What happens when the company that spent the past three years convincing Wall Street that it could build cars without human quality inspectors hires 300 of them back? What happens when executives who are writing the checks for all of these AI tools and platforms out there can't tell you what they're actually paying for until the bill finally arrives and it's 10 times bigger than what they actually budgeted for? And what happens when the CEO of one of the most connected, most respected, most known AI companies on planet Earth sits down on live television and says the quiet part out loud? He says the thing that all the CEOs in the country are only willing to say in private, but he goes out and says it for the public. Today, all three of those things actually happened almost at the same time. Ford admitted its AI-driven quality checks didn't work out and they brought humans back into the company. Enterprise tech leaders around the world are discovering that they've lost visibility into their own AI spending. And Palantir's Alex Karp went on CNBC in perhaps the best, I don't know, 15, 20 minutes of live unedited interview telling corporate America that the industry is betting billions of dollars on things and it's all gone completely wrong. It was one of the best interviews that I have seen for 2026 so far. So those are going to be the top three stories of the day. Welcome to Future Ready Today. It is Wednesday, July 1st, 2026. And if you enjoy the content on this podcast, please consider rating and reviewing the show on Apple or on Spotify or whatever your preferred platform or channel is. It is greatly appreciated. Okay, so let's get into the top story of the day or the first story of the day, I should say. This is from the Wall Street Journal. corporate tech leaders are reaching for cost control strategies they sharpened during the rise of cloud computing in order to keep ai spending in check as the shift from prompt based chatbots to always autonomous and always on ai agents is driving token consumptions higher to ridiculous levels that a lot of people did not predict at priceline for example dashboards are tracking employee token usage with monthly reports going directly to the CFO and CTO. High usage triggers a conversation with the employee, though limits can flex for revenue generating work. Priceline's Chris Reed compared giving employees AI access to putting a credit card in the end user's hands without controls. I actually love that analogy. Completing a task with an AI agent can require 50 times the computing power of a single chatbot question, according to Goldman Sachs analyst Jim Schneider, who projects that AI agents will increase token consumption 24-fold over the next four years. I don't even see why it's not going to be 30, 40, 50-fold. And business AI agents will increase consumption 55-fold by 2040. Again, by 2040, I would imagine it's going to be even higher than that. Model prices fell roughly 50% 5-0 between December 2024 and December 2025, but token costs grew four and a half times over that same window, according to Bain & Company. This is something I've said before as well. It's great that we are going to optimize token usage, meaning the efficiency of using these tokens, because as the token costs go down, It's going to be cheaper for us. But the ratio at which the cost of tokens is going down is nowhere near the ratio of how much more tokens, how much more AI we are using. And so if token costs go down by 50%, but our usage goes up by 450%, well, you're still going to be spending a lot more money than you're going to be saving. Bristol Myers Squibb's Chief Digital and Technology officer Greg Myers said he expects token costs to run orders of magnitude higher than today, though he's already prepared the company CFO and board for high consumption and believes the return on investment will be positive. Other companies are adopting cloud era FinOps practices. Smart Sheet has a FinOps team, which owns its AI budget with automated alerts before employees hit their token limits. Qualcomm caps token usage by team and uses a showback system tying dollar costs to token consumption. OpenText's CIO said the chargeback approach can lower an organization's token cost by 20 to 30 percent. And Lowe's is using smaller and open source models to avoid what it calls token wastage. Now, the futurist lens here is there's one number that I came across which was just staggering. And that number is 73.7 trillion. So almost 74 trillion. That is how many tokens that Meta's own employees burn through in a single month. These are not on customer-facing products. This is on an internal leaderboard nicknamed Cloudonomics that quietly turned token consumption into a status symbol. Now, from what I recall, this has since been switched off. They're not using this anymore. But the problem with Meta was not that they had a technology problem. It had a measurement problem. It optimized for exactly the wrong variable, and that is the result that they got. So when you make the metric the goal, then that's what's going to happen. If you tell employees that we want you to use AI as much as humanly possible, and the metric is going to be token consumption, then everyone is going to do whatever they can to maximize tokens. If you don't tell them that it is about generating business impact and value, making customers' lives easier and better, streamlining workflows. If that's not the goal, but token consumption is, well, then we're just going to use more tokens. And that's exactly what's happening to a lot of these different companies out there. Priceline, Qualcomm, Lowe's, many of these companies out there are all solving the same root problem with governance. Meta is basically just one of the first companies that hit it first at scale, and it made it pretty much impossible to ignore. Now I think there are gonna be three paths that are going to emerge and we already starting to see these happen And this is going to be when every company which this will happen every company is going to run into this exact same wall that Meta and every other organization is running into that's going all in in the AI and tokens. And so the three paths I see out there. So first, companies are going to build their own, meaning your own AI tools where you're not going to have to rely on anybody else. Meta's pivot towards MetaCode, that's its in-house coding assistant, is a company deciding it would rather own a mediocre model internally than rent a great one that it can't control. So I think that's going to be one path some companies will take. Build your own. Number two, go open wait. Lowe's, Qualcomm, and several other organizations out there are already routing lower stakes work to smaller open source models, and they're reserving the frontier models for tasks that actually are deemed worthy or necessary of those frontier models. And so cost-based selection is becoming a default architecture decision for a lot of these companies. And then third is going to be the routing layer, where a company is going to sit above all of these models and kind of act as a switchboard. And so when you put something into CLOD or ChadGPT, there's going to be some sort of a switchboard that basically says, based on this task, you don't need to use Fable 5 or Mythos or Opus 4.8. You can use Sonnet, or you can use almost a free version of ChatGPT. You know, if you want to ask a basic question, you want to know the weather, you want to understand, you know, what's going on in the world today. Yeah, you don't need to use Opus 4.8 or Fable 5 for that. You can just use something basic, probably even something that's free out there. But if you're looking to review code, to make changes to code, yes, you need Opus 4.8 for that. And it's going to basically like a switchboard point you in the right direction and automatically direct you to the right model and let that model get used. So those are kind of the three paths. And I guess there's a fourth one, which is you stick with your current frontier models, your subscriptions with the Anthropic, the OpenAIs. Number two will be build your own model. Number three will use open source models. and number four is have a kind of switchboard out there that will be automatically redirecting you to the right model that you should be using. So I think that is the direction that we're going to start to see more organizations take. It's still kind of playing out, you know, and as we get into story number three where I mentioned what happened with Alex Karp, it'll be clearer why these are going to be four distinct paths that are going to be emerging. Story two, employers who laid off workers for AI are actually revisiting, reversing their decisions. This was published CNBC, BBC, a couple other news outlets cover this as well. So a growing number of employers that cut jobs in favor of AI are reversing course. Now, yesterday I talked about how AI, I'm sorry, how Amazon is spending a billion dollars to hire thousands of new roles, not bringing back old ones that they let go, but creating an entire new category of job. Today, it's companies that are bringing back employees that they fired. Ford, the clearest example of this. So the automaker hired roughly 300 veteran quality inspectors and engineers after its AI-driven quality check systems, including 900 AI-powered cameras rolled out across the plants, failed to catch problems the way experienced staff could catch them. Vice President of Vehicle Hardware Engineering Charles Poon said the company had assumed that simply feeding design requirements into AI tools would produce a high-quality product of its own, and that assumption proved to be wrong. Many of the experienced technicians whose knowledge the systems needed had already left the company before that expertise could be captured. The rehired engineers are now training Ford's AI and machine learning tools, as well as mentoring younger staff. Now, my question is going to be, what happens after these 300 seasoned engineers who were brought back? What happens after they're done training the AI? Is Ford going to say, okay, you know, thanks for doing that. Goodbye again. is this going to be a little bit of a yo-yo back and forth? Now, this reversal comes despite CEO Jim Farley's earlier bullishness on AI's disruptive potential, and it lands alongside a win because Ford just reclaimed the top spot in the J.D. Power initial quality study for the first time since 2010, crediting the ranking partly to what the company called a talent refresh that included bringing back its veteran engineers. Common Bank of Australia, they laid off more than 40 customer service representatives last year and replaced them with, that is correct, you guessed it, AI voice bots. But the system couldn't handle the demand, and the bank reversed the layoffs. IBM previously used AI to replace certain HR functions. The system handled about 94% of routine requests, but couldn't manage the remaining 6%, including ethical dilemmas. So the data backs up the pattern, a report by OrgView, which found that 39% of business leaders admit, I'm sorry, 39% of business leaders made employees redundant due to AI deployment, and 55% of that group are now admitting their redundancy decisions were wrong. Separately, 32% of U.S. hiring managers said they eliminated a role primarily due to AI and later rehired for the same or similar position, according to Robert Half Data. Now, the futurist lens here is that mistake was not installing the cameras by Ford. The mistake here was letting your most seasoned veteran engineers walk out the door before anybody even asked them what these cameras needed to know what they were supposed to be paying attention to. Ford didn't lose quality control because AI is bad at quality control. It lost control because it automated the visible task that the engineers did, spotting the defect while quietly being able to discard the invisible one. That is knowing why the defect actually happens to begin with. What does it predict and which one of these defects actually matter? That's something that a seasoned engineer is going to do much better than AI. Now the knowledge for this never actually lived in a training manual It never lived in an online course or in a video or anything that these seasoned engineers and veterans went through It's something that they had to live through, something that they had to work through. It lived in the people who were doing this work for decades inside of Ford. The people who spent time working through many product cycles, who spent time going through iterations, seeing different products, different cars, different models and versions get created that they've worked on. And that is very different than just kind of, you know, plugging data into AI. And so once these seasoned engineers left, there was nobody left to ask these questions to. They were gone. And that is the big mistake that Ford and I think other companies are making. Now, if you put this next to story number one, I think you start to see a little bit of a pattern that starts to emerge, and that is companies that are optimizing for a measurable proxy that they have. So in the case of Meta, this is token volume. That's what Meta was focusing on. In the case of Ford, it was automated defect catch rate. So how quickly can they use automation to catch any defects inside the company. But while the thing that ultimately creates value, though, is the judgment that sits underneath that. So it's the judgment that is required from employees that sits underneath that token usage. It's the judgment from your veteran engineers that sits underneath the automated defect detection that your AI tools are you're putting into place. So Ford is rehiring these 300 people. Meta is cracking down on their AI token usage. So is Uber. So are many, many other companies out there. And so this is pretty much the same correction, but wearing different clothes. Both companies built a metric, which was looking at AI usage and AI consumption and things like that. And they watched people or systems game it or do things incorrectly. and then they had to walk things back. But they had to walk things back once they saw the gap emerge between the metric that they were striving to optimize for versus what humans were actually able to do. And I suspect that this pattern will keep showing up across a lot of organizations. Now, remember one of the things that I talked about yesterday because there was another story that came out today looking at the job market for entry-level graduates. and again the story was that it's becoming harder and harder for entry-level grads to find jobs but if you listen to the podcast yesterday which if you haven't i highly recommend that you do data came out which actually showed that in organizations that are heavy users and adopters and investors in ai tools those companies are actually hiring 10 to 12 percent more and growing 10 to 12 percent at a greater rate every year than companies that are just dipping their toe in the AI pool, so to speak. So again, this job growth or slowdown, if you want to look at it that way, is not evenly distributed. It's only showing up in the organizations that are hesitant and tentative and lukewarm and are piloting these AI tools. But in the other organizations that are going all in and really focusing and doubling down on this. In those companies, they're not seeing the slowdown. They're seeing more entry-level grads. They're seeing greater growth over year. And so it's not an evenly distributed story. So getting back to the Ford piece, I think any organization out there, before they're thinking of layoffs and going through any of these types of AI transformations, you really need to start asking yourself, whose tacit knowledge is going to disappear when somebody leaves the organization? And did you actually capture that tacit knowledge before they leave? The forward example, that is what happens when an organization finds out too late that they didn't capture that tacit knowledge. So hopefully going forward, organizations will still do a better job making sure that does not happen to them. The last story of the day that I want to get to, and of course, before I get into this one, if you're a chief human resource or chief people officer, please consider checking out my CHRO group called Future of Work Leaders. Don't spend $10,000, $15,000, $20,000, $30,000 a year joining these CHRO groups. We're getting access to outdated information and webinars and speeches and blah, blah, blah, blah, blah. You can spend a fraction of that cost and actually engage with peers, chief human resource and chief people officers who are shaping the future of work, going beyond traditional HR. You can learn more about what we are doing and working on by going to futureofworkleaders.com. Again, that is futureofworkleaders.com. Now, the last story of the day. This, to me, was my favorite one because I actually watched this entire interview. I think it was on CNBC, if I'm not mistaken. CNBC or MSNBC. I think it was CNBC. And it was quite an interview. I mean, I enjoyed it a lot. So Palantir's CEO, Alex Karp, which if you haven't seen any of his interviews before, He's a very animated speaker. That's, I think, probably the best way to put it. He's fun to watch. He's got a lot of great points, very smart guy. But he called the AI industry, quote, effing insane in a heated CNBC interview, which took place today. He accused leading AI firms of overcharging, exploiting customer data, and jeopardizing U.S. national security. Karp appeared to discuss Palantir's expanded NVIDIA partnership. That's why he was brought on the show, which integrates NVIDIA's Nemetron AI models into Palantir's sovereign AI platform, letting government agencies and enterprise customers today deploy... Where did I leave off on this? Ah, deploy AI in secure environments, while retaining control over their data and model weights. And so the interview turned into a very interesting broadside argument against OpenAI, against Anthropics, token pricing. And what ended up happening is that CARP said that the leaders and organizations are telling them, quote, I'm paying for tokens and I'm not getting any value. He argued that a lot of the enterprises fear their proprietary business value is being extracted by frontier labs without adequate protection and that customers want ownership over their own compute over their own models and over their own data, not a setup that quietly transfers value to a third party. And this is a big concern, of course, because you can imagine if you're using something like ChadGPT or Claude, I mean, you're giving it access to all of your most sensitive, confidential information and data about customers, about employees. You're training it. You're feeding it. And the concern is, well, what are these frontier labs doing with that information? Are they keeping it? Are they going to build their own things? Is it secure? Is it safe? Like what's actually happening with all these things? Am I being overcharged for tokens? And so Alex Karp is basically making the argument that the entire narrative for AI from all these vendors was completely oversold, overblown. It was exaggerated. You know, the end of the world, it just, he's making the argument that it just became too much. And that what ends up happening in that kind of a situation is you create a, um, you know, a massive tidal wave of everybody that wants these tools. And now they're struggling to find the value and the ROI from it. So it's kind of like, uh, you know, telling everybody you're going to come to the greatest party ever. This is going to be the best. We're going to have the best food and the best blah, blah, blah. And everybody shows up and you're like, yeah, all right. I mean, it's okay. I've been to better parties. I mean, this is fun, but it's not the best thing that I've ever been to. You kind of overhyped it a little bit for me. You know, you show up a little bit overdressed, your fancy tuxedo. Other people are walking around in shorts and t-shirts. And that's kind of the vibe that Alex Karp is making it sound that we are all experiencing now. He also talked about national security. He was very livid in this regard because he said that leading AI companies are criticizing U.S. reliance on these AI companies to develop military technology by asking, quote, are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley. That is effing insane. And when pressed on this, Alex Karp said, quote, this is the voice of American business that is being channeled through me. So very, very interesting interview. I mean, he did make some great and some valid points in there, which I think are worth paying attention to. If you get time, go onto YouTube, you can find the full interview in there. Now, the futurist lens here is that if you take Alex Karp's anger out for a second, which I think is justified, he's explicitly pitching the third path that I'm talking about from story number one. That's the router, because he described Palantir's product as letting customers switch between models rather than being locked into OneLabs pricing and OneLabs access to their data. And so he has this application layer, which he was obviously on the show promoting, an application layer that acts as this kind of a switchboard that allows data and information. It goes through their application, and then through the application, it goes into the different AI systems. And this is the emerging category that I flagged a couple minutes ago. And Karp is trying to be the first one out there to really define this category publicly, of course, on live television in the most combative way possible, which is his reputation. Now, it's also worth noting, Anthropic reportedly spent around $2.66 billion on AWS in the first nine months of 2025, more than its estimated revenue. And OpenAI is reportedly spending $2.25 for every $1 that it earns. So the frontier labs out there are burning cash to hold the lead, which means that token prices are likely to keep falling as they compete for enterprise volume. But as I talked about a couple of minutes ago, the cost of the tokens doesn't matter if the usage is increasing by 400x. So I think Karp is making his pitch at the exact moment when a lot of these frontier labs out there are most desperate for enterprise revenue. And they're least able to walk away from customers' demands. So the Anthropics, the ChatGPTs, they need the enterprise revenue. And to be honest, when we talk about frontier models, I think most companies out there are really kind of pitting ChatGPT against Anthropic. Yeah, Gemini is somewhat in the mix. Yes, Grok is somewhat in the mix. Perplexity is tossed around in there a little bit as well. But I think most organizations, when they think of these frontier models, they're really thinking about Anthropic. They're thinking about ChatGPT. Yeah, Copilot as well. but I don't think people view any of those other platforms at the level of Anthropic and ChatGPT. So very, very interesting interview, to say the least. And so if CARP's numbers, I think, are directionally right, then the CEOs who are livid right now, they're sitting on potential decisions that they haven't made yet, and that is the direction that they want to go with these AI models. Do they want to put their money into the Anthropics? Are they going to build their own? Are they going to use open source? They're going to go the switchboard direction. And so I think the important trends to be paying attention to here are not going to be necessarily the ones who are making the biggest noise, but kind of the quiet investments, the quiet decisions, paying attention to what some of your peers out there are doing, what they're thinking about, not just paying attention to what's making it on CNBC. So those are the top stories of the day. I want to leave you with a quote from Henry Ford quality means doing it right when no one is looking so that is future ready for today Wednesday July 1st my email if you have questions comments or want to sponsor the show jacob at the future organization dot com thank you for tuning in for watching and listening I will be back tomorrow enjoy the World Cup games today go team USA marketing is hard but i'll tell you a little secret it doesn't have to be let me point something out you're listening to a podcast right now and it's great you love the host you seek it out and download it you listen to it while driving working out cooking even going to the bathroom podcasts are a pretty close companion and this is a podcast ad did i get your attention you can reach great listeners like yourself with podcast advertising from libsyn ads choose from hundreds of top podcasts offering host endorsements or run a pre-produced ad like this one across thousands of shows to reach your target audience in their favorite podcasts with Libsyn ads. Go to Libsyn ads.com. That's L I B S Y N ads.com today.
