground and walk away. It's like he's worked really hard not to interface with that guy. So but I mean to me that is you're going to see more and more of that now. Whether that ever has a challenge to elevate up into senior leadership. I think you start seeing two casts of employees. You see those tactical strong executors and then you see those visionary entrepreneurial leadership business growth. Because the goal for this isn't to cut costs. And I and this is, this is really the big preach here and we talked about at the beginning, what do I want to leave them with? If your goal is to add AI so you can cut headcount for cost, you've already lost and you don't even know it. Keep as many people as you can and go for asymmetric growth. Go back to the customer success model we talked about out I've got agents. I mean I'm not looking to cut any of my customer success. I want to cut my churn in half. I want to double the amount of NRR net revenue retention I can keep every year. Now at the end of that, how far can I grow that? How many more customers can I put into that bag and have those stewards continue to service it before I have to grow? And now at the end of the day, if I have totally dominated my TAM and I can't grow any further, by all means, Mr. And Mrs. Custer, right, size your business, but don't start There start with asymmetric growth. How do I get maximum free cash flow? How do I get maximum net revenue retention? How do I reduce my churn with the people I've got? It is a new muscle. It's not a job description you can put out. No one knows how to be as well, except for my son. But no one else knows how to be a steward. Right now there may be a few of these guys have come in and cracked the code. I see it all over the social media. I've got agents to doing everything for me. There's a few early adopters, but the general population is going to need to be taught.
- [42:53]
B
Okay, so this is interesting because I've always, when companies ask about that, I say, whoa, don't fire everybody. Because they have, they understand how your organization works.
- [43:02]
A
Right?
- [43:02]
B
And if you get rid of them, even if you want to hire somebody that knows AI, they don't understand your company culture. That Diane has the keys for this. They don't understand all those things.
- [43:13]
A
So the context that you need to steward with that AI, because it's going to change, entropy is going to kick in. You kill all of your tribal lore before you. That's the point. AI will never. I'm not going to say never. That's probably inappropriate, but probably in our lifetimes, it will never have the ability to absorb context the way we do. I go into a business, there is no communication from my boss's boss's boss's boss that says, hey, the water cooler is not working today. You need to go down. That's organic context that we as humans, just through interaction get and the business takes on. But an AI is going to keep doing the same thing because it never gets that it needs its steward, its humans care. We're the bridge to reality for it to stay efficient, to stay on track, to stay doing the things that worked really good in the lab but isn't working so great out in the world. That's what the steward is really doing. And now that you realize that my steward can achieve 10 times the work and I just spent a bunch of money training my entire department to be stewards. I'm not exiting those guys. I'm going to go grow my business 10x because I just spent a lot of money training this entire department how to be stewards. It's a new muscle. There's nowhere in the market to get it today. No one's putting that skill on their resume, probably not for three to four years. So you better keep your resources, learn how to invest in Them, they've got your domain knowledge, they know your culture, and go for asymmetric growth. Pick that once you're done growing. I don't know any business that's ever said I've grown big enough. I don't want to grow. But maybe you are. Then by all means. Right. Size your business.
- [44:52]
B
Interesting. Okay, this is my head swimming a little bit because this is an approach that I hadn't considered yet because it's the whole question about what happens when AI starts taking the jobs and are people, you know, is it UBI or as Elon calls it, universal high income? Which sounds intriguing, but. And I don't have any answer, I try specifically not to think about what does that look like? Because I'm focused on how do I get the most out of it today. I'm not here to necessarily predict the future. I want to predict tomorrow. Not.
- [45:30]
A
Look, my goal is to provide a stewardship model, an operations model, a security model, and a governance for AI for your business to operate. But it does involve holding people until you've gotten all of your gains and then you. Right size, that's really the trick. It's not much harder than that. Right. So if your private equity, your board is. You need to, you know, cut 50% of your workforce. Calm down. Why don't I double my revenues first and then we can decide whether I should cut my work workforce. Yeah, let's target the up. That's an infinite game, as opposed to the best I could get cut in my workforce is zero. And I still now have the cost of AI, so I didn't get it to zero. And then the second entropy sets in. The context of the world of reality shifts one way or another. We go off the edge, we list the ship horribly. And now you've got nobody driving these agents.
- [46:26]
B
You know, it's interesting because yesterday, as I mentioned at the beginning of this episode, we were sponsoring and speaking at a. A vistage event. And our speaker, Adam, is clever in general, but his approach to AI is unique. And in that, in that example, in the demo, the presentation yesterday, he did something I wasn't expecting. He said he was trying to explain to people that artificial intelligence truly by definition doesn't exist. This is predictive engines and blah, blah, blah.
- [46:57]
A
Sure.
- [46:57]
B
And people were like, oh, you know. And he said, well, let me prove it to you. I want you to take out your phone and open up your chat GPT app and ask it to give you a random number between 1 and 25. And I was like, oh, my gosh, where's this going? Right, Because I mean like a random number. I didn't know where he was going. Come to find out, and I want anybody listening, do this. I'm not going to tell you what the number is until the end of the episode. Sure enough, about half the room, when he asked, was your number blank? About half the room raised their hands. And this is a room of 80 people. Right. So that means, you know, 50% got the same quote, unquote, random number.
- [47:38]
A
Now but why is that? It does that because it's a predictive engine.
- [47:43]
B
Yes.
- [47:44]
A
Based on models of statistics, you could have rephrased that question to say, write a piece of code that generates a random number, execute it and give me that. And even then the purists will tell you a computer can't really generate random number. It creates a.
- [48:00]
B
And that's what I'm going with with this. If we get rid of the people, then there will not be the opportunity to introduce true creativity. Whatever. Whatever the equivalent is of randomness. Maybe it's creativity.
- [48:13]
A
Yeah, it's called stochasticness. It's cellular automata. It's formed from chaos. Right. It's creativity. It's. It's je ne sais quoi.
- [48:23]
B
Yeah, but the model doesn't, the model will predict randomness.
- [48:29]
A
It'll give you the probability if you were. Because look, I, I'm with your guy because I actually teach on how to build these deep neural networks.
- [48:36]
B
Yeah.
- [48:36]
A
At its simplest, it's a normalized curve, plus or minus Z formations with a linear model going through it. And you're predicting outside the model. If I have, have this many on X, I go up to Y, here's a likely outcome.
- [48:49]
B
Yeah.
- [48:49]
A
Now it's that. It's just that on steroids, when they talk about, you know, dimensions and billions of parameters, that's a, that's a two parameter model, right? Yeah. How much money, how many tables I have in my restaurant at this point, I've tracked five. I get this much money. So can I produce that? At this number of tables, I'd make this much money? That's a predictive model. We do it in finance all the time. Take that to the end, the of degree and that's where you're getting AI. So asking it to understand chaos and make form out of it is uniquely a human feature. It's called secular automata for a reason. It'll do predictive probability. The normalization curve is our best predictive model for chaos systems. Just about. I mean, I'm sure you've seen the ball bearings where you flip them in it. It creates that little hill no matter how many times you do it. That's a normalization curve. We. It's just a statistical anomaly that is like a property of our reality.
- [49:42]
B
So that reinforces on why I would want the human involved in my business. Not just because of morals or ethics or, you know, so green as people or whatever, but because if I eliminate the individuals, I can be tricked into thinking that my business is adapting to what's happening when the reality is it's not. It's not.
- [50:07]
A
Yeah. Take that exact same example. If we want to beat this dead horse, I got customer success. I'm trying to reduce Churn, and I've actually put a call service agent out there to cut half of my inbound calls that statistics measured. I've cut half of my inbound calls. My costs go down green, green, green, green, green. And then I come to find out when I get my quarterly reports, I've lost half my customers. My inbound call didn't get dropped because of the agent. It got dropped because no one wanted to talk to an agent. And so I lost half of my customers to Churn because they didn't like it. And I wasn't measuring that. So you just got to be very careful.
- [50:41]
B
Yeah, I didn't know I was going to go in this direction. And I never do on these, like, the conversations. I know that I'm bringing on people who are thinking, dreaming, sleeping, eating AI. And I'm always curious, like, what are you. Like, how are you thinking? That's kind of the main purpose of this podcast, is to peek into the minds of other people who are thinking about AI as much as I am. Am right. And so I never know where it's going to go. But this is not the direction that I wanted. But, sorry, no, no, it's perfect because this is something that. This is a. This, this organization of the future or of today or whatever the timeline looks like with AI is a question that's relevant to every organization. Yeah, everybody listening to this. Because if you're listening to this, you're either at the top, you're somewhere at. Either at the top or at the bottom of that. That, that pyramid of the org chart. So this, this understanding of this leaner, certainly not. Not as, not as vertical as it was. Right? Like, yeah, only a few layers. Because I want the answer. I want to start building my company that way. I want. And I want to advise other companies on how they should be building it. But if I haven't like if I don't have the answer, we're kind of like, well, let's try this. Right? I mean, we could can or we can, we can adapt the technology to fit into the current paradigm. But is that's not really necessarily preparing somebody for what it looks like in five years. It's keeping them participating at a, at a competitive environment. But they're not necessarily, hey guys, here's the real way to go. They're not going that direction. So. And I don't have that answer. And so I'm always looking for it. And I didn't know that that's where we were going to go with it today. But I've gotten some definite takeaways or new thought considerations on what does this mean for the organization of the future. So thank you for that. And you kept mentioning ju, but that's Jacksonville University.
- [52:40]
A
Jackson University, sorry. Yeah.
- [52:42]
B
Jacksonville, Florida.
- [52:43]
A
Jacksonville, Florida.
- [52:44]
B
Okay. And that's where you're currently the adjunct teaching.
- [52:49]
A
So I'm an adjunct professor. I teach technical project management because these things need to be adopted. Architecture, we're looking at how to still get architects when there's no coders, because that's our current escalation path. Right. So you got to figure out to your point, how do I get these guys if I've taken this, if I've clipped them at the knees, I got to come up with another progression path. And then of course, my role at AMCS group is just a pleasure. Every day we're inventing ways to save the planet, come up with more efficiency. One of our AI models saves 17 gallons a year off of a waste truck that goes around and picks it up. You got 14,000 trucks. That's a lot of diesel not burning. That's bottom line impact. And all that was, was just rethinking about how you do the pickup.
- [53:32]
B
Thank you for that from the rest of the world. Thank you for.
- [53:36]
A
And that's just one of those efficiencies, the whole point of an ERP. I've been doing ERPs since the late 80s, early 90s. And it's just, it's an efficiency engine. It's. It's meant to make businesses more efficient. And so AI is just another rung in the long rung of tools to do this. It's just going to look a little bit different. The OCM component of AI is very different than the classic ERP where I still have humans being very expensive. Modems, modulate, demodulate. I'm taking data from one thing. I'm hand Jamming it into a screen and it's transferred. So that's a very expensive modem. And those are going to go away as we plug in IoT and we give it some context. But it's, it can AI will get confused by the context of inbound. There's a difference between data and context. Let me just be very clear. Sure, you can get all kinds of data in, but if you don't have someone, a human being providing that context, you know, what does that mean? Randomly give it context. I assure you it will do it.
- [54:35]
B
Oh, and before we close, I want to give everybody that number. If you've run that experiment, give me a chat GPT, particularly give me a random number between 1 and 25. If your number was 17, then you are, you're like half of the room from yesterday. That random number was the same for everybody. It was, it was pretty incredible. So anyway, Evan, thank you so much. This is fascinating and I really appreciate the perspective because, because it gives me another model to consider when answering that question by organizations like how do we prepare for tomorrow? And if you're a listener and you're not ready for that level of like what do we do with our organization? I would suggest that keep this one in mind because it seems extremely viable to me. So again, Evan, thank you so much. Any closing remarks for everybody?
- [55:25]
A
Look, I love to do this. My goal is to make sure that businesses, I don't care what your business model is, understands and still going to be here in the future. We do not have to pick this apocalyptic view that everyone is painting. It's a choice. Life is happening because of us, not to us. It's a choice. So if you, if you're looking for details, I do speaking engagements on this all the time. I'm happy to just give you my talking deck. So if you want to just look at it. I don't charge anything for this. You can go to my website, evanjsports.com I'd happy to give it. If you're in one of the resource intensive industries and you're looking for something, give me a call and I'll get you hooked up with someone at amcs. So other than that, sounds great. Stick with the person plus AI because at the end of the day, I don't care how abundantly productive you are. There's no one to buy your product. What are we doing this for?
- [56:13]
B
Yeah. Agreed. And everybody listening. I'm going to have Evan's contact information, at least his LinkedIn and everything available in the show. Notes and like he said. If if your industry is resource intensive, the stuff that they're doing at AMCS Group is. I don't hear a lot of people talking and thinking about it that way. So I'd encourage you to follow what they're up to and even reach out to Evan to get some perspective on this.
- [56:39]
A
We're sustainable and profitable.
- [56:41]
B
Love it.
- [56:42]
A
You don't have to be one or the other.
- [56:44]
B
And thank you so much for listening everyone. Again, if you have anybody that you know is on the AI journey and they're looking for broader perspectives, please share this with them. And if you have a moment to take, take some time and give us a rating on whatever your podcast platform is of choice. That really matters a lot. It gives me feedback and it also helps others find what we're up to here. And my advice is always go out there, start using AI. So thanks everybody. See you soon. Thanks for tuning in to Using AI at Work. Don't forget to subscribe for more conversations about how to use AI at work and a special thank you to our sponsor, Chief AI Officer for Empowering Businesses with AI Education and Training. Visit their website for a free AI Readiness Assessment and AI Strategy Guide to help you get started using AI at Work. That's www.chiefai officer.com. follow us on Twitter at the handle Using AI at Work and visit www.usingaiatwork.com for free resources to help you harness AI in your role.