CEO-Led AI Gets 3X the ROI

Today on the AI Daily Brief, why companies where CEO owns the AI strategy are seeing 3 times as much ROI. Before that, in the headlines, so much is going on, including OpenAI announcing their first custom-designed chip. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, KPMG, Super Intelligent, Mission Cloud, and OutSystems. To get an ad-free version of the show, go to patreon.com/aidailybrief, or you can subscribe on Apple Podcasts. And if you want to learn more about sponsoring the show, send us a note at sponsors@aidailybrief.ai. Lastly, check out our new enterprise upgraded training programs at training.beesuper.ai. The new Executive Agent Leadership Program kicks off next week and is registering now. Man, some days there is one dominant story that just absolutely kicks into dust all the small things that are happening around it. And then there are other days where there's no one singular story that everyone is talking about, but about a million smaller stories that, if we're looking for them correctly, tell us all sorts about what's actually happening in the world of AI and almost serve as tea leaves for what might happen next. Today is one of those days, so these headlines might be a little bit longer than 5 minutes. First up, OpenAI has unveiled their first in-house chip. The chip is codenamed Jalapeño and was produced in collaboration with Broadcom. OpenAI referred to Jalapeño as an integrated processor and described it as the first AI accelerator in a multi-generation compute platform. In a statement, OpenAI President Greg Brockman said, "The world is moving to a compute-powered economy. Jalapeño is part of our long-term full-stack infrastructure strategy to make compute more abundant, resulting in AI which is faster, more reliable, more affordable for people and businesses, and can be used to solve more important problems. By designing more of the stack ourselves, we can serve more intelligence with greater efficiency and keep pushing advanced AI towards broader access." The chip is an ASIC, similar to Google's TPUs, which means it's designed for the specific task of serving inference for LLMs. By contrast, NVIDIA's GPUs are much more general in their application. OpenAI said that they believe this to be the fastest development cycle ever for a high-performance ASIC, going from initial design to manufacturing tape-out in 9 months. In an interview with CNBC, Brockman credited the speed to AI-enhanced design, commenting, "The degree to which our models have been able to accelerate it was very surprising to us." Now, while OpenAI will begin deploying the chips as soon as they're ready, this almost certainly does not mean they'll cut down on Nvidia orders. Brockman reaffirmed that OpenAI, quote, cannot get compute fast enough. Appearing alongside Brockman, Broadcom CEO Hock Tan agreed, stating that compute demand from all of their customers is, quote, simply insatiable. Tan added, it's just much more than we can address. And this is not just '26, not '27. We're seeing that same and even elevated demand in '28 as well. Now, in a little bit, when we talk about Micron, we will come back to why the long duration nature of that demand is one of its most significant aspects. One additional small update from OpenAI. At this point, any model update that isn't at the edge of the frontier is likely to fall on fairly deaf ears, but OpenAI has handed free users another upgrade with a new version of their GPT-5.5 Instant. For the portion of people who are on the free plan, which is the vast majority of ChatGPT users, these sort of updates to Instant can make a big difference. OpenAI claimed the model is much more fun to talk to, saying, "Our most used model is now better at understanding the intent behind a question and adapting its response accordingly. It also handles complex constraints more reliably and makes shopping and local recommendations more useful and cohesive." Now, for those trying to understand where general consumers and free users fit alongside the clear increase in importance of enterprise customers, OpenAI has now released upgrades to their Instant model every month or two since February. Whether that's because they really care about free users as a category or because they see them as top of funnel for their bigger enterprise use doesn't really matter. In practice, the models that the free users have access to continue to improve as well. Now, on the question of the model that we are really waiting for, we continue to experience rumor whiplash as prediction markets went from very dreary about Fable 5 to all of a sudden a massive increase in the chance that we get Fable 5 back soon. Around 2 PM on Wednesday, the odds of a Fable return by July 1st skyrocketed from 15% all the way up to 63%. Assuming that there was some sort of insider knowledge going on, Greg Eisenberg posted, "Someone knows something." Now there have been at least a few signs that things are moving. Earlier in the day, Synthwave posted a code snippet from a Claude code update, which they believed, quote, "hints at preparations for a Fable 5 return, with it being permanently included in subscriptions with weekly usage." The code snippet adds a warning for using up weekly Fable 5 limits and removes a reference to separately purchasing access to the model. A little later, they noted that Fable is also reappearing on Amazon Bedrock. Ever hopeful, Chubby posted, "This Fable 5 update sounds almost too good to be true. Including Fable in subscriptions would be fantastic, and I hope it's true insofar as Anthropic generates good PR with it." Still, a couple of hours later, the headline that likely drove the shift in the market came out, with Wired reporting that the Trump administration is on the one hand sick of Dario Amadei, but on the other seemingly more than happy to deal with co-founder and chief compute officer Tom Brown. According to one White House source, with characteristic parrhesia, they said, Tom Brown is not being a weirdo like Dario and can actually engage. Now, Brown traveled to Washington last Monday to participate in negotiations, and reports state that Dario has now been sidelined from the discussion as talks continue by phone. TR Taxis commented, one of the two must stay in the locker, either Fable or Dario. Dario is a weirdo, so Fable gets to walk. Now, aside from that, the reporting was pretty vague on how much progress is being made. It reported that there's still no timeline for reinstating Fable, but said talks are ongoing with leadership and technical teams at the White House. Sources said a big part of the conversation has shifted to establishing what level of proof Anthropic could provide to alleviate the administration's concerns about the jailbreak. I don't know, man. I guess it's better than bad news, but I'm kinda with Ran Longevity when he writes, I'm gonna believe Fable is imminently coming back and I'm ready to get hurt again. Now staying on Anthropic news, in a follow-up to our discussion yesterday of Claude Tag, the release is proving to be a little bit more controversial than I would have guessed, and I think it has to do with a couple of things. First, the response, at least among the highly enfranchised AI community, shows that Anthropic's reputation in the community right now is kind of at a low ebb. Secondly, I think people responded fairly negatively to how much folks from Anthropic were trying to say that this thing is much more than the Slackbot it seems at the surface. Specifically, Andrej Karpathy caught a lot of guff for calling it a new paradigm. In a discussion on another post, he said, I think a number of people on the timeline didn't read past the title and made inferences and comparisons that are just wrong and then use it as an opportunity to take cheap shots. This is not a quote-unquote feature like some crappy Slack bot, and it's certainly not a CLAW, though it has aspects of it. It is an org-level harness. The difference will become clearer over time. Now, if you listened to yesterday's episode, you will have heard the argument for why this is indeed potentially more than just a Slack bot, but there was another strand of critique which I think is a little bit more interesting, especially in the context of everything else happening in the AI industry. Ashwin Gopinath of Centra suggested that Claude Tag will start off as a handy feature, but quickly transform into vendor lock-in. Now in some ways, this is just a natural byproduct of anything that gets more deeply integrated into organizational context. You might remember when people were concerned about memory as a lock-in when it came to individual accounts, but those concerns petered away a little bit, at least when Labs started to introduce one-click migrations. But what people are recognizing is that it's going to be a lot more difficult to migrate away from something like CloudTag once it's fully embedded in the organization. Summing up this point of view, Mark Gajenstadt wrote, Claude Tag is turning your company's context into vendor lock-in, and it looks like convenience until you try to cancel. Herbie Bradley had a lengthy take on the pros and cons that essentially boil down to a combination of pricing anxiety and a lack of user control. His post expressed the reality that we don't know a great deal about how expensive it will be to deploy Claude in this way across the organization, but that people's assumption is that it's unlikely to be cheap. Now, I don't think that these conversations are unreasonable, and I think that for anyone who's been starting to flirt with the idea of different model architectures or even using local models, this might be another example of why that could be valuable. At the same time, I do think a little bit of the concerns that people are sharing are sort of just the inevitable outcome of AI getting more deeply integrated, whether it was Claude or ChatGPT or something else. It is, yes, definitively the case that when you take all sorts of time to give organizational agentic systems access to lots of important context and permissions, it creates a very high barrier to switching. But that's not Anthropic-specific and it's not Claude Tag-specific. It is just an unavoidable consequence of AI doing what we hope it will, which is making the organization work better. Capturing the nature of the challenge, Ethan Mollick writes, decisions about how to use AI in your organization are increasingly organizational design and strategy decisions, not IT choices. How do you integrate agents into your firm? What intelligence will you outsource? What are the boundaries of the firm? What is the role of people? Now, one other interesting story from Anthropic, the company has accused Alibaba of illicitly accessing their models in order to distill Claude's capabilities. In a letter to the Senate Banking Committee, Anthropic has accused Alibaba of, quote, brazenly and illicitly carrying out what they describe as the largest distillation attack ever detected. Anthropic says that Alibaba accessed their models almost 29 million times through a network of 25,000 fraudulent accounts. They say the attack ran from mid-April through to early June before it was shut down. The letter states, these distillation attacks are carried out illicitly, systematically, and at an industrial scale to harvest US AI capabilities across frontier labs and repackage them as their own without incurring the training and R&D costs required to train US frontier models. Anthropic warned the senators that models created via distillation often lack safety guardrails, posing broader security risks. The letter also noted previous attacks from the Chinese AI sector, including a major campaign from Deepseek, which they publicly disclosed in February. Anthropic claimed these distillation attacks pose a threat to the US military and broader competition with China, commenting, "Distillation attacks turn hundreds of billions of dollars in American investment and R&D into a massive subsidy for our geopolitical competitors." Now, a couple interesting things about this letter. First of all, Anthropic is clearly ratcheting up the rhetoric when it comes to Chinese model distillation. The attacks that they're discussing are ultimately nothing more than using Claude, recording the outputs, and repurposing them as training data. In other words, they don't degrade Anthropic's product in any way. Distillation does allow the Chinese labs to catch up quickly, but calling them attacks is a deliberate choice in how Anthropic communicates with Washington. Second, Anthropic describes the attacks as illicit largely because it's unclear that anything actually illegal is going on. In other words, this is a breach of Anthropic's terms of service, but not necessarily the law, although that could change soon. Senators Hagerty and Kim have proposed a bipartisan bill addressing distillation to be included in this year's Defense Authorization Act. If passed, the bill would blacklist or sanction any Chinese lab found to be distilling US AI models. For now, Anthropic seems to be agitating for more action. Now, none of this is to say that Chinese distillation isn't a big problem. A post on Hacker News this week discussed a thriving underground economy reselling Anthropic tokens. "Formally, both Anthropic and OpenAI block access in China, but informally there is a huge market for discounted AI tokens farmed from Max subscriptions." Commenting on the post, Chubby wrote: "There may be an entire gray market economy around Claude access in China. Resellers allegedly pool Claude Max accounts, operate bot networks, and sell access far below official API prices. The more interesting claim: user logs and reasoning traces may be resold as training data. If true, this is not just API abuse but model access arbitrage turning Frontier AI usage into a shadow data pipeline." Meanwhile, separately, Alibaba has sued the Department of Defense over a decision to designate them an affiliate of the Chinese military. Earlier this month, the Pentagon updated their list of firms with ties to the Chinese military, adding more than a dozen firms. This includes every Chinese cloud giant alongside multiple electric vehicle companies, robotics labs, and chipmakers. The designation blocks these firms from doing business with the Pentagon and restricts lobbying activities. Many analysts also view the designation as a precursor to these firms being blocked for civilian use, as occurred with Huawei. In the lawsuit filed on Tuesday, Alibaba claimed they had no affiliation with the Chinese military and that the Pentagon had acted unlawfully in applying the designation. The lawsuit stated, "The designation thus does not merely impose commercial costs, it strips Alibaba of its ability to petition the government through its chosen representatives." Alibaba claimed its relationship with the Chinese government is purely regulatory and no different to any other firm operating in China. The Chinese government has also spoken out against the expanded list of designated firms. In comments earlier this month, the Chinese Ministry of Commerce said the US had, quote, "disregarded the consensus reached during the recent trade summit." One more quick bit of lab news and then a little market news and then we're out of here. Google continues to bleed talent as two additional senior researchers head for the exits. Last week, DeepMind was rocked by the departure of AI luminary Noam Shazeer and Nobel laureate John Jumper, who joined OpenAI and Anthropic respectively. On Wednesday, Bloomberg reported that Jonas Adler and Alexander Pritzl were also leaving to join Anthropic. The report described Adler and Pritzl as senior researchers who were viewed as key contributors to Gemini. After scraping social media, ChrisGPT found two more Googlers parting ways this week. He argued, typically this is a sign when a company is about to release a subpar model. This happened with OpenAI, xAI, and is now happening with Google. While we don't know whether Gemini 3.5 Pro will actually be subpar, we now do have confirmation that it's been delayed. After speaking with sources, Business Insider reported that the model will not be released this month as planned. Instead, DeepMind is now indeed targeting a July launch. With their sources suggesting that they're using the additional time to tweak the model based on feedback from early testers. In particular, testers are being asked to stress test the model in real-world coding use cases using antigravity. Now interestingly, Meta researcher Lukas Beyer suggested this might not be just about Google falling behind. He commented, "One thing I noticed with the big departures lately is that most of them are longtime Londoners leaving Google DeepMind. This would be consistent with laments I've heard about the center of gravity for pre-training slowly but surely shifting to Mountain View." Mountain View is of course Google's main campus south of San Francisco, but since DeepMind was founded in London, that's historically been a big locus for them. Perhaps notably, Anthropic opened a major office in London in April with space for 800 employees, conveniently just a few miles away from the DeepMind office. Finally, over in markets, it is bubble on, bubble off, as blowout earnings from Micron send the markets in the opposite direction. Throughout this week, concerns had been steadily growing that the bull market for AI stocks was coming to a close. The narrative began with SpaceX falling 16% on Monday and accelerated as Cerebras fell below its IPO price for the first time on Wednesday. Even well-established hardware stocks were taking a hit as the drawdown spread. SanDisk, Micron, and Arm all fell more than 10% on Tuesday and bled lower on Wednesday, dragging the NASDAQ down 3.8% so far this week. Now, there are of course plenty of external catalysts to explain the plunge, including the on-again-off-again peace deal with Iran and growing expectations of rate hikes from the Federal Reserve. But for the AI-centric narrative, analysts seem to believe it's just time for the semiannual bubble jitters set in. Gravity strikes, proclaimed JPMorgan analysts in a Tuesday note. Dan Ives of Wedbush wrote, with Micron set to report earnings this Wednesday, there is some added nervousness on the important memory chip trade in this market. We will continue to go through a number of gut-check moments in the tech trade as the AI revolution remains in the third inning. This morning is another one of those moments. Now, with the stakes established, analysts held their breath on Wednesday night to see if Micron could come through, and the results were much stronger than anyone expected. Micron delivered a beat on top-line revenue and profits, reporting 445% year-over-year revenue— and a 74% jump from last quarter. More importantly, Micron hiked forecasts, guiding another 22% jump in revenue for next quarter. They also disclosed 4 long-term contracts with what they described as very large customers that lock in current memory prices, which are historically high and deliver 56% gross margins. Micron executives said that they expect the memory market to be undersupplied for at least the next year, forecasting that gross margins will expand to 86% in Q4. The market was quick to pivot in overnight trading, setting the stock up 14% in overnight trading, recovering the entire drawdown from this week. Now, for months, the bearish case for memory and storage companies has been their boom-and-bust track record. Any sign of weakness led to an aggressive sell-off based on the belief that a bust is inevitable. The Wall Street Journal noted that even though several semiconductor firms are up 10x over the past year, they're, quote, still cheap, trading below 10 times projected earnings over the next 12 months because investors are skeptical that the good times will keep rolling. This is the narrative that Micron disrupted on Wednesday night. It is now clear that their massive growth surge in Q1 wasn't a one-off. It appears the AI industry is instead driving a structural shift in memory demand, and the suppliers are struggling to keep up. Goldman Sachs seems to have the correct read on the market, warning that consensus forecasts are underestimating the size of the AI buildout by as much as 50%. In a note earlier this month, they wrote, the investment boom is likely to extend, and near-term expectations of its scope may still need to rise. But with a lot of value already built in, markets are more vulnerable to news that challenges an optimistic view. We'll see what the summer has in store for markets, but for now, that is going to do it for the headlines. Phew. Like I said, extended headlines today. Next up, the main episode. One of the most important AI questions right now isn't who's using AI, it's who's using it well. KPMG and the University of Texas at Austin just analyzed 1.4 million real workplace AI interactions and found something surprising. The highest impact users aren't better prompt engineers. They treat AI like a reasoning partner. They frame problems, guide thinking, iterate, and push for better answers. And the good news? These behaviors are teachable at scale. If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time. Learn more at kpmg.com/us/sophisticated. That's kpmg.com/us/sophisticated. Today's episode is brought to you by the new Executive Agent Leadership Program produced by Superintelligent and by frequent AIDB operators guest Nufar Gaspar. To tell you a little bit more about the Executive Agent Leadership Program, here is Nufar. The best predictor of agent adoption in an organization is how hands-on their leaders are. Talking about agents is completely different than building them. Our participants, ICs all the way to C-suite, have built working agent fleets, governance frameworks, and the playbooks to scale it. Executive Agent Leadership is the evolution of Enterprise Claw. Everything we've learned across 3 cohorts, rebuilt for right now. The token economy, security, vendor resilience, and the architecture to lead agent adoption at scale. The next cohort of the Executive Agent Leadership Program is signing up now and will launch on June 29th. You can find out more at training.bsuper.ai. The average enterprise is spending $11.5 million on AI this year, and most of them can't prove a single dollar came back. What does AI actually look like when it produces ROI? Ask the healthcare company that just made their payment processing 320 times faster, or the law firm whose document research went from 3 months to 10 minutes. Or the contact center who reduced wait times by 99%. These are real Mission Cloud customers with real results. Mission Cloud is a CDW company and an AWS Premier Tier partner. They're the AI-first, outcomes-obsessed AWS experts who build AI solutions that drive your business forward. Whether you're flooded with AI ambitions but no idea where to start, or 6 months into a deployment that's going sideways, they've seen it and they've fixed it. Stop burning your budgets on AI that doesn't produce results. Start at missioncloud.com. This episode of the AI Daily Brief is brought to you by OutSystems, a leading agentic systems platform built for the enterprise. Organizations all over the world are building, orchestrating, and governing agentic systems on the OutSystems platform, and with good reason. OutSystems' open and unified platform allows teams to architect, deliver, and scale governed agentic systems with agility. Teams of any size and technical depth can use OutSystems to build, deploy, and manage AI apps and agents quickly and cost-effectively without compromising reliability. Scalability and security. With OutSystems, you can rapidly launch ideas from concept to completion. It's the leading agentic systems platform that is unified, agile, and enterprise-proven, allowing you to accelerate growth, reduce operational friction, and deliver real enterprise impact with AI. OutSystems—build your agentic future. Welcome back to the AI Daily Brief. In today's episode, we are looking at the latest KPMG Quarterly Pulse Survey. Now, one of the things that's been challenging this year about enterprise data is that there was such a massive shift that happened at the beginning of this year for early adopters, kind of between November and January, and then for everyone else starting from January on, that a lot of the surveys that companies have done just really don't reflect the reality anymore. Now, overly simplifying, it's the shift from non-agentic to true agentic AI, but everything from the use cases to the patterns of how we interact with it changes so dramatically that I haven't found a lot of studies that I think provide a lot of signal. What's useful then about the KPMG study is first, that it is a quarterly repeated survey, so you get a more longitudinal view. And two, that these survey results were actually collected in that agentic period. It wasn't back in the before times. And there are some pretty interesting findings in this. So let's dig in. The story is very much of AI on the rise. You see confidence in AI rising. You also see where it sits in organizational strategy, increasing who's in charge of it. Shifting even higher in the organization. And for the first time, we're starting to see some of the trends that we've been discussing on this show recently, including cost considerations at the frontier, start to actually find their way into AI strategy discussions. Let's start on the confidence department. One of the most encouraging things is that the percentage of these senior leader respondents who say that AI is currently driving meaningful business value at the organization level has jumped 12 points from 64 to 76%. Now, importantly, as we will see, this does not mean that they have perfectly precise ROI metrics, but it is still a powerful indicator of the reported sensibility among executives about how AI is working at an organizational level. Perhaps unsurprisingly then, we also saw a couple of interesting shifts in where on the maturity spectrum organizations self-report. The spectrum that KPMG uses from early stage to mature stage is research and development, experimentation, strategic planning, scaling the technology, driving adoption, and established ROI. Now, of course, one challenge with this is that there is so much variety across different parts of the organization. And also with AI, there's usually overlapping waves based on the type of technology. So for example, certain types of use cases from the pre-agentic era might be an established ROI, while some of the more advanced agentic uses might now be in experimentation. That's the limits of the self-assessment, but I still think it's overall an interesting way to see where organizations see themselves. Both research and development and experimentation are down because organizations are moving farther. The percentage that are in the strategic planning stage has stayed the same. And actually, the number who are in the scaling the technology stage has also gone down a little bit from 26 to 22%. But that's because by far the biggest jump was seen in the fourth of five stages driving adoption, i.e., embedding AI across the organization. That jumped 9 percentage points from 13% of respondents to 22% of respondents. And in one of the clearest indications I've seen yet, outside, of course, of the AIDB Usage Pulse surveys, which have shown this throughout the year, is that opportunity AI, in other words, strategic opportunity generating use cases for AI, is on the rise, while efficiency AI is proportionally on the decline. So in terms of where organizations' priorities are when it comes to AI, faster, better decisions declined from 41% to 36% between Q1 and Q2. Productivity gains declined from 42% to 35%, and cost reduction declined from 31% to 29%. Now remember, I don't think any of those things like productivity or cost reduction are meaningless or not important. I just think that they are the amuse-bouche of what you can really get out of AI. And on the priorities rising side, you have human-AI collaboration and fluency going up from 28% to 30%, responsible AI and governance going up from 26% to 28%, adaptability and resilience going up to 20% from 18%, and ecosystem and partnerships going up from 12% to 16%. KPMG sums this up as AI priorities becoming more strategic. Now, what about the big concerns? Data security, privacy— that has remained pretty consistent for a long time as the top concern among these enterprises. But interestingly, you are very, very much starting to see the AI subsidy era ending showing up in the numbers in terms of organizations that have these different concerns. Pressure to demonstrate value jumped from 19% to 24%. Limitations on hiring and upskilling jumped from 18% to 22%. Access to lower-cost LLMs had a big jump from 15% to 22%. And I would anticipate, of course, that we're going to see that do nothing but increase in the immediate term. But it's interesting to see that even in this period before we had the most dramatic shifts to usage-based models, that cost and that interest in lower-cost LLMs is already on the rise. Now, when it comes to leadership of AI, a couple really interesting things. First of all, the percentage who say their CEO actively owns AI as a strategic priority is very high, all the way at 75%. To me, this is a very strong indicator of just how significant of organizations getting the idea that this is not a tool selection problem, but an organizational design challenge. Now, what's interesting is that although three-quarters say that their CEO actively owns AI as a priority, the actual accountability is somewhat diffuse and distributed. Which makes sense given how different AI is going to interact with different parts of the organization. KPMG found very few organizations that have a single point of accountability for AI-informed decisions. It's usually spread between a CEO or executive committee, some named C-suite executive, or other groups like the business unit leader or a centralized AI governance group. However, whatever combination of accountability there is, organizations that had clear accountability were 3x more likely to report ROI from their AI. So if you are an enterprise and you are looking for quick wins after you listen to this, making sure that everyone knows who is accountable for what decisions when it comes to AI seems to be a clear indicator of a stronger AI organization. And man, when the CEO is accountable for key parts of AI, it massively changes the outcomes. I talked about the established ROI where the CEO is accountable. 14% of respondents reported seeing established ROI, but when the CEO is less or not accountable, that number dips all the way down to 4%. When asked whether AI is currently delivering meaningful business value, only 21% of organizations who had the CEO who wasn't accountable said yes versus 57% where the CEO was accountable. Same with confidence in their organization's ability to future-proof its AI strategy. Where the CEO wasn't or was less involved, it was only 22%, whereas when the CEO was accountable, it was 60%. So quick win number 2, sorry CEOs, if you are listening, it is your job or else your organization is going to have a much tougher time. We're also seeing a growing maturity in AI deployments. One example of this is that just about half of the responding organizations had rephased AI deployments when they discovered that costs had outweighed expected value. This to me doesn't threaten the overall trajectory of AI as all these other numbers show. It just shows that organizations aren't just buying hype and dreams. They're actually figuring out what works for them. It should be a reminder though to organizations that even if you feel behind, That does not mean that every AI implementation you're going to do is going to work, and you do need to be comfortable cutting your losses and saying, let's repurpose those funds in time for something else. Yet there are still some big challenges. Only about one-third of organizations report having full visibility into their AI operating costs and actively monitoring them, which I think is going to be hugely challenging when it comes to this new token efficiency era. I would not put this in the category of quick wins, but if you are looking for strategic ideas coming out of this, If you are in the two-thirds of organizations that don't have that full visibility, I would suggest even before you shift any strategy, creating systems to actually have active monitoring around costs is going to pay dividends in the long run. Right now, about 54% of organizations have a cost review as part of AI approval processes. 53% have AI cost monitoring dashboards, and about 40% have usage or token budgets. Even as someone who is completely convinced that AI is going to change everything, with enough time, all of those numbers will be at 100%. Now, one interesting last note is around the human side of AI scaling. And frankly, this is one area where I really think that these executive surveys can only tell half of the story. One very common thing that we've seen across so many different surveys is that bosses tend to radically overestimate the excitement around AI relative to their employees. So 71% of these executives report making good progress towards becoming a fully integrated AI-human workforce. Which is great, but I'd like to see that number from individual contributors inside those organizations. And while globally significant employee adoption of AI agents rose from 25% to 28% and resistance a little bit to 14%, in the United States there was a big difference where resistance to agents increased from 5% to 20%. Now, by next quarter, we'll be able to find out whether that is noise in the data or whether that represents something more significant. But it is certainly something to keep an eye on. Ultimately, I think there is a lot to be excited about in this survey, a lot that reflects what we're seeing more broadly in the trends, and a lot that indicates that organizations are starting to think about things in a smarter, more long-term sort of way. We will of course report on Q3 when it comes out, but for now, that is gonna do it for today's AI Daily Brief. Appreciate you listening or watching as always, and until next time, peace.