I have an AI workforce with 34 AI agents that are working around the clock for me. 34 agents. And I'm now saying, actually that was two chapters. There are AI super users, they have higher risk tolerance, and they are going to sprint past you. The Majority of CEOs I've talked to, they might actually show up pretty well on these shareholder calls, but that's a speech that was written by someone else.The pace of change is going to be faster than I think it is. It is always faster. Should this be forced on workers to incentivize the heck out of your workforce and inspire them and motivate them and make sure that they're seeing positive examples so that they want to do this.

Howhas this changed corporate America? Oh,

so many things. One that is starting to happen right now.

All right. Welcome to a new episode of Power Players. Uh, I'm really psyched for my next guest here. We're gonna talk about all things AI and how I'm probably using it totally, absolutely, completely wrong. Uh, Ali K. Miller is here, Open Machine CEO. Good to see you.

Good to see you. I, that is the purpose of this podcast is actually to make fun of you.

Well, also too, like, like you are real. Like there's not your AI. You're real. I'm real. Like

thereare a lot of social comments asking me if I'm AI. This is real.

So, so for those not familiar with the company, talk to me about Open Machine.

I work with a lot of Fortune 500 companies, private financial institutions, AI labs, AI startups, and just help everyone transform into the business that they're meant to be leading or career that they're supposed to be leading in the AI age. So it's a lot of, there's some education, there's a lot of advising, uh, and it's a lot of conversations behind closed doors that I try and bring the learnings onto chairs like these.

What did you, what got you to Open Machine? Where did you start your career?

Oh, I started an AI almost 20 years ago and was doing, yeah, just more acronyms. Uh, so I started doing uh ML research in college and I was doing stuff in natural language processing.And, um, about 10, more than 10 years ago, decided to dedicate like the rest of my life to AI and every single day for the last decade plus I've been working in it. So I was at IBM, launched the first multimodal AI team there was at AWS, was the head of uh AI for startups and venture capital, built that into a multi-billion dollar business. And then summer 2022, I was like, the next model is gonna blow our minds, right? Because GBT.3 had come out end of 2020. And it was just so obvious what the scaling law was looking like. It was so obvious to people who had been working in this space every day.And so I quit my job to build a startup, and then the next thing that came out was Chachi PT and not a model, and I literally burned our business model, like took the business plan, lit it on fire on a stove, moved to New York 5 days later because I was on a road trip, and from that point on, I don't think I've slept.

Like I know it goes like that's, I feel you. I feel you. So I feel as though a lot of people that are watching this or listening to this, theyStill not very familiar where like where did AI even come from? Someone who's been doing this for decades, like decades, or like 20 years 20 years, um, how has it changed over that time period? So

we've had the term AI for 70 years and for the first many decades it was, uh, a lot of rules-based systems hoping to make computers look and behave like humans would, but a lot of it was, you take this in and every time you see this, please give me that. And it wasAgain, meant to appear as if it were doing human tasks. Um, starting really around like I would say 2012, we started to have some big shifts, uh, around very large data sets where, um, we started to have a lot more learning around deep learning. And so you could take these really, really large data sets and start to extract.Some interesting patterns, which then when faced with brand new data, you went, wait a second, I've seen stuff like that before. Let me jump in and help. And so maybe it was in spaces like image understanding or maybe reading comprehension. And really in the last couple years we've seen an explosion in image generation, video generation, video editing, uh, coding.Certain. Um, and so that that generative AI has been around for decades, but like good generative AI because of the scale, because of algorithms, because of how much data we have and cleaning process, all that stuff, really only in the last couple years, I would say since, you know, GBT 2, have we been able to see this like really big hockey stick moment. And then basically we've hadHockey stick moments every six months since, like chat to BT. I think so many people thought of chatty tea is like, oh, that was the thing. And now it is mid 2026 and now it's changing again. No, we had a really big change end of 2024 and another really big change end of 2025. And the average consumer and the average business leader did not noticethat.

What changes should they expect next? What are you seeing?

So it's, it's easy when we think about what has happened up until now, and then you can kind of predict the future a little bit more easily. Um, and for what it's worth, I do, uh, publish my AI predictions every year. I've done it for the last 8 years. So you can see what I predicted.

They've all been right.

Listen, my batting average is actually really strong.

If I was in the MLB, you'd

want to draft me.

Iwas not doubting you atall.

Um, OK, so, so end of 2022 we get Chat GBT, which is also getting a new model in there. It was GBT 3.5, and all of a sudden you have this model that can yap back and forth with you for thousands of tokens, doesn't forget every single thing you've ever said, has some guardrails where it's not telling you how to make bombs, whatever, and so it's, it's decent.End of 2024, we start to get reasoning models. These are models who would kind of be like, thank you, Brian, for your question, and we kind of go off here, start to think through step by step things on its own.Um, then would come back to you with, uh, hopefully a better answer because of those reasoning steps, as these models got, uh, we're able to take on more information at once and we're able to kind of paralyze, uh, out different workflows and bring it back to itself. We have the ability for AI agents to actually perform because you would give me a very vague task, me being theThe AI. I would go over here and I would go, OK, Brian wants me to plan for this interview with Allie, which I'm sure you did it on your own, but let's just say this. So wants me to plan my interview with Allie. First thing I got to figure out who Ali is. And so therefore I might fire off a research task. Then I should figure out what, uh, episodes have performed well. So I'm gonna look at the YouTube and see all the ones that got over 100,000 views, whatever. And so that end of 2025 was really taking the harnesses that we have, which is like the claw code codex of the world.Combining it with very, very strong reasoning models, and the combination of those two now gets us to, I have an AI workforce with 34 AI agents that are working around the clock for me.Right, I, I literally, even if I had had that idea three years ago,

I agents.

So I only have to manage one technically. Like the, the whole thing that people should be doing right now. If you want to be the most up to date in 2026 in AI, what should you be doing? You should be building the system that allows for work and tasks and goals and whatever to be optimized. And so you should be thinking, how am I putting goals into this system? That's a really big one.How am I connecting the system into context? So for the example with you, does it connect into YouTube and can it see all your previous things? Does it connect to your desktop and can it see all the upcoming shows that are not yet published on YouTube, but that you're working on? Does it have access to your email so it can see all the conversations that you've had with your producers to see what shows are getting delayed or what topics are really hot if it's looking, you know, at Reddit or online forums to be able to see what's trending.So as you're thinking about giving it goals, gathering context, that is really that system that I'm building up. So I manage just my AI chief of staff

who name is Simon.

I do.

You can't, you can't hirehim. I'll try not to curse. I like I have to realize like I'm not in a bar right now. Like I, I don't wanna curs effing awesome. That's just reallycool.

And, and I think what is cooler because any anyone can have that like single AI agent that's doing, um, uh, wide tasks for them.I think where it gets interesting is that so Simon has 6 directs that are named after the Friends characters. Simon's chief of staff. Simon's the AI chief of staff, yeah. And then underneath Simon there's Chandler, Joey, Monica, Rachel from Friends for you out there. Rachel does the client work because she handles clients very well. Monica is obviously operations. Phoebe is just like the wacko do in the corner just like dreaming Ross's education, Joey's product, Chandler's marketing, and underneath them they have like task-based agents.The coolest person, AI person that I brought into the workforce is not anyone that I've already named. I've named 33 now. The 34th is that I gave Simon an AI assistant. And so this AI assistant's entire job is just to be like a watchdog looking over the whole system going, that's wrong. That's a lot of friction. How come every single time we do this task, she keeps giving us feedback like this. Every time we do this task, how come we're missing memory on this? And so it's like aGap analyzer, it's a watchdog. And I think what people are realizing now is that we used to hire jobs like marketing manager or producer or whatever, and because of AI, because at the margin, you can basically hire any AI agent you want.All of a sudden you can hire for any role ever, right? Maybe you're only getting 5% of the value shout to

my producers. I love you all. You're, you're all human and you're all my favorite. Love you guys.

But, but in addition to that, like human producer, maybe that human producer wishes that they had this research watchdog that could look over all of your competitors or contemporaries to see what they're doing.They wouldn't be able to hire a human for that, right? Because maybe the revenue didn't make sense, or uh the time for how long a human would take to do that task didn't make sense, but now you can just bring that in. And then by the way, if there's no value behind it, you can spin it down. But all these things are fractions of a dollar to be able to test out. All of my agents run within my $200 a month subscription, so this is not like the stories, you know, tokenomics of like, oh my God, she's burning millions, no.Everyone in my company, we spend about like 6-7K per head, and all of the workforce stuff that I'm talking about is within that

subscription. So to create one agent, let's say I am new to AI today. How do I create an agent?

Yeah.Like, like, truly new, you've never opened

up. I don't know anything about really technology. Let's say, let's say I'm my, let's say I'm my mom. I'm not gonna give my mom's age away. Let's just say she's older than me for obvious reasons. How will she go about creating an agent?

OK, if I am.Mrs. Sazi, um, which, what a dream. I, I bet she's amazing. OK, I'm going to step one, open up my laptop. Uh, I'm going to probably navigate to something like cloud.AI or chat GT.com as a default just to create an account.Then I would download the desktop app for either of these things, and for, you know, ease of the rest of the example, let's just say it's cloud. So you download the cloud desktop app, there's gonna be tabs that you can switch between, you switch into cloud code. Now basically you're sitting in the agentic platform. Instead of sitting in chatbot land, you're sitting in agent land, you're sitting in task land.The next thing that I would do if I was your mom or someone like that, is I would open it up and I would literally just say into that black box, white box, tan box, whatever. What do you do? Right? I think a lot of people jump in and they think that they should prompt with like build me an AI agent that every morning at 9:17 a.m. tracks my daily morning briefing and summarizes my calendar.Like the average person, the best first step that you can do is open this and just rant about a problem, like switch to dictation mode, rant about a problem, and then end it with help me think of something really creative, interview me, ask me questions, learn about me, you know, don't jump into action, take your time, plan, and the way to build an agent is literally in any of these systems, you say the word build me an agent that.I had a session with some CHROs yesterday. So heads of HR of companies that have tens of thousands of people, and I showed them a screenshot of a prompt where I said, you know, build me an AI agent that and I literally got a text afterwards going, I have heard the term AI agent for now over a year. No one told me that I just open up this thing and say, build me an AI agent that.All these things are very natural language based. And again, in the absence of your mom knowing exactly what problems she wants to solve, I just tell people complain and then ask the AI how you can be supported.

How do you keepyour 34 employees, including your chief of staff, your AI chief of staff? How do you keep them current? How do they stay updated?

So when that assistant finds faults, then I'll jump in, but I also, again, at the margin, you can hire anyone. I have constant watchdogs looking at AI watchdogs looking over my workflows. So at the end of every single week, it's tracking all the things that I worked on and comparing it to my goals.And saying, did we just get closer? If not, why not? What are the gaps? What should we be solving for? What could solve for that? Should we redo a workflow that isn't working anymore? Did she used to use this workflow 5 times a week and now she hasn't used it in the last month and we can spin it down. So it is constantly analyzing that on a weekly basis. And so my Fridays are actually kind of crazy because I get all this influx from my AI systems.I get a readout of how goal aligned I am, a readout of every single urgent email I haven't replied to. I have a revenue engineering workflow where every single week, I think on Tuesday mornings, I get a readout of how much money is waiting for me in my inbox, and it tracks where every single dollar is coming from in my business, and it is calculating likelihood of clothes.Likelihood of joy, right? I'm not just chasing revenue. I'm chasing whether that revenue is ally aligned. It's gonna be

joyful. Yes, yes, yes. How do you, do you unplug it all? How do you like, do you ever pull yourself out of this AI digitalworld?

So, um, what, what I have learned that as I talked to more people that are, you know, these AI super users, which I would absolutely put myself in that category, um, what I've heard from them and what I feel too is that as our work lives become very, very, very AI enabled.It is actually allowing us to completely separate when we want to separate. So like I went on a one month expedition to Antarctica and didn't have internet that entire time and had technically for that I had AI like posting for me while I was gone. Um,

your teamis still working hard for you, these 34 AI thisthis team of AI and

so I think like I, you know, again, I'm going on hikes, I'm stepping away and things like AI on my mobile phone, AI through dictation.I am able to now walk like I feel like I'm more physically fit than I was a couple years ago because I'm able to bring this stuff away from my laptop. Like I think if you had looked at me working at a very large tech company, you know, a couple years ago, I was glued to my chair until 2 a.m. and now I can bring that work, not that I, you know, always want to bring my work out, but I can get outside. I canBe more active. So I think that that's helpful. Um, and the, the bigger change, which, um, is, you know, you can decide whether to have a good work-life balance or not, but the bigger change is in 2026, the really, really intense super users are not actually delegating work off to AI. This is a really weird thing, right? Because most people haven't even built AI agents, and I'm now saying actually that was two chapters ago.So build AI agents, build multiple AI agents, run it as a workforce, delegate stuff to the workforce. The next step that almost no one has gotten to is that actually you're building the system where the AI agents are going, um, let me check on that. Let me take this off her plate. Why did this just come in? Let me go ahead and handle that for. And I am managing the escalations that come back to me. So I might have 4 or 5 parallel work streams happening.And then I might be on a walk, and my AI agent is gonna say, hey, we just found these 5 things. We went ahead and drafted these 4 replies. How do you want to manage this last one?So it feels a little bit less like I am the direct manager of the system and more that I'm like the SVP above thesystem.

If we were having this conversation 40 years ago, you're 34, 40 years ago, I'm saying 40 years, 40 years, your 34 agents would have been humans. Like when you're inside these bigcompanies, my

34 agents would have been humans, and I only would have had a couple of them because of how much I could afford.Of how much I could manage, of how many resources they had, so I wouldn't have had 3.

So what is this, what does this AI agent world mean for the future of the workforce, especially as you're talking to HR departments. These in a lot of cases where these layoffs are coming, it's HR, it's backroom technology, it's support functions. Yeah,

and I don't believe that the layoffs are attributed to, uh, I don't believe that they're actually caused by AI. I believe that they're heavily attributed to AI because it helps stock prices, butUm, I think that if I were to look at the AI agents and conversations that I've had with people and HR leaders, and also finance leaders, marketing leaders, the, the general sense that people have right now is that we're going to have a hybrid workforce and that if you look at systems today, like the ability to run something called dynamic workflows inside of cloud, if anyone wants to look it up.It's kind of the ability to run massive parallel workflows that can spin up thousands of agents. So right now, I told you that my AI chief of staff has a name and his directors have names.Within that, they, Rachel or Ross might fire off a request for 1000 agents to all run in parallel to all search, you know, different forums, different websites, um, maybe like, uh, if you're going to a conference and you have 500 names that are attending the conference, you could ask Chachi BT to look up name one. Wait, wait, wait, look up name two, wait, wait.Or you could ask Codex and say, here's the list, fire off 500 agents and you get back the answer in 30 seconds, instead of waiting an hour for everything to finish. So that hybrid workforce, some of them are, are named and shared entities where you and I are teammates, human teammates, and we are both, uh, enabled and supported by AI and I want to talk more about that.And also, there's these temporary agent support systems of thousands of agents that are being spun up to help on coding projects or whatever.

Where, you know, I think I really do believe this is going to require a massive retraining of the workforce. Yeah,

and not just the tool part, it's so much on the mindset. Should,

should this be, should this be forced on workers? Like, you need to know how to make a frigging agent or you shouldn't be here because I'm listening to you and these agents are driving massive productivity. And if I'm the CEO, uh, or the CFO of a company, I'm like, everybody needs to know how to create 15,000 agents. Yeah,

if it were forced on you, would you do it

joyfully?Well, if I want to keep my job, yeah, I mean, forced on me. I, I want, I want to learn because if it's not forced on me, I, me, maybe I don't try it.

Yeah, so I think here, here's, so I, I'm blessed with, you know, millions of followers, and what that means is that they DM me with what they're feeling. General sentiment is that when you come from a work culture that has never required any training other than like, you know, don't bribe the government type training. Um great, very important.What happens in those workforces is that they go, whoa, whoa, whoa, why are you requiring this? I feel like I'm being forced into it. You're not letting me move at my own pace, and there tends to be a stronger hesitancy. Um, and so in cultures that have never had that requirement, I, I actually don't recommend that it gets required. I recommend that you incentivize the heck out of your workforce and inspire them and motivate them and make sure that they're seeing positive examples so that they want to do this, right? And then I think you'll be able to inspire them and I can give tips there.The um, the other piece though is that I have worked at companies like Amazon. We were required to take quizzes, tech quizzes every single month. Our VPs were measured on that. If we finished it earlier, they might have gotten a bonus from it. Like that was part of the culture was like hardcore test takers, and of course you would have required that sort of training at a company like that, cause it was already baked into the culture, so why wouldn't you take advantage of that?Vast majority of companies are not going to fully require it, but they might incorporate it into their performance management system where they say part of the way that you're getting evaluated is how you're using AI to multiply yourself or to improve the quality of your work, or to improve the morale of other people around you, or to uh grow the revenue that your team is pulling in, or to decrease churn rate, like whatever other metrics matter.But you would expect that they make it easy for you to get the actual training, make it easy for you to see great examples, make it easy for you to access agentic platforms and not just chatbots.And make it easier for you to do the right thing. And that means also having good governance in place. But most companies have not given all those things on a silver platter. And so they still have low, low, high quality adoption.

How do you see this,these agents or this technology changing the role of the CEO?

Well, OK, so right now, well, forever. This, the role of the CEO is I set the vision. I am corralling this system, right, which largely is people to be able to drive toward this, and I am responsible for shareholder return or stakeholder if it's more privately held.If I'm the CEO, the first thing that I have to think about is how does my vision change or how does the velocity of that vision change with AI in mind, and that is a veryUh, uh, not a hollow view, but it is a very like a blind blinders view where I am wholly focused on what I do as a company, me, me, me. What I think the majority of CEOs are not paying enough attention to is the fact that if you have AI and you're enabling everyone and everyone just got 10% more productive and you feel really good about that.What you're forgetting is that there are AI super users who are much higher, you know, they have higher risk tolerance. They are way deeper into these tools than your workforce is, and they are going to sprint past you. So the threat of smaller businesses against incumbents is higher than it has ever been and will only continue to increase. The moats of what that CEO is thinking of are plummeting.And so if you're only thinking about this in a CEO vacuum, and you're only thinking about how do I make my workforce more effective so that we can drive more shareholder value, how can I think about our customers and how to serve them more efficiently or effectively, you are going to lose the, the bullet train like flying by you, um, and you're only going to focus on like the very slow bike that you're on. So if I'm a CEO, one of the most important things is actually to test out these tools. Majority of CEOs I talked to have never built an AI agent.And they talk a big game, right? They're in these boardrooms

and I think they're still stuck in a decade, you know, past, you know, they, they have at least they're public company. Earnings calls this way, they're gonna meet investors, they're gonna do, they, they've still not.Adapted to this day and age, and they're not going to, and they might

actually, they might actually show up pretty, pretty well on these shareholder calls and talk about AI and talk about how they're using it, but that's a speech that was written by someone else. That's correct. And so if I'm presenting, I speak in board meetings, I speak with execs and exec teams. If I sit down with an executive team, I might hear, oh, we're all in, we can't wait. And then I pull them into one on one meetings and I learned two things very, very quickly.Number one, they have three things I guess. Number one, they have not yet caught up to 2026 AI and enterprises are always going to have a little bit of a lag, so it's to be, you know, expected. So they are not yet at today's AI levels. Number two is at least one person in that Xcom on the leadership team, the executive committee.is definitely not using AI and probably a detractor and doesn't want their department using it and they think their whole department's gonna go away if they don't use it. And the third is that it's probably, it's probably the case that the point person or one of the point people, again, assuming they're, it's outside the CTO role, but a non-technical executive, I pull them aside and they go, friend to friend, what the hell is an AI agent, right? And these are brilliant business minds that are running.Companies that are making tens of billions, $100 billion plus dollars, and they are brilliant. They just are so busy that they're looking for that signal among noise. And so one of the things that I do as an advisor is I go, let me summarize the last five months for you and give you the three things that you actually have to know. Let us build this thing together so I can watch you have that aha moment, and then you can go back to leading in the way that you need to lead.So I think a lot of CEOs, again, they're not using the tools. If they are, they're not using it at 2026 levels. And if they're even doing that, then they haven't figured out how the moats and competitive advantages have completely flipped toward compounding gains, right? My AI system and your future AI system, because obviously we're going to get you to build one out as well.

Thank you, as they say, we'll take that offline. We'll take that offline,

yeah.OK, you're gonna have an AI system. I have mine, and the, the glorious part of this, the delicious, amazing part as a business techie, when I look at this, is watching that system improve without me whipping the hand of every single AI agent going, uh, can you make this 2% better? It is improving without me having to constantly jump in, much in the way that a human workforce does as well, that a CEO doesn't have to constantly jump in, that VPs and directors and senior managers take that on.If you can imagine running a very, very large enterprise, and your challengers have these systems that are not only allowing them to do the work of 1000 people with only 30, but that system is improving every day or every week, you are at a greater disadvantage with every single coming week.Like it's crazy to me that people have 401ks and they can't figure out the idea of compounding gains with AI.

Yeah,one more for you, um, and I'm gonna lean into your expertise at really good predictions. Five years from now, how has this technology changed corporate America?

Oh, so many things, um.So I think one that is starting to happen right now is the blurring of lines of jobs. And so there was a recent, um, open AI report that is fascinating that I encourage everyone to read, where OpenAI, they looked into 800,000 chat GBT messages. And remember this is standard chats, not codex, whatever, but they looked into 800,000 messages, and they looked at the work related messages, and what they found is across the board.That nearly half of people's messages, so like if I'm in whatever sales, nearly half of my messages were not related to sales, and they're related to other functions. And if you look at this graph, designers are using AI more for engineering than they are for design. Finance, they're using AI more for marketing than they are for finance. HR they're using AI more for finance than they are for HR.So all of these lines are getting blurred. I have a really hard time imagining that anyone is going to want to hire someone with as niche of a title as like SEO specialist or something in 5 years, because what you're actually trying to do right now, what workforces are trying to aim toward is getting as many players as possible, and giving them wider scope, and giving them these growth oriented tasks, right, not just productivity, and that is kind of what the the work of the future looks like.I have very strange predictions around voice. I think voice AI is actually the interface of the future. I have a hard time picturing that open offices are gonna work when you have a lot of people yapping to their AIs. I went into the headquarters of Whisperflow in San Francisco. All of them have gooseneck microphones.And they're all just whispering into

the, so no more keyboard.

It's, I mean, they still have one, but mouse. I would, I could chuck my mouse across. I like, I just, I use VR still and like I just want to like pinch. I want to do the minority Report, like zoom in, zoom.That is all I want. And so I think, I think in 5 years, I can imagine there are a lot more voice on subways, in the street, at work, which I think has negative consequences. Like I don't want to walk around hearing people yap to their AIs all day.Um, and so companies are gonna have to think about that. I think we'll have a good early sense of what an AI first device looks like, right? So OpenAI has shared that they're going to show what the device looks like in 2026 and start selling it in 2027. So it's gonna

be a pendent. It's gonna be a pendant.We'll, we'll see, we'llsee.

Um, so I think we'll get some senses of device, a lot happening in voice, blurring of jobs. Um, those are, those are some of the stronger predictions. But my, my louder prediction that I hope every single person feels is thatIf I'm a CEO or CFO or or a busy parent of 5, and I'm just trying to survive my workday, the number one thing that I would be paying attention to is the pace of change, right? How fast are these models getting better? How fast are new models getting rolled out? How fast are we able to add on new features inside of these AI tools, etc.And I would be looking at that pace of change, and I would be thinking, how can I build a little bit more flexibility, a little bit more adaptation into my workday, into my career plans, and I would be assuming that the pace of change is gonna be faster than I think it is. If you would, if we had had this conversation 10 years ago, I would have said that we would have been here in like 2040. And so even all of my friends where we've worked in AI every single day.I have not talked to a single person whose timeline has had to get longer, right? Maybe in like weeks, months sort of trade off, fine, but in terms of AI capabilities on an annual let's check in with each other basis, it is always faster. Things like robots, hardware, I think it's way far off.We should expect that these systems are going to be very reliable at a quarter's worth of work that it would have taken an entire team to accomplish, and that it'll be able to complete all of that in hours.

Sorry team, I know I'm over time, but I can't help myself. Did you have your agent like research me before you came on here?No, no, I'm just curious. I'm just curious. I had to ask, like, you just like,

I had a true.

You are. You are, you are a true expert.I feel as though, uh, you just took me to school on all of this. Um, let's go create that AI system stuff. Uh, good to see you, Ali. I appreciate it. Thank you for having me. You rock. That's it for the latest episode of Power Players. Appreciate all that love.