#215: Musk v. OpenAI Round 3, AI's Hot New Job, The AI Jobs Apocalypse Debate & The 2026 State of AI for Business Report

Introduction

0:00What happens when you have a nation of people that used to be doing okay with white-collar jobs and now have to drive DoorDash or Uber or something? You're going to have a lot of angry people. Welcome to the Artificial Intelligence Show, the podcast that helps your business grow smarter by making AI approachable and actionable. My name is Paul Reitzer. I'm the founder and CEO of SmarterX and Marketing AI Institute, and I'm your host. Each week, I'm joined by my co-host and SmarterX Chief Content Officer, Mike Kaput, as we break down all the AI news that matters

0:33and give you insights and perspectives that you can use to advance your company and your career.

Episode Theme

0:39Join us as we accelerate AI literacy for all.

0:46Welcome to Episode 215 of the Artificial Intelligence Show. I'm your host, Paul Reitzer, along with my co-host, Mike Kaput. We're recording on Monday, May 18th, 9.30 a.m., which could be relevant to our first main topic, which is going to be the Musk versus OpenAI, because Jerry is in deliberations, and I don't know how long that's going to take. It might be one day, it might be five days. So, by the time you listen to this, who knows? We may already have the verdict in the case. We've got some big picture things to talk about today. I don't, Mike, like, I guess there's a bunch

1:17of updates in the product and funding. No major models last week, but just some massive topics that I think are going to, as the year progresses, take on far greater importance, I would say, within business and society. So, buckle up for today. We've got some big topics.

Sponsor Message

1:37Today's episode is brought to us by AI Academy by SmarterX, which helps individuals and businesses accelerate their AI literacy and transformation through personalized learning journeys and an AI powered learning platform. As I will note today, when we're going through a couple of our topics, one of the really important things about our AI Academy approach is it is a human-centered approach. So, we are trying to teach responsible use of the technology, and that is going to be increasingly important, especially to this next generation of professionals who are coming into the workforce

2:09as we speak, graduating in May, and who aren't really loving AI, as we will discuss. So, if you're

Responsible AI Use

2:17hiring that next generation, you need to be thinking about how do we teach AI in a responsible, human-centered way, because they are not going to have it any other way. It's going to be a really important balance people are going to have to figure out. So, with AI Academy, new educational content is added weekly, so you are always up to date with the latest AI trends and technologies. We feature a number of collections of courses, and so one of those is AI for Industries, which has seven course series and certificates available on demand right now. These are designed to jumpstart AI

2:48understanding and adoption. Those collection features certificates in professional services, healthcare, software and tech, insurance, financial services, retail and CPG, and manufacturing. And we are adding new industries every month. So, stay informed. If you're not hearing your industry there, there's other collections for you, including AI for Departments and the Foundations collection. And like I said, new stuff getting added every month. So, these series are an ideal launchpad for

3:18organizations that want to level up their teams and accelerate responsible AI adoption and impact in their organizations. Individual plans are available, as well as business accounts for five or more licenses. Visit academy.smarterex.ai to learn more. And for individual plans, you can use pod100 for $100 off your annual subscription. So, again, that's pod100 at academy.smarterex.ai. All right. So,

AI Pulse

3:47every week we do an AI Pulse. This is just a quick informal poll of our audience to see how they're feeling about topics we've covered on that podcast. Last week, we had a poll. The first question was, should new powerful AI models be vetted by the U.S. government before they are released to the public? This one is almost exactly split. We have, yes, but only for the most powerful or highest risk models got 40%. We have, no, voluntary safety testing by labs is enough at 30%. And then we have,

4:21no, the government should not be involved at all at 27%. So, that's 57% saying no. And then we had a very small sliver that says, yes, mandatory pre-release vetting for all frontier models. So, yeah, again, informal poll of our listeners. The second question is, is your organization actively replacing roles with AI today? Okay. This is dominantly, no, not yet. 69, 70%. No, we are not doing that yet.

4:51Next closest is 15%. Yes, but unofficially through attrition or hiring freezes. And then 9%. We are considering it, but have not acted yet. And then a sliver at yes, we have cut headcount specifically because of AI. It's around, what, 7% or 8%? Yeah. It looks like it. Yeah. Yeah. So, interesting. So, keep that one in mind today. We are going to be talking about jobs a little bit. There's been a lot of news in the last week related to jobs. So, right now, a very informal poll. No, not yet. It's

5:2370% in terms of organizations. All right, Mike. On to the main topics. If you're new to the podcast,

Main Topic 1

5:29again, we have new listeners every week. So, it's helpful reset here. What Mike and I do is we go through probably between, I don't know, 70 and 100 sources each week between podcasts and expos and research reports and articles and videos and keynotes and all these things we consume throughout the week. We curate those down. We pick three main topics. Mike does an amazing job of this on Sundays. He puts in the work on a Sunday. He goes through, curates the three main topics, and then we try and do about seven rapid fire items. And those, so we try and get to 10 total,

6:02sometimes a little more, a little less. And then we consolidate all the product and funding news into one update at the end because that alone could be 15 to 20 topics every week. It's becoming burdensome to try and manage just selecting like the 10 or 15 we feature. So, that's just how we run this. Three main topics, and then we get into the rapid fires from there. So, Mike, kick us off with

Main Topic 2

6:25main topic number one. All right. Yes, Paul. So, a lot going on this week. We talked first about this ongoing Elon Musk versus OpenAI trial. We have talked about this on the last couple of episodes. And in the days since, of course, the trial wrapped closing arguments, and a nine-person jury is now in deliberation. So, OpenAI co-founder and former chief scientist Ilya Sutskova took the stand Monday. He testified that he spent roughly a year gathering evidence of what he called Sam Altman's

6:56consistent pattern of lying. And he had prepared a 52-page document for OpenAI's board detailing those concerns, which we've talked about before. Interestingly, for those keeping score at home, Sutskova disclosed that his OpenAI stake is now worth approximately $7 billion, up from about $5 billion in November 2025. Microsoft CEO Satya Nadella also testified Monday. He said he never believed Microsoft's $13 billion in investments violated OpenAI's nonprofit mission, and that Musk never

7:26once contacted him to object, despite the fact that they, he says, have each other's phone numbers. Sam Altman himself testified Tuesday. He said Musk tried multiple times in 2017 and 2018 to merge OpenAI into Tesla or convert it into a for-profit he would majority own. Altman called one moment during that time, particularly hair-raising. Musk told co-founders reportedly that control over OpenAI could pass to his children when he died. In Thursday's closing arguments, Musk's lawyer Stephen Mollo accused Altman

8:02and OpenAI president Greg Bachman of stealing a charity. OpenAI's lawyer countered that Musk never cared about the nonprofit structure. What he cared about was winning and showed evidence that Musk himself proposed turning OpenAI into a for-profit in 2017, in which he would have held over 50% ownership. Now, as a reminder, Musk is asking for about $150 billion in damages, Altman's removal from the OpenAI board, and an unwinding of OpenAI's for-profit conversion. So in this case, actually, the jury's role is

8:33technically advisory. Judge Yvonne Gonzalez-Rogers will decide the actual penalties. Obviously, everything's kind of on the table. The options range from granting Musk everything he wants to throwing the verdict out entirely. So Paul, we've been covering this trial now for weeks. What is your read on what actually matters here? Where do you think this is going to go? The verdict itself, I'd be curious on your thoughts on where that's headed. Yeah, I have no idea where the verdict goes. I think, I mean,

9:05what largely has been accomplished is a whole bunch of private text messages, personal journals, emails from these tech companies that the public would never see or generally even know how these people think were just laid bare for everyone. I don't know that either side wins in the court of public opinion. I don't know that either of them came out looking great here. Microsoft spent most of the trial trying to distance themselves from everything. But other than that, it's like billionaires

9:36fighting over trillions of dollars. And so I think that that's one of the things that will play into one of our topics a little bit later is just how society's feeling about all of this. And I don't know that for the most part, people really care. Like we care because it's in the AI industry and open AI is obviously a major player, has a Elon Musk. So it is like intriguing, certainly from just the show of it all and the inside information. And, you know, a lot of validation of these, you know, journalists who were uncovering these stories along the way and being told that that

10:10wasn't true and it was misinformation. And then all of a sudden they have to admit to all this stuff in court. So I think that's all fascinating. I don't know where it goes. I don't know. I can't imagine that it ends up leading to an actual changing of the structure of open AI or the ouster of Altman and Brockman or, you know, the dissolving of the for-profit. Like, I just can't see that.

10:40But I don't know. Like, it's going to be really interesting to see how the judge, because like you said, the jury doesn't decide this. They're basically, the nine-person jury is making recommendations in essence to the judge. And then the judge is going to actually make these decisions. I thought it was notable that Altman and Brockman were there last week for the final stages of the trial while Musk was in China with Trump. So, and apparently he wasn't allowed to leave because he was on a callback, whatever, I don't know, legal term for it, but he needed to be available for callbacks. I'm sure he doesn't care about that. Like, I almost feel like Musk feels like he

11:15accomplished what he wanted to, which was probably to embarrass them and like make it really, their lives really difficult. And now he just doesn't even care. Like, whatever they decide, they decide I'm off to China now. That was at least the perception I had of it. So, I don't know. We could be shocked. We could be sitting here next week talking about an unexpected outcome that throws their plans for an IPO into chaos and resets the competitive structure of the AI landscape, maybe, but I just don't expect that. I could see some sort of penalties, but who knows?

11:50I haven't heard any good real analysis of like what the likely outcomes are. It doesn't seem like anybody has a clue. And I've been following the journalists who've been in the room, you know, following along and everybody's just kind of documenting what's going on. No one seems to have a clue how it actually plays out. I'm glad we're not the only ones who are not able to see around the corner on this one. All right. So, Paul, our second big topic this week, this past week, a title called Forward Deployed Engineers or FDEs became basically the most talked about role in enterprise AI. So,

12:25these are essentially engineers who embed inside customer organizations to design and deploy AI systems alongside frontline teams. So, the reason this is becoming such a hot topic now is that OpenAI made a huge move on Monday that we had kind of teased last week with the launch of this thing they call the deployment company. So, this new business unit is dedicated to embedding FDEs inside customer organizations to identify high-value AI workflows, redesign critical processes around them,

12:57and turn the gains into durable production systems. So, this deployment company actually launched with over $4 billion of initial investment. It is on the investing side led by TPG with Advent, Bain Capital, Brookfield, and Goldman Sachs among the capital backers, alongside also partners who are systems integrators, including Bain & Company, Capgemini, and McKinsey. Additionally, OpenAI is acquiring a company called Tomorrow, an applied AI consulting firm whose clients include Tesco,

13:30Virgin Atlantic, and Supercell. They're bringing literally 150 experienced FDEs into this from day one through that acquisition. On the same day as that was announced, Google Cloud CEO Thomas Kurian announced a new AI-focused organization inside Google Cloud's go-to-market team. They have plans to hire a ton of additional FDEs to scale customer AI transformation. This sits alongside a previously announced $750 million ecosystem commitment to help Google's 120,000 member partner network deploy

14:03agentic AI. And finally, we had a ton of commentary online about this role. So, Box CEO Aaron Levy posted that FDEs are about to become one of the most in-demand jobs in tech. He's arguing that deploying agents is far more technical than deploying traditional software because vendors need to deeply understand the customer's business process and deliver work output, not just software. He actually also urged college career counselors to start steering students towards FDE roles. Allie K. Miller,

14:33a prominent AI voice, however, did push back on this, warning enterprises that treating FDEs as their entire transformation plan can be a very expensive mistake due to all of the other change management, communication, and education that is required to achieve these changes, not just technological deployment. So, Paul, there's a ton of talk about FDEs now. It seems like everyone is chatting about this or investing in this. What is your read on this? Because we've kind of, I think, talked about the

15:03need for some type of role that might look and feel like this in the past, but is it purely an engineering role? What does that look like in your mind? We did touch on this topic last week. I think I mentioned I was kind of getting tired of this term already, but I don't think that's going to change anything. I think we're only going to hear more and more about this term. As we mentioned then, Palantir sort of made this job title popular, you know, 10, 15 years ago. It is, in essence, if you're wondering, isn't it just a consultant? Yes. The answer is yes. It's a consultant who has

15:34technical ability to actually go in and, like, customize software in this case, AI models and agents to solve problems for businesses. So they have more technical ability to go in and do that kind of stuff. I've talked with friends who work with Palantir and organizations like that and worked with these FDEs. And what they're often doing is going in and just trying to find business problems to solve to charge money for. So now the beauty with the FDE role is they can actually charge on an outcome pricing basis. You know, we talk a lot about what are the pricing models of these

16:06AI companies and how they're kind of metering it and charging by tokens. The beauty of the FDE role is they can go in and say, what is a business problem? They identify one they can solve with their AI technology. And then they say, okay, this is worth $100 million to you. So we're going to charge you $25 million or whatever that is. So you can do true outcome-based pricing, which makes a massive difference. Now, at a high level, while they're not going to say this outright, they're going to tell you they're working with consulting firms, you know, the Deloitte's and the McKinsey's and things like that, but they're a hundred percent going after them. Like they're coming for your work. Like

16:39they can be partners as much as they, you know, want, but the reality is there's six trillion in labor wages in the United States for knowledge work roughly, and they're coming for it. So the FDE serves a couple of roles. One is it's hard to do this adoption. Like it's hard to apply these AI models and the agentic capabilities within organizations with the existing staff of those firms. They don't, you know, a lot of enterprises aren't structured to know how to do this deeply.

17:11And so there's absolutely like a functional need for this, but there is also a capitalistic need. There's a need to generate massive amounts of revenue to justify their valuations and their IPOs. And so FDEs is a bit of a Trojan horse, in my opinion, for what they're trying to do. Yeah, I was doing a little prep work on this when I found an article from Salesforce in March of 26. And we, I don't think we covered it at the time. We weren't really talking too much about this, but we'll put the link in the show notes. It was a blog post called today's hottest

17:44role forward deployed engineer. And I thought they had some good context in here. So I'll just read a few excerpts. He said, consider the role a natural reaction to the times AI has burst onto the scene faster than many companies can adapt. FDEs are like a personal tech guru, business consultant, and handholder all in one. They work closely with companies to remove blockers and accelerate AI adoption, as well as share customer feedback with product teams to make AI agents better. The software company Palantir pioneered the FDE in early 2010s when it embedded engineers directly

18:16with customers, mostly government agencies at that time, to help implement products. Palantir called these engineers deltas at the time. Until 2016, they had more deltas than software engineers. The role evolved and gained huge traction this year when tech giants like OpenAI announced they were hiring FDE teams. Analysis by Indeed and the Financial Times found that job postings for this role soared by more than 800% between January and September 2025. Salesforce alone has committed

18:47to building a team of 1,000 FDEs. The role varies from company to company, but at Salesforce, which launched its team in April 25, FDEs work in several ways. Many work individually with a customer, but some are starting to work in pods that consist of one deployment strategist and two FDEs. I found that interesting. The deployment strategist identifies the best use cases for a company and creates the overall AI strategy. The FDEs then design, build, and deploy the agent. They're the team's technical

19:17architects and primary coders. Pods focus full-time on one client for about three months, thus the forward deployed part. They go in and just literally work with this client for three months, or as long as it takes to successfully deploy an agent for one or two use cases. So they're very focused in their efforts. Sometimes they even travel to the customer and embed themselves in the customer's day-to-day work. It may sound like FDEs do the same work as Salesforce partners. This is an interesting commentary, but they play different roles and their collaboration can help customers launch agents

19:49more successfully. Partners still do the nitty-gritty work of helping customers implement technology, but FDEs can provide behind-the-scenes knowledge from Salesforce that a partner may not have. A quick context there, what they're doing is they're talking to their solutions partners who are value-added resellers of Salesforce. These are agencies and technical partners who make their living providing services on top of Salesforce. Salesforce is trying to thread a line here saying, we're not competing with you, we're enhancing what you're doing, not taking work from you. That's a

20:23tough line to draw. I'll come back to that one in a moment. And then they finished, one reason competition for FDEs is so fierce is that the role requires an unusual combination of skills. Not only do you need to be a tech whiz, you also have to communicate well and be comfortable in a customer-facing role. FDEs and deployment strategists have to listen to customers, understand how business works, and offer solutions in language non-techies can understand. That is a very rare mix of capabilities. It is not often the person who can build the solution is also a strong customer-facing

20:56professional. So you can get at it. So why all of a sudden, like what is going on here? A couple of notes. So my basic take is the technology is advancing faster than enterprises are able to understand and adopt it. And agentic AI, which as we've talked about many times on this show, starting in December of 2025 in particular, agentic AI capabilities have accelerated the urgency and the complexity of scaling AI. In the Levy post you mentioned, he wrote, deploying agents is far

21:28more technical of a task than most people realize, often far more involved than deploying software. Software generally works the same way every time. And generally for the past few decades has been updated versions of an existing technology or concept, which basically means easier to the enterprise to adopt their workflows on a newer system. With agents, you're actually deploying the equivalent of work output within the enterprise. The customer is effectively using you as a professional services provider for a task, which they expect to get solved nearly end-to-end now.

22:02This means you need to actually deeply understand the business process as a vendor and get the customer from the current to the end state seamlessly. So FDs are technical, but the same opportunity and need exists for non-technical consultants. So this is my commentary here. So we think about agencies. So again, if you're not familiar with our history, I owned a marketing agency for 16 years. We were HubSpot's first VAR, first value-added reseller solutions partner back in 2007. So I grew up in this ecosystem, this idea of partner ecosystem and a solutions partner ecosystem. And so when I look at

22:35the needs right now, these FDs that OpenAI and Google Cloud and everybody is building, Salesforce, HubSpot, I would imagine is probably getting ready to announce they've got one. What they're building is solving the technical side of this. And the agentic stuff is making the technical side higher in demand. But as Allie Miller alluded to, it's the non-technical stuff, the support around adoption and change management and communications and education and training. That need is going to be as big or bigger than the technical lead, in my opinion. And all of this is based on the premise.

23:09Like when I was in the HubSpot ecosystem, they had a report, and I think I might have mentioned this last week, $1 in software sales equals $6 in services. So the premise is if you're spending a million dollars with OpenAI, you're going to spend $6 to $10 million on the services to make that million dollars work. And so that's the money everyone is now going after. So OpenAI and Google Cloud and others can sit back and wait for a solution partner ecosystem to emerge that can provide this level

23:43of service. But they don't have time to wait. They need their AI models and agents working right now. So they're just going to go hire all these people. And so even if they say they're not directly competing with the consulting firms, they're 100% competing for the talent that those consulting firms would otherwise hire. So this is what happens. This is how it works. Like HubSpot in the early days, we were part of, in essence, a pilot test to prove to them that solutions partners could drive retention and growth of accounts. And so what then ended up happening is over time,

24:14as the ecosystem sort of leveled out and you had different levels of capabilities across the different partners, HubSpot realized that they could not depend on the same level of service and quality from all the partners. So they started offering the onboarding services themselves and started charging, I don't know what they charge today, but it used to be like $5,000 to do an onboarding installation because that's how they ensured that the client would actually get value. So this is a constant conflict going back decades where technology companies sell a solution,

24:45but then they need services on top of it. And your options are build a solution partner ecosystem or do it yourself. And sometimes they try and balance between those two. And then the final note, Mike, here I'll make is, you know, we released our state of AI for business report last week. We'll put a link in the show notes to it. I think it's stateofbusiness.ai. I think you can go and get it. So the two, yeah, the two key things we asked the question, what are the barriers to adoption? Number one, lack of education and training. Number two, lack of awareness and understanding. And so this is the stuff that has to get solved. They don't know how to do it. They don't know what

25:18the models are fully capable of and they can't train their people fast enough to do it. So a complimentary approach is you bring in the experts to do it. And like I said, there's just massive amounts of revenue and they're going to look at that $1 equals $6 to $10 in services. But the bigger number they're all looking at is the $6 trillion in labor that every year. So the US is like $11 trillion, I think overall, and about $4 to $6 trillion is knowledge work. And so they're totally going for that. I don't blame them. Like it's a logical thing. You want to call it an FDE or a technical

25:52consultant or whatever you want to call it, it's needed. You know, when you were talking, I just couldn't help but like nodding along, especially at the non-technical point, because I read about FDEs and I think it makes sense as a concept and I can certainly see the need for it. But if we think that this is the only role that needs to be in this kind of forward deployment of AI, I think we're wrong on that. Like this doesn't work just with FDEs. Or to your point, it has to be an FDE who's uncommonly good at all those other things, which is not a lot of

26:25engineering-minded people, I would say. Totally. Yeah. I mean, we're doing something similar. I don't think I've talked about this publicly, but we're doing something similar with AI Academy. So, you know, AI Academy of SmartRx, the way I think about this is in essence, I told the team months ago, we basically have to build an internal agency to help organizations do the education and training. So there's a lot of enterprises that have L&D teams in place that, you know, can take AI Academy and run with it. But then there's a lot of organizations that don't or that like we're working with a marketing team that maybe doesn't

26:57have the full support of the L&D team to do what they're trying to do. And so we think about what we do through Academy is AI transformation, not just selling courses. So again, I've mentioned before, like Coursera, LinkedIn Learning, like there's great places to get courses. That's not what we're trying to do. We're trying to actually get in there, identify what are your goals, what does success look like? What are the barriers to preventing that, you know, in terms of communications with employees, dealing with cohorts of employees who don't want to do this stuff, don't want to learn AI. So I see it

27:28as like we have to build a similar concept. I would not call them FDEs, but we have to function as advisors to help our AI Academy customers drive transformation and change within their organizations. And you have to be sensitive to all these complex issues. And so I think about a similar approach where we have to build advisory services on top of AI Academy to go do this. And then we may also empower an ecosystem of partners to support in that way. But similar concept, like we're offering a solution that on its own

27:59needs additional guidance within many organizations to make it happen. And so I'm building a system to do that, you know, and supported by different technologies. But we're also looking at the service side and realizing we're going to have to build that capability pretty quickly. Yeah. And I'm just curious, as we round this out, what your thoughts are, I kind of found myself agreeing with Levy when he was like, hey, this is a huge opportunity for possibly recent college grads or more entry level people. He focused on the computer science aspect of it. Hey, like, if you have a computer science background, FDE is an amazing path that we should be counseling people to go down.

28:35I could almost see whether you call it an AI accelerator, just an advisor, consultant, whatever. I think this is a potentially large opportunity for not just entry level careers, but if you are an AI forward version of anything, a marketer, a salesperson, operations person, there's a real interesting path here in my mind of how you could add value to an organization doing this kind of thing. At an individual level, definitely. You know, I think the premise holds true. I also look at it as, you know, from an organizational perspective, if you're an

29:09outside solutions partner and you want to specialize in open AI or anthropic, like, again, like we did with HubSpot back in the day, I just bet everything, the future of our company, I just bet it on HubSpot. We went all in. We didn't try and learn Marketo and Salesforce and Pardot and all the players at that time. We just said, let's just get great at HubSpot. And so every employee was trained deeply in HubSpot's capabilities. We would often get on calls with their customer support team and we would solve things that they didn't even know were problems. And that's what made us a great solutions partner is you get to know the product oftentimes better than the own

29:44marketing sales and customer success teams of the technology company. And so I think there's a tremendous opportunity right now for service firms to go in that direction, whether you specialize in capabilities across a collection of AI platforms and companies, or you just make your bet. We're going to go in Google Cloud. We're going to go in, you know, with open AI or anthropic or whatever it is, and you just get great at it. These companies aren't going to hire the FDs fast enough. The issue you're going to have is them then coming and hiring your people. That's a real concern.

30:17All right. Our third big topic this week. So in the past week, we've seen a range of very prominent voices pushing back very hard against the narrative that AI could fuel a jobs apocalypse. So this included people at Andreessen Horowitz, the notable commentator, Scott Galloway, Andrew Ng, Derek Thompson, the journalist, all of them made this case from kind of different angles that the fear of job loss from AI is largely manufactured. So Scott Galloway wrote an essay that framed the AI

30:48jobs apocalypse as narrative driven, engineered by people who profit when you're scared. He pointed out that US tech employment has stayed flat at 9.6 million for three years. He argued that recent layoffs from Meta, Microsoft, and Oracle are mostly returns to pre-pandemic headcount, not AI displacement. And he basically claimed that the AI job apocalypse narrative is just a marketing strategy for the labs to make their products seem more valuable and to be able to charge more for them. Over at Andreessen Horowitz, David George argued that the AI doomer position is the old lump of labor

31:24fallacy with new branding. So this fallacy assumes a fixed amount of work. So if AI does more, humans must do less. George pointed out that history contradicts this. For instance, agriculture went from a third of US employment in the early 20th century to about 2% today, but output tripled and workers flowed into entirely new industries. Andrew Ng called the AI employment narrative irresponsible and damaging, pointing out that software engineering is the field most directly affected by AI coding tools. Yet,

31:56hiring remains strong and overall US unemployment is still a healthy 4.3%. He also argued that AI labs have an incentive to make AI sound powerful enough to replace workers so that they can justify much higher pricing by anchoring the value of their solution to salaries rather than your typical SaaS company pricing. He actually expects we're going to have an AI job-a-palooza where we're going to see a ton more jobs created from AI rather than destroyed. Derek Thompson released a podcast titled

32:28The Smartest Case Against the AI Jobs Apocalypse. He argued that human desire and status-seeking are insatiable and in his view, even if AI automates many tasks, new categories of work will emerge wherever people are willing to pay for scarcity or status. And finally, kind of on the other side, Brookings researcher Molly Kinder, who we've talked about before, published an essay called The Messy Middle, where she argued the binary between today's intact labor market and a future post-AGI world of abundance, should we get to something like that, ignores the long stretch of disruption in between.

33:01She warned that AI may reverse the decades-long skill premium for knowledge workers with losses concentrated in the highest paid cognitive roles. Paul, curious what you think is going on here. We've got some really prominent people arguing, it seems like all in the same week, that we're not going to lose as many jobs due to AI as people, I guess the doomers we might call them, or just maybe realistic people might think. What's going on here? By last Wednesday, I was trying to figure out,

33:32like did an email go out to everyone saying counter the jobs narrative this week? It was just like everywhere. So, I don't know. I mean, they can say whatever they want, but it's not changing public sentiment. It's going in the opposite direction. So, the state of AI for business research that I just mentioned, we ask a question in there, and we've been doing this for six years now. We've been asking this question now for six years. What do you believe the net effect of AI will be on jobs over the next three years? This was actually the starkest finding in this year's data. 71% of respondents

34:02believe AI will eliminate more jobs than it creates in the next three years. Only 13% expect net job creation, and then 12% say they simply don't know. And then we noted in the report that what makes the finding so notable is the consistency. So, the belief that AI will be a net job eliminator does not vary meaningfully by roles. CEOs and VPs, 73%. Think, you know, eliminate more than it creates. Directors and managers, 71%, and specialists, 64%. So, it also doesn't vary by function. Marketing, engineering,

34:37sales, operations, they all converge within a few points of the 71% average. And then when we put it in context, this is where it gets really interesting, Mike. In 2023, when we asked this question, so this would have been right around the time GPT-4 came out. We would have had the survey in the field around that spring of 23. So, right after ChatGPT in 2022, 40% of respondents said AI would eliminate more jobs than it creates. The following year, and we asked the surveys in the field, again, roughly late spring

35:08every year. So, this is cyclically around the same time. The number was 47%. So, nothing crazy, is seven percentage point jump. That's about 20% or so. Last year, it was 53%. So, 2025, 53%. So, it's inching up six, seven points each year. And then this year, it jumps 18 percentage points or 34% it jumps up. So, that is a significant change in the sentiment and how the public is viewing this.

35:40And we asked more than 2,100 professionals across industries, geographies. And most of the people who are answering this, again, there's some bias in every data set, but ours is that they're more likely to be AI for professionals and leaders because they're taking our survey, which means they're seeing it through our podcast, they're hearing about it in our newsletter, things like that. So, these generally are more informed people who are using AI themselves and seeing what's possible and realizing this is significantly going to impact jobs. And then I think there's just the reality within companies. And I'm not sure, like the people that are saying this are

36:14incredibly educated and well-informed people whose opinions we listen to on a lot of issues. So, this is not like, you know, people who generally just like to say the opposite of what's actually happening. For the most part, there's definitely some in there who I think just say the opposite regardless. But everybody we talk to and every company I talk to and across every industry, flat headcount is a common goal. Like that's their best case scenario is they're trying to slow down hiring and they're trying to keep it flat while increasing revenue. The people who are trying

36:51to take this alternate position that it's going to be amazing, we're just going to create a ton of jobs, it always works out because historically it always has. The job displacement we're seeing is from over hiring is the thing they keep saying. And I just don't get it. Like I don't see how that is like a standard belief system that makes you just ignore the possibility that that isn't what's going to happen. So, on a positive note, Mike, the one thing we are seeing is more companies that are

37:22displacing workers, but saying, listen, we're going to hire, we're just going to hire AI forward employees moving forward. And the trick there is they're just not going to need as many of them. So, we had this happen with General Motors. TechCrunch reported this. General Motors has laid off more than 10% of its IT department or about 600 salaried employees in what they're calling a deliberate skills swap, clearing out workers who expertise no longer fits and making room for some with AI-focused backgrounds. So, I think this is something we're going to start hearing more about. They said the most

37:54sought-after capabilities are AI-native deployment, data engineering and analytics. This is sounding a lot like the FDE thing. Cloud-based engineering and agent and model development, prompt engineering, and new AI workflows. So, they're looking for the people who are AI-literate. Now, the thing that I think is going to be interesting, Mike, is all these people who are very, very confidently saying jobs aren't going anywhere, that's just going to all work out. My general feeling is one by one, they're going to come to the realization that they might have been giving false hope to people.

38:25This happened with Ken Griffin just this week. So, Ken Griffin, the CEO of Citadel, was a prominent AI skeptic, and he now says AI is real. So, this was Business Insider. We'll put a link in. So, many CEOs have been saying for years that AI can do the work of many white-collar professionals and remaking their companies accordingly. Griffin has been a notable holdout. At the beginning of the year, during a panel discussion at the World Economic Forum Davos, which we reported on at the time, the hedge fund billionaire said AI was impressive on the surface, but as soon as you

38:58dug deeper, it's all garbage. This is like three months ago. So, their article at the time, AI is not going to be a game changer for investment. So, this was the investment business. This is from May or March of 2025. He said, it saves some time. It's a productivity enhancement tool. It's nice. I don't think it's going to revolutionize most of what we do in finance. So, machine learning is going to come with a cost to society. So, even though he's saying that it's not going to impact their business, he was acknowledging that there was like a bigger picture here. So, he quote,

39:31machine learning is going to come with a cost to society, a cost that we need to understand. How do we help these people land on their feet so we don't end up with a backlash against AI and machine learning? So, then coming back to his comments at the Stanford business school this past week or earlier this month, he took this starkly different view. He said, I got to tell you, I went home one Friday actually fairly depressed. You could just see how this was going to have such a dramatic impact on society. Again, this is three months later. Griffin said AI had become, quote, profoundly more powerful than it was nine months ago and had allowed the hedge fund

40:08to unleash a wider range of use cases for the technology. He said, for the first time, AI is real. And then he said, quote, to be blunt, work that we would usually do with people with masters and PhDs in finance over the course of weeks or months is being done by AI agents over the course of hours or days. He emphasized that this goes beyond what he called mid-tier white-collar jobs that are now being automated with agentic AI. So, my argument, Mike, summarized, and I've made this

40:39argument now for multiple years, but again, I know we have new listeners to the podcast all the time, so I'm going to kind of highlight our view on this. Once you have a deeper understanding of the full capabilities of today's AI models, including the agentic capabilities, and a general concept of how these models are going to improve over the next one to two years, your view of the future of work and your business changes. Specifically, as I said, this is applying to the increasing autonomy and reliability of agents. That's what's happening to Ken Griffin. He's coming to this realization of

41:12this isn't the same technology that we had nine months ago. He's now seeing what you should have been seeing nine months ago, which is where these agents were going. And what if you stop now and look ahead and say, oh my gosh, where are we going to be nine months from now? I don't understand this argument that we won't have fewer jobs. Like, I really just struggle tremendously with this. So, if you aren't growing. So, I think what happens oftentimes is they look at Anthropic and OpenAI and Salesforce. They look at these companies like, well, they're hiring. They're hiring developers nonstop. They're hiring FTEs. Yes, and they're also growing 40 to 60 percent a year.

41:45So, like, of course they're hiring. If you aren't growing, or if you are in a business that's growing, let's say, less than 10 percent a year, if you don't need fewer people, then you're not properly applying AI. Like, you just don't need as many people to do the same level of output or the same amount of revenue. So, employees who are AI forward and able to drive efficiency and productivity gains are going to reduce the total number of people needed to do the same amount of work. That's what General Motors is doing. We're going to get rid of these IT people, the 600 of them, who don't like

42:19AI or aren't good at it, and we're going to hire 200 back who are like AI forward, and they're going to do the work of 2,000. Like, that's the premise here. So, if you are in big tech and you're growing 20 percent, or if you're outside of big tech and you're growing 20, 30, 40 percent, then yes, you are probably still hiring, and you can get this feeling like everybody should be hiring because you live in a world where growth happens. Right, right. But if you work for a private equity-owned company, a VC-backed company, or a public company,

42:49the pressure to increase revenue per employee is going to be massive. So, if you just look at a simple metric of how much revenue does each employee in our company generate, everybody is going to pressure that KPI. And there's two ways to improve it. You reduce your costs largely by reducing payroll, or you increase revenue with the same headcount. That's it. So, I don't, again, I want to be wrong on this. I really want to look out three years from now and saying, nope, they were right. Like,

43:20jobs just kept growing. It was awesome. Everything worked out great. But I also don't understand the premise that a responsible leader of any organization wouldn't plan for the alternative. That maybe there is this chance that we do go through a one to three to five-year period where we're just going to need fewer people and it's going to cause displacement. It's going to cause underemployment across the economy. And yet, they seem to all have this unshakable confidence that they're right. And so, I guess that's my big thing is like, we need AI to drive innovation and growth. We need to be

43:53realistic about this and human-centered in the fact that we should be preparing for the possibility that there's at least displacement, that college students are going to have a hard time getting jobs, that highly paid people like PhDs and people with masters aren't going to be needed as much in companies like Citadel. And I don't understand the premise of pretending like that's not a possibility. It just seems like a disservice to humanity to just not at least say, maybe. Maybe one to three to five years is going to be kind of messy. Maybe 10 is going to be amazing. 10 years out, it's all going to work

44:25out. But that's easy to say when you're making a half a million a year or five million, whatever, and it doesn't affect you. But I just feel like, and we're going to talk about it in the next topic, there's a hell of a lot more people who don't feel that way. Our own research says 71% of jump, 34% increase in people who think it's going to eliminate more jobs in the next one to three years. Um, that's a major problem that these people are basically just glossing over with a future of

44:55abundance. And historically, it's just always worked out. I don't understand that premise. Yeah, I was going to mention to that last point, this just strikes me as such a tone deaf argument, because like, people are already upset right now. And I'm no economist, and I agree with you. This is the number one issue that keeps me up at night. I genuinely hope I'm proven so wrong. Like it would be my most joyful outcome would be like five years now, I'm being like, I worried so much about that. And it was not a thing to worry about. Great. I you can poke fun at me for the rest of my life for being wrong on this. But like, it's really hard to hear these

45:30arguments of like, well, you know, people get displaced and move into other jobs like, well, let's talk about the nature of that displacement. Are you going to move into another white collar job that pays you the same amount or more in some cases, perhaps, but probably not? What happens when you have a nation of people that used to be doing okay with white collar jobs and now have to drive DoorDash or Uber or something like you're going to have a lot of angry people. And as we're going to talk about, they're already really upset about this issue. Again, they keep coming back to,

46:01but look at software developers, look at the need for it. It's like, okay, well, the out of work marketing manager isn't becoming an FDE. Like, I'm sorry. Right. And like, and again, like we're doing our best to hire as many people as we can at SmarterX. I'm not hiring people who aren't AI forward ever again. Like it's, and that's just the reality. And if I talk to other companies, it's what I advise them. You cannot hire people who aren't AI forward. It's a disservice to your

46:31organization and you have a responsibility as leaders to do what's right for the organization. So you have to offer the ability to reskill and upskill people. But if they choose not to learn the technology, there's not a hell of a lot more you're going to be able to do as a leader. But I don't see job opportunities for people moving forward that don't buy into the technology, which as we're about to talk about is going to become a major problem for the next generation entering the workforce this year. All right, Paul. So before we dive into that and all of

47:06our other rapid fires, just one other announcement. This episode is also brought to you by Macon, our marketing AI conference that we run every year. This year's is happening October 13th to the 15th. Here in our home base of Cleveland, Ohio, we are super excited to announce that Dan Slagan, SVP of marketing at Zapier has joined the 2026 speaker lineup. He leads growth for one of the world's most connected AI orchestration platforms. Previously, Dan was at CMO at Tomorrow.io.

47:36He's been recognized by Forbes as a top 50 entrepreneurial CMO. His work has been featured all over the place with what he was doing at Tomorrow.io. He has previously spoken at Macon. We can't wait to have him back on stage. And what's really cool is Dan joins a 2026 lineup that Paul is just shaping up to be incredible. We've got Karen Howe, the award-winning AI journalist and author of Empire of AI. Andrew Yang, former presidential candidate and tech and economic futurist. Of course, Paul, you yourself are speaking at Macon. Plus, we've got a ton more speakers

48:09ready to be announced. So this is three days of workshops, keynotes and sessions and conversations built specifically for marketing and business leaders who are actively figuring out how to adopt, operationalize and scale AI across their organizations. So one important note here, if you are thinking about joining us at Macon, this week and next week are a really good time to do that. Ticket prices go up on May 30th. If you register before then, you'll get the best pricing available. Also, you can use the code POD100 at checkout and save an additional $100 on top of

48:40the current rate. So go visit macon.ai to register. Okay, Paul, so let's dive into this first rapid fire we've been alluding to throughout our conversation so far. So an Axios piece this past week was published called, quote, an AI hate wave is here. And it captures what is now starting to look like sustained backlash against AI in the US. And they actually write in their opening here, if AI were a candidate

49:12for political office, it would be losing in a landslide because they cite a bunch of recent data that does not look great for sentiment about AI. So a recent Gallup survey found only 18% of young people ages 14 to 29 feel hopeful about AI. An economist slash YouGov poll released this past week found that over 70% of Americans think AI is advancing too quickly with the figure consistent across parties. 68% are Republicans, 77% Democrats. YouGov's tracking shows negative views of AI have risen

49:44from 34% three years ago to just over 50% today. Axios opened their piece as well with a Florida commencement address earlier this month that went a bit viral because real estate executive Gloria Caulfield sparked a chorus of boos when she told graduates that AI is the next industrial revolution. The backlash is starting to have real effects, too. There's a record number of data centers that were canceled in the first quarter of 2026 amid community resistance. Morgan Stanley analysts wrote

50:16that public pushback is emerging as a binding constraint when we're thinking about investments, particularly around data center build out. And Jeffrey has told clients these setbacks are sapping investor confidence. Paul, let's dive into this. This really does seem like sentiment is going not only the wrong direction, but is going to stay there for a bit. This doesn't seem like a passing fad anymore. As you were going through this, I had this thought. The students graduate right now would have started in 2022, right?

50:48Yeah. 23, 24, 25? Yeah. So, they went through college with at minimum chat GPT. Like, they're, you know, came into being in their freshman year of college, basically. High school students today won't know education without it. I mean, they've gone through high school, they'll go through college with it. And one of the problems we faced in the early days of generative AI, and even today in many schools, is these students have been told it's cheating. They've been raised, raised, I'm going to use

51:21loosely over the last three or four years, that AI is cheating. They're being taught by professors who want nothing to do with it. There was a lack of effort at a higher education level to move with urgency to embrace it in a responsible way. And we just like outlawed it or called it plagiarism, whatever. So, part of this issue, Mike, again, I'm kind of thinking out loud here is, if you're told something is bad long enough, like, you come to believe it's bad.

51:51Right. Um, I don't, I'm actually trying to think like, where else would the negative sentiment have come from? Because up until the last six months, nobody gave a shit about this politically. Like, it wasn't, it wasn't like this was like some big effort by one side or the other to vilify it. I mean, certainly we're heading in that direction now, but I don't know. Like, I'd be really fascinated actually to like, dig into why the hate exists, I guess, or why the distrust exists. But it was the first thing that came to my mind is it's just,

52:24that's what they've been ingrained in them. I would imagine the school piece is significant. I could imagine, I've no way to prove this, that as certain things like data centers have become more prominent, that there's probably a solid ecosystem of commentary online that perhaps people are seeing videos on TikTok and things that are blowing up around these kind of polarizing issues. But to your point, that wasn't the case a year ago. Yeah, someone's funding campaigns to make it. Yeah, I wouldn't surprise me.

52:54Yeah. Yeah. So to bring this point home, we'll put a link in the show notes for this one. But you, you got to watch the six and a half minutes or so of Google's former CEO, Eric Schmidt, giving the commencement speech at the University of Arizona last week. I'll read a couple of quick excerpts here from a CNBC article. It says, Schmidt, who led Google for a decade, opened his remarks by reflecting on his own student years and the rise of the computer of a device named Time Magazine's Person of the Year in 1982, back in his day. He traced its evolution into the laptop and smartphone and

53:29its proliferation through the internet and social media. While the computer connected people, democratized knowledge, and lifted many out of poverty, it also carried a darker side, Schmidt said. The same platforms that gave everyone a voice like you're using now, as they were booing him, also degraded the public square, he said. They rewarded outrage. They amplified our worst instincts. They coarsen the way we speak to each other and that way, and in the way that we treat each other, is in the essence of a society. Schmidt then drew a parallel between AI and the transformative

54:02impact of the computer and was immediately met with boos. And they are, if you listen to the clip, it is awkwardly painful. Like the boos got louder and louder as he kept talking. So he said, I know what many of you are feeling about that. I can hear you. There is a fear. There is a fear in your generation that the future has already been written, that the machines are coming for the jobs. Schmidt's reception was not an isolated incident. Earlier this month, real estate executive Gloria Caulfield was similarly booed at a commencement speech at the University of Central Florida

54:35after mentioning the controversial technology. The rise of artificial intelligence is the next industrial revolution, she said, as the crowd erupted in booze. So again, when you watch the Schmidt talk, it is like, honest to God, it was like a dystopian feel. Like I found myself having a flood of thoughts about it. I've mentioned on previous episodes that the industry has a major PR problem. Like they have for too long

55:06ignored their own impact on jobs and impact on society with this whole future of abundance thing and just assumed it was going to just work out. And I think they're now realizing that society doesn't work that way. Even when Schmidt through the booze transitioned, I thought he was just going to walk off stage at one point, but he transitioned to the good it's doing, how it's going to solve cancer. It's going to do all these things. They booed louder when he tried to highlight the good stuff. And I was like, oh my God, like this is maybe a larger problem than I was expecting it to become.

55:43And then I had this moment where, well, I guess I'll play it out for a second.

55:50If you're a politician and you want to win, we are about to see a tripling down of those people on the fear mongering. Like we are going to hear so much negativity around data centers and job loss and all this stuff because these booze opened the floodgates for that messaging to work. And that is a very dangerous thing to do to shift power, I guess. The societal backlash is very real, though. And the thing that really worries me, I mean, there's a lot of this that bothers me.

56:21But the thing that I really was worrying about Mike as I was listening to those boos was, you aren't going to find a job. Like if you are sitting there booing him, you can boo Eric Schmidt, you can hate Google, you can hate Eric Schmidt, I don't, whatever. Like I'm not standing up for him or like, you know, whatever he's done. I don't know the guy personally. I don't think they're booing him though. I think they're booing what he stands for as a leader of an industry that builds this technology that they feel threatened by. Yeah. But if you're in that crowd or any commencement

56:56crowd or you're having these feelings as a 21 year old, a 22 year old, I am sorry, but you're not getting a job. Like you can have those feelings towards AI and I empathize with them a hundred percent. Like I understand the fear, the anxiety, even the hatred towards it. The tech isn't going to stop. It's not going to go away. And unless you find ways to embrace it and go work for companies that are doing it in a responsible way. But if you show up to an interview and you say you hate AI,

57:31they're not going to show you the door fast enough. No. And so that is a really, I don't know how we solve this. Like it's going to be hard enough to find jobs for these, the students graduating right now. If you layer on top of that, a hatred for AI and a refusal to learn how to use it in a responsible way, you have no job prospects, none. Like it, this just doesn't work. And this is like, now I'm like, I'm concerned about jobs on a totally different level after the

58:02last 24 hours. The more I've kind of sat on this and watched that video of what happens to these students who just resist the change that is inevitable. Your, your booing isn't going to stop AI from changing work. And I don't, I don't know what to do with that. And Mike, I really don't. I'm just like, I'm, I'm really conflicted right now. I'm like struggling with this one. You know, what's also really disturbing to me to consider is look, I get, again, couldn't echo more what you just said, but there's the kids booing. And then there's the kids who are friends with

58:36the kids booing who might on their own be curious about AI. But I would imagine if the boos are this bad, there's a lot of peer pressure where I would imagine among your friend group, your cohorts, it's probably not a good idea to be talking about this stuff. And I think that's such a shame where like there might be people who are put off on experimenting or exploring AI because of the sentiment, even if they don't agree with the boos, which I would just encourage you, if you feel like you're in that spot or have a kid who is like, try to work against that.

59:07Yeah. We've, we've had these conversations internally. So again, you have to keep in mind, like at SmarterX, we are not an AI model company. We're not building this technology. We're trying to teach how to use it in a human set of responsible way. And we have humans within our company who are scared about the future too. Like these are real conversations that we have all the time where you know, either we have an employee who's like, I am, I'm actually terrified of what's going to do to jobs. And we like, okay, that's why we're doing what we're doing. You have employees we talk to, like coworkers of ours, Mike, whose friends don't like the fact that they work in AI, who judge them

59:42based on that and, and like hate it. And so we are, we are living in this world. We are not living in the Silicon Valley tech bubble of like all progress is amazing and abundance is, you know, guaranteed. And it's just going to work out Kumbaya. We're living in the world where we don't necessarily all feel super positive about the future and where we have friends and family members who don't know why we're in AI. Cause they don't understand what we're trying to do in AI.

1:00:15Um, it's, it's a very weird time and I, I don't, I, this is definitely one where I just don't have the answers, but I, it weighs on me. Yeah. Like every day. Yeah. I couldn't agree more. I think we're, this is not the last time, unfortunately, we're going to be exploring the implications of this. No. Okay. Next up this past week, Anthropic released a paper called 2028, two scenarios for global AI leadership. And in it, they laid out the view, their view of the U S China

1:00:45AI race and what policymakers need to be doing right now. Now, this was interesting timing because this dropped right in the middle of president Trump's visit to Beijing. So in the paper, in the first scenario, they outline a pathway where America defends its compute advantage. Policymakers tighten chip export controls. They disrupt China's distillation attacks on us models and accelerate democratic AI adoption. Anthropic projects that this locks into a 12 to 24 month U S lead in AI by

1:01:162028. In the second scenario, the U S fails to act. The Chinese communist party catches up to near frontier intelligence, deploys subsidized AI globally, and authoritarian regimes shape the rules and norms of the technology. Now, Anthropic argues the window here to pick a path or get on a path is closing fast. The paper says there is a high likelihood that we will look back on 2026 as the breakaway opportunity for American AI. And they pointed to their own mythos model as a wake up call for the

1:01:48acceleration ahead. Now, meanwhile, in Beijing, the summit between President Trump and Xi, the leader of China produced no signed AI agreement. Treasury Secretary Scott Besant told CNBC, the two AI superpowers will start talking and set up a protocol in terms of how do we go forward with best practices for AI to make sure non-state actors don't get a hold of these models. Trump told reporters aboard Air Force One, the two sides discussed possibly working together for guardrails. Interestingly, NVIDIA CEO Jensen Huang was

1:02:20initially left off the CEO delegation that accompanied Trump in China to avoid some awkward optics on chip controls. After the snub leaked to the press, Trump personally called him and Huang boarded Air Force One at an Alaska refueling stop. So Paul, it's interesting Anthropic releasing this while Trump is in China. There is, it sounds like from Anthropic's view, a window that is closing for the US to establish American AI dominance. What did you make of all this? I am nowhere near enough of an expert on geopolitics to like comment deeply on this one. I always

1:02:56suggest AI superpowers by Kai-Fu Lee if you want to understand US and China relationships. Chip War is another great one if you want to understand the significance of Taiwan and, you know, why the conversations are centering around China's ambitions with Taiwan. The, I guess all I'll say from just from observing and studying this stuff through the years is whatever they would agree on to work on together with AI safety, I think is probably, probably not worth the paper it's printed

1:03:34on. Yeah, right, right. They're so deeply embedded in like espionage and stuff with each other and I think Trump even boasted at one point to Xi, whatever you you're doing to us, just know we're doing it way worse to you or I don't know what the exact words were, but yeah, like, yeah, you think you've got us, wait, wait, you should know what we've done to you kind of thing. And so that stuff isn't stopping, like they're not gonna just all of a sudden be like, all right, let's just work together. This is amazing. I figured it all out. It's like, no, they need each other at a high

1:04:07level for lots of different things. But, you know, if you think that they're gonna stop trying to steal the weights of the mythos model and things like that, like, come on. So, whatever. It's a trade trip. It's trying to open up, you know, money and trying to demonstrate power and things like that. I don't, I just don't have high hopes for it personally. And I think Taiwan, again, I don't want to get into like super hot button political issues, but if you don't know why Taiwan is such

1:04:43a sticking point, you should learn it. And they said, Chip War is a great book to understand it, at least at a technological perspective. But I think it's safe to assume that it's gonna become a much more mainstream story in the not too distant future based on things the US is doing in other countries and what China's ambitions are with Taiwan. I just, I feel like it's probably an

1:05:13important topic for Americans to be educated on. I'll just kind of leave it at that for right now. So, next up, some more politically related news. Axios reported this past week that 32 House lawmakers in a bipartisan letter led by Republican Bob Latta urged the White House to act on AI cybersecurity threats. So, one of the triggers here was Anthropik's mythos model has already identified thousands of high severity zero-day vulnerabilities across major operating systems and web browsers, which we've talked about. These include flaws that survived years of human reviews.

1:05:48Palo Alto Networks, one of the few companies with early access to something like Mythos, told Axios it found 75 vulnerabilities in its own products in the past month. That is seven times its normal rate. And Palo Alto estimates organizations have just three to five months before attackers gain broad access to these capabilities. Separately, Google's Threat Intelligence Group reported the first known case of cyber criminals using AI to develop a zero-day exploit in the wild.

1:06:18Google's chief analyst said the AI vulnerability race has already begun and that for every zero day, we can trace back to AI. There are probably many more out there. Their broader report also detailed Russia, North Korea, and Beijing-backed hackers using AI to scale up cyber attacks. Alongside all of this, OpenAI announced Daybreak, its broader cybersecurity initiative, which has cyber-capable versions of GPT 5.5 being provided to launch partners that include Cloudflare, Cisco, CrowdStrike, and Palo Alto

1:06:52Networks. On top of all this, the Trump administration has been weighing executive action on frontier model cybersecurity, though Axios reports that process has been delayed by internal disagreements and the run-up to Trump's China trip. So, Paul, we've talked about this before, but just especially with the Google research coming out, are companies anywhere close to ready for what is about to hit them from a cybersecurity perspective? It doesn't seem like it. It really doesn't. No, I mean, if you're not going

1:07:23to be a forward-to-point engineer, then be a cybersecurity expert. Yeah, right. Those jobs are going to be in really, really high demand. Might be stressful, but probably in demand. Yeah. So, if you're that out-of-work marketing manager, you know, maybe level up on the cybersecurity side of things, I don't know. I've said it before in the pocket, this part terrifies me. Like, whatever the most advanced models today are doing, six, nine months from now, the open source models will be doing. And based on the reports we're getting, which is just the stuff that's

1:07:55leaking out of this. Yes. Yeah. Like, this isn't even the worst of it. It is, these models are going to be prolific at causing major problems for companies. And I don't know what the solutions are. Google has their IO conference this week. Yeah. And when I was at Google Next, they talked a lot about security and governance. So, I think you're going to hear a ton about it from these big companies. I just don't know that they're going to be able to move fast enough. But cybersecurity has always been that way, where

1:08:28it's just like you're trying to racing to sort of stay ahead of the bad guys. Yeah. So, man, this is not like the most uplifting episode, to be honest. So far, it is not. Like, we've got a couple of things that aren't like directly gloomy in here, I promise. But yeah. I'm going to grow up in a ball after we're done with this episode for a little while. Yeah. This one's hitting all the usual suspects of things that scare the hell out of me. Yeah. Really? And then the next one, I'm looking at headless. It's like, oh, wow. Yeah. I know what it means. But just like immediately, I'm like, this is just like

1:09:00continuing. Yeah. Don't worry. This one's not as morbid as it sounds. So, what we're talking about here is last month, Salesforce launched what people are calling a headless product. So, basically, opening its APIs and betting that in an agentic world, its value as a business lies in the data layer rather than the UI. So, this past week, we actually got a piece from Andreessen Horowitz that used that announcement to kind of ask a broader question here, which is, when agents replace humans as the primary users of business software, what is actually defensible?

1:09:36A16Z's argument is that the user interface is actually doing far more work than it got credit for. So, the interface enforced data hygiene, created shared vocabulary like leads, opportunities, and accounts, and built muscle memory across thousands of users. That stickiness, they claim, is why something like Salesforce gets brought from job to job. So, agents, however, do not need a UI. They read and write directly to the underlying data. So, Andreessen Horowitz argues, defensibility in this environment shifts in two directions. So, some of the old moats stay,

1:10:10the operational logic, compliance critical data like payroll, connectivity across siloed systems, all stay very defensible and important. But there are several factors that become newly important for AI-native startups. So, for instance, proprietary data the product uniquely generates, owning the full action loop from decision to outcome, the network effects across multiple parties, and how all this is executed in the real world. So, their bottom line is the next generation of systems of record will likely not just be databases of human entered data. They will be

1:10:44agentic systems that capture context, initiate actions, and produce their own data exhaust. The most interesting ones will actually extend into real world execution or the mediation of multi-party workflows. So, Paul, it's a little bit kind of in the weeds and technical here on like the insider baseball of like software companies and moats. But it is an interesting overall trend that when agents start using software, what changes and what do we need to be thinking about as business leaders?

1:11:15I'll just try and like I'm processing this as you're reading this one, Mike. I'll use HubSpot as an example. So, HubSpot is our CRM. It powers a lot. Our marketing, our sales, our customer success, operations to a degree. And increasingly, you're starting to say, okay, well, if Gemini is connected to HubSpot or Claude is connected to HubSpot or ChatGPT or like whatever the interface is, because those companies don't want you to leave their interface. Like their AI companies want you to live within whatever the UI is going to be that they're going to create. It might be the chat like

1:11:49we see today, or it might be some new way to interact with information. But their whole premise is like whatever you want is just living there. And whether you want to run a workflow, say, hey, launch a campaign for me, send an email to this. Like you're just talking to your AI assistant all day long. And it just goes and does stuff. And so, the premise here is I don't have to log into HubSpot anymore. Maybe every day like Claude just pulls my HubSpot dashboard and shows me the 10 KPIs that matter and tells me what to do about them and then writes my customer emails and does my sales. Like it just all happens right within my Claude interface as an example. And I never go to

1:12:24HubSpot anymore. And then like, let's say we do that for 20-some employees. Like, hey, everybody, like all your workflows, everything you need, it's just going to live right within Claude. Just log in every morning and boom, like it just does it all for you. And all of a sudden, like they don't have to go to HubSpot anymore and do things. Like they just do it all right through there. So that if you're a software company, it's like, well, what is our purpose then? And so, the premise here, if I'm understanding correctly, is like, well, where the data lives. It's like, we are. So then the question becomes like, okay, but if you're where the data lives,

1:12:57because historically that's where the data lived. But if moving forward, there's new interfaces that exist where the data starts to live, then do I need that system that was built for humans to work within? I don't have an answer. I'm just like, I'm sort of like regurgitating here how I'm synthesizing the current situation. It's a very tricky place for these software companies to be. And it is why there's so much uncertainty around the future of software. And they're all trying to kind of figure

1:13:28this out. But I do think that it's reasonable to have this assumption that at some point, it's going to be agents that are going into these software programs and extracting the data more than it's going to be humans. It's kind of like, I assume we're going to have more agents coming to our website than humans at some point, based on what industry you're in. And the agents are going to do the work for the people. And so this is a really intriguing thing to follow along. Which is why for a Salesforce and or a HubSpot, if they can somehow figure out how to get the

1:13:58ownership of the agents on their platform, and your ability to build them be on their platform, that's interesting. I'm not sure if that will be the case, if that will happen. But it seems like where they're headed. Because otherwise, yeah, you're just a database or a data repository that someone else's agents are accessing. Yeah. And some of this, I feel like it's just above my pay grade to comprehend, or I don't live in this world every day and think about this challenge. I observe what all these software companies are doing. All I know is Wall Street's not convinced yet. The software stocks aren't

1:14:33throttling up like the way the AI companies are. And so I think that there's naturally just a lot of unknown about, okay, well, maybe what they're doing is going to work. Maybe this switch plays out, and it ends up being good. But yeah, I think most of it is just uncertainty right now. Yeah. Next up, this past week, Publicis Group agreed to acquire LiveRamp for $2.2 billion and an all-cash deal. The French advertising holding company announced the agreement on Sunday. And Publicis

1:15:06is betting the acquisition positions it as a leader in what the company is calling agentic transformation, meaning the use of AI agents to automate and collaborate on corporate workflows. Publicis estimates the agentic transformation opportunity at roughly $1 trillion. LiveRamp specializes in what Adweek describes as data collaboration. It basically lets different companies share and build new data sets and data models together. Publicis is framing those capabilities as a foundation that can power agentic AI frameworks. Publicis Group Chairman and CEO

1:15:38Arthur Sedun told Adweek the deal is core to that strategy. So Paul, it really seems like Publicis just dropped a couple billion dollars to position itself for the agentic transformation market. Did this make sense from your perspective as an ad holding company kind of positioning itself to win in agentic AI? Yeah, it's going to be an interesting partnership acquisition. I actually just, I did a keynote for LiveRamp, I think it was like two months ago, I got a chance to spend time with their

1:16:08CEO. You know, I was really impressed with the organization. But yeah, they do like data collaboration where, you know, brands can kind of input their data anonymously and then you can kind of share and build off of that data to do better targeting, things like that. So it's a, I mean, I guess it's a pretty logical play for Publicis because a lot of, you know, they probably have a lot of mutual customers. And so being able to kind of go deeper into the data side of that and the agentic side makes a ton of sense. All right, next up, we've got three different stories that surfaced this past week that when

1:16:39you kind of put them together, point to some interesting ways that AI is beginning to change how we're all starting to do our work. So first, go to market leader, Kieran Flanagan published a long post on the, what he calls the AI second brain he built before taking over a 400 person team. So this system runs on a piece of software called Obsidian and Cloud Code. It basically has hooks that load all of his notes, his strategic context at the start of every session and write structured

1:17:10summaries back to his file. So he could basically query this kind of ongoing second brain of all his information, notes, what's been happening in his day and in his work. Second, we got a post from Shopify CEO Toby Lutke that shares how his company built an AI agent called River that lives in public Slack channels. So in the past 30 days, just about 6,000 Shopify employees worked with River across almost 4,500 channels. And about one in eight pull requests merged into Shopify's code base last week

1:17:41were authored by River. So River only works in public, never in DMs. So the whole company can essentially use this through Slack and watch and learn what everyone else is doing. Also interestingly, AI researcher Andre Karpathy posted this interesting viral tip. At the end of any LLM query, go ask the model to structure its response as HTML and view the file in your browser. Karpathy argued markdown is today's default for AI output, but HTML is the better next step because vision is how the human

1:18:13brain prefers to receive information. So Paul, the reason we're kind of talking about all these, we've got one go-to-market leader building essentially an AI second brain for his job, his function, where he can have this instantly queryable and conversational database about basically everything going on. Shopify is doing an interesting version of something in public where employees can essentially query and use a centralized agent. And then I just thought the HTML thing was interesting because I know you and I both have often used Claude to output HTML as a better way to kind of

1:18:46visualize and review output. So all of this is really just pointing to like there's these weird interesting nuances that are changing how we work because of what AI seems to enable. Yeah. The HTML trick is like a key on lock. I love it because you can make it interactive and everything. Yeah. So that's huge. Yeah. So I've alluded to this a couple of times, but I think Mike, I'll just like explain what I've been working on real quick because I think it's super relevant here. So about two months ago, I had an interesting day. I had a breakfast with a higher education leader,

1:19:23a top leader at a major university. And then that same day I went and visited NASA. So we're lucky in Cleveland to have NASA Glenn in our backyard about 10 minutes from my house right next to the airport. And as I was touring different labs within NASA, there was one I went in where they work on the wheels for the rovers that you send to the moon and Mars. And as I was getting the tour, a scientist walked by and somebody was following him and I got introduced and the guy following him was an apprentice.

1:19:57And so I'd spent the morning trying to talk to a higher education leader about what the future of jobs look like and what their graduating students might be doing when they got out into school, out of school. And then I go to NASA and I'm seeing this idea of this apprentice. And so I left NASA, and I stopped on my way home and I was getting something to eat and was waiting for the food to take up for my family. And I started thinking about this apprentice idea. And so I actually went into ChatGPT and I was like, let's talk about the history of the apprentice. And like, what is the

1:20:27origin of that role? Where does it come from? What are the possible applications? And I'm just like, just having a conversation. And so I started landing on this idea and I messaged, I think, Mike and some others at the company, like in the moment. And I was like, I think I have the idea for my Macon talk this year. And so this idea of like the organizational structure is something I've been thinking deeply about for a while. We did that AI talent report earlier this year, Mike, where we kind of shared some of the findings from our council, or maybe it was the year of last year. It's all running together. Yeah, it was beginning of this year, actually. It seems like a million years ago.

1:20:58It really does. So I'm just going to read something because again, I alluded to this on an event last week. And so I've kind of like publicly said I was doing this. I'm just going to explain what I'm doing. And I'll leave it at that. We'll come back to this later on. I don't think this is publicly on the Macon site yet, but it's going to get announced soon enough. So I'll just explain it. So here is the premise behind my keynote for Macon. The title is The Architect, The Orchestrator, and The Apprentice Rethinking Work in the Age of AI and Agents. As AI and autonomous agents transform

1:21:32the workplace, they are doing far more than automating tasks. They are redefining expertise, reshaping organizational structures, and changing how professionals learn, contribute, and advance. In the process, they are giving rise to three defining roles that may shape the future of knowledge work, the architect, the orchestrator, and the apprentice. The architect designs the system. This role defines the workflows, governance, team structures, and human machine operating models that make AI useful, scalable, and aligned with business goals. The orchestrator runs the system.

1:22:03This role frames problems, directs AI agents and human collaborators, evaluates outputs, and turns distributed intelligence into decisions and actions. The apprentice learns within the system. This role develops judgment, context, and discernment in a world where much of the entry of the work people once learned from is being absorbed by AI. For decades, organizations relied on a familiar model of professional development. Junior employees handled research, drafting, analysis, coordination, and execution. Over time, they built the experience and judgment required to lead.

1:22:37But as AI systems take on more of that foundational work, the traditional ladder from novice to expert begins to break. If machines do the work people once learned from, how will organizations develop future experts, managers, and executives? In the architect, the orchestrator, and the apprentice, I'm going to try and present a new framework for understanding this next era of work. The future of competitive advantage will not belong simply to the companies that adopt AI tools fastest, but to those that redesign work management and talent development around these roles. We're going to look at historical traditional

1:23:10hierarchies, why judgment becomes more valuable as execution becomes more automated, how middle management must evolve from supervising production to scaling expertise, and why apprentice may become more important, not less in the age of intelligent machines. So the way I'm basically thinking about the future of teams and pods and orgs, like everyone think about it is like, you have the leaders, obviously at the top, you have these architects or planners that are strategists, you have builders who are like develop the apps and agents, which is why we created SmarterX Labs. So that that's actually

1:23:41like where this labs concept came from. Then you have the orchestrators who are the human and agent project managers, basically the apprentice, and then you have agents. So while I'm focusing on architects, orchestrators, and apprentices, when you layer in the builders and the agents, you actually have the forming of an organizational structure. So it's kind of like a working hypothesis, I would say. I'm like observing all of these different models that we're hearing about from Dorsey and others and trying to like for how that aligns with the direction I'm thinking. I'm having conversations with leaders who are like working on this within major enterprises and bouncing ideas around. So

1:24:15yeah, like if you're working on something along these lines, if you're thinking about this at an organization, reach out to me on LinkedIn, like I'd love to like compare notes and, you know, just chat a little bit, because I want to do a discovery process behind this. I don't want to just like write this and say, here you go. So yeah, reach out if you've got something interesting to talk about on this. I love that. Can't wait to hear more. All right. So next up, we have our AI use case spotlight where every week we give you a quick look under the hood at the real AI use cases we are exploring,

1:24:48building or deploying in our own work at SmarterX. So Paul, I'm going to share a quick use case. Then if you have anything to share, we can talk through that too. Sounds good. So for me, I'm going to actually just read a couple excerpts out of a message I sent to our team, because as we've talked about several times, we launched this past week, our state of AI for business report. So kudos to our director of research, Taylor Rady, for taking the lead on that. As part of that, me and Taylor actually sat down to build out how we activate this report

1:25:18internally. So I'm going to share a little bit of what I shared with the team about how we're using AI to do that. So I started the message saying a testament to the velocity at which we can now move thanks to AI. Taylor Rady and I met today to finalize our plans for activating the state of AI for business report, both internally and externally. These plans also serve as the template for how we will activate all our research moving forward. So knowing what AI was capable of, we expressly set out not just to plan all this out, but solve as much of it as possible during a single

1:25:49meeting. So we had about two and a half hours blocked off for this, and we used the following playbook. First, the first half of the meeting was spent on a deep strategy session architecting what, you know, what we might call an activation machine could look like. Every bit of that was recorded. The recording, number two, was then fed into Claude along with the context into what we were trying to build. Number three, while that processed, we listed out and gathered all the context documents we'd need to build the activation machine we envisioned. And number four, we spent the second half of the

1:26:22meeting actually building and iterating directly in Claude. The result was that in that time, we had a working prototype of a Claude project that takes any piece of research and in minutes generates an internal activation brief that presents customized data, takeaways, and next actions for each individual team at SmarterX so they can go make optimal use of the research for things like marketing, sales, etc. So we still have a lot of iteration to do to fully get this right, but it's pretty incredible. We're just blown away by when you really set out with that mindset and that approach,

1:26:54you can make so much progress in building things and just getting them solved and done in a single meeting. Yeah, this was super impressive. This is one of those where I just wanted to call Mike and be like, okay, explain exactly what's happening here, how this is working, what are the implications of the other workflows we have. So yeah, it was really cool. And you and Taylor are just doing awesome stuff from a research perspective and content creation perspective. Mine was actually a pretty cool one, super simple though. I'll put the prompt in the show notes. A few weeks back, Anthropic shared a

1:27:29prompt you can use to extract memories and personal context from another platform. So if you wanted to shift from OpenAI over to Anthropic, I think they did this when OpenAI was getting in a mess for something. I forget what they had done. Yeah, I forget what prompted it, but they did this pretty recently. Yeah, there was some backlash to OpenAI. And so Anthropic was like, hey, if you want to switch over, here's a prompt. And so I was like, wait a second. So I have a Claude account that I've been using for about a year. And I wanted to switch because we got Claude for all SmarterX employees

1:28:04now. So at SmarterX, we have ChatGPT, Gemini, and Claude. Everybody gets all of them. And so I was not using the SmarterX Claude account, but I didn't want to have to maintain two work accounts. So I was like, oh, let me try this. So I go grab Anthropic's prompt. I put it into the historical Claude account I had. And then I took the output and put it into the new Claude to train it on all the history. And it was insane. It was so good. That's awesome. And the prompt is no more than, I don't know, 250 words. It literally says, export all of my stored memories in any context you've

1:28:39learned about me from past conversations, preserve my words verbatim where possible, especially for instruction and preferences. And then it gives categories, output in this order, instructions, identity, career, projects, preferences, and then like how to organize it and things like that and what the output it in. So I just, I dropped that in to my old Claude account. And about a minute later, I get like the output and it's, it's way more than 250 words, but it's, it was really good. And so I just

1:29:10copied that and I put it into like the system instructions or the memory within the new Claude and boom. And I just picked up where I left off and started working on projects in the new one. So pretty cool little trick. Again, sometimes knowing how to prompt is, is still a super valuable skill. Yeah, that's super valuable. That's definitely going to help a few people out, I bet. All right. Final item for this week, Paul, is our roundup of AI product and funding updates. So I'm going to rapid fire through a bunch of these as we wrap up today's episode. So first up,

1:29:43Anthropic is raising new funding at a reported valuation of 900 to 950 billion, according to some reports from the Financial Times and the New York Times. At the same time, Anthropic announced a $200 million partnership with the Gates Foundation focused on applying Claude to global health and development priorities. As if that weren't enough, Anthropic also launched Claude for Small Business, a new offering bringing Claude's enterprise capabilities to smaller companies with a range of different features. OpenAI launched a preview of a new personal finance experience in ChatGPT for

1:30:18US pro users that lets them securely connect financial accounts through a service called Plaid. Across more than 12,000 institutions, they can then see a dashboard of spending, investments, subscriptions, and upcoming payments, and ask ChatGPT questions grounded in their real financial context. There is support from the fintech company Intuit, and they're planning a rollout to the plus tier next. OpenAI also reorganized its executive ranks, officially making President Greg Brockman head

1:30:49of product strategy in addition to his work on AI infrastructure, merging ChatGPT Codex and the developer-facing API into one core product team that is now led by Codex head Thibaut Sotio and moving longtime ChatGPT head Nick Turley to lead enterprise products and naming former Instagram VP Ashley Alexander head of consumer products. SpaceX is aiming to go public on June 12th and was expected to be the biggest IPO of all time with Elon Musk's rocket company planning to raise as much as $80 billion

1:31:21or more and list on the NASDAQ. The information reports a couple items related to OpenAI and Microsoft. OpenAI apparently expects to save approximately $97 billion by 2030 under its latest revised deal with Microsoft, which we talked about in past weeks, and they reported that Microsoft has now spent over $100 billion on its OpenAI partnership to date. At the same time, OpenAI announced work with Codex from anywhere, expanding access to its Codex coding agent across more environments and

1:31:53surfaces beyond the original command line interface. Google was busy as well. They launched Gemini Intelligence on Android, an upgraded set of Gemini AI features directly integrated into the Android operating system. And Google DeepMind unveiled an early research concept that reimagines the computer mouse pointer for the AI era. They designed it in a way to give users a more intuitive way to interact with AI across applications. Amazon has launched Alexa for Shopping, a personalized agentic AI

1:32:25assistant designed to help users discover products and complete purchases on Amazon. This past week also saw the launch of Recursive Superintelligence, a new AI startup focused on building AI that safely conducts experiments to improve itself. And last but not least, isomorphic labs. The DeepMind drug discovery spin-out led by Demis Hassabis announced a Series B funding round to accelerate its AI-driven drug design pipeline. So, Paul, that is all we've got. One more quick announcement. We talked at the top of

1:32:59the episode about our AI pulse survey. Go take this week's survey at smarterx.ai forward slash pulse. This week, we are going to ask about how concerned you are about AI's impact on your own job over the next year and asking a question to kind of gauge where you personally are at with AI versus where your organization is at. So, go to smarterx.ai forward slash pulse to take the survey. Yeah, one quick note. Just don't sleep on, as I said before, those AI product and funding notes that Mike does at the end here. There's so much significant information in just the ones from

1:33:35this week. Like the isomorphic labs is major. If you're not following what they're doing, follow them. The Recursive Superintelligence is a huge deal. Some of these things are just very significant and will mean more in the near future for people. But yeah, don't breeze past those. It's good information to know. Well, Paul, thanks again for breaking everything down. Sorry it was a little gloomy this week, but I think it was an important conversation nonetheless. Yeah, they all are. I mean,

1:34:07it's important we're having these conversations, but if you're feeling the weight of them listening, trust us, we're feeling the weight of them talking about them sometimes. Yeah. Next week, we're going to try and try and get some bright spots. I will do my best to find some bright spots. All right. Thanks, Mike. Thanks, Paul. Appreciate it. Thanks for listening to the Artificial Intelligence Show. Visit smarterx.ai to continue on your AI learning journey and join more than 100,000 professionals and business leaders who have

1:34:37subscribed to our weekly newsletters, downloaded AI blueprints, attended virtual and in-person events, taken online AI courses and earned professional certificates from our AI academy and engaged in the SmarterX Slack community. Until next time, stay curious and explore AI.