Speaker 1If you survey just the mainstream media, you might have the impression that business adoption of AI was in some way impaired or slowing down. The stories that make the news about enterprise adoption of AI seem primarily right now to be about companies restricting how much their employees can spend on it. Like Apple, for example, cutting the amount that engineers can spend per day, which was a story this week. And yet, when you take a step back, not only has AI adoption in the business sector continued unabated, the speed at which it is growing is actually increasing, defying all logic and expectation as it starts from a bigger base. The story of enterprise AI adoption this year is very clearly the story of agents, and that's what we're looking at on today's episode. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, quick announcements before we dive in. First of all, thank you to today's sponsors, KPMG, CDATA, Granola, and Blitzy. To get an ad-free version of the show, go to patreon.com slash ai-daily-brief, or you can subscribe on Apple Podcasts. And if you want to learn more about sponsoring the show, send us a note at sponsors at ai-daily-brief.ai. One more quick shout out before we dive in. Our newest premium AI training program is starting soon. It's called the AI Champions Boot Camp. It's three weeks long, and this one has six live sessions led by Nufar Gaspar, frequent contributor to the show, and will not just help people use agents more effectively, but have the skills and resources to do so. So if you're interested in learning more about AI, have the skills and resources to bring that use to their entire teams. You can get all the information for that program at besuper.ai, and the boot camp begins on November 16th. Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. Well, it looks like the markets got a little bit annoyed at my overly optimistic analysis, because yesterday, AI stocks plummeted in the worst drawdown since 2020. Overall, the MAG-7 lost a combined trillion dollars in market cap, with Nvidia falling 9.3% and Microsoft dropping 3.6%. The pain was even greater in the AI supply chain, where Micron lost 17.5%, and SanDisk plummeted 22.3%. The S&P 500 fell by 4.4%, and the NASDAQ collapsed by 6.4% in the largest single-day drop for the major indices since March of 2020. Fatalistic takes were all over X. Kevin Hsu from Alpha AI wrote, We're effed. Absolutely effed. The AI trade is over. We all top-blasted and holding bags now. Tech investor Steven Sanofsky added, This is not the AI trade or AI-lost-the-election trade, as you might read or see, but rather the what-happens-when-USAI-is-not-going-to-happen trade. In fact, one of the most striking parts of the drawdown was how broad-based it was. Completely unrelated stocks like Walmart and ExxonMobil took massive hits alongside tech. Some argued that semiconductors are now cheap, with Nvidia trading at a 13x multiple, making it one of the quote-unquote cheapest stocks on the market. Then again, investor Jeff Park that boom-and-bust semiconductor stocks always look cheap as they head into a bust, commenting, Nothing outs a newbie faster than valuing semis on price-to-earnings. Now, the curious part is that nothing substantial changed between Wednesday and Thursday that would explain the sudden collapse. And whenever that happens, when there's no narrative explanation, we have to look towards market structure and market dynamics as the driver. And indeed, Fundstrat's Tom Lee told CNBC, I think there are stories of some large head funds actually being liquidated now. And according to reports, Walleye has blown up and returned their capital. Lone Pine is also rumored to have collapsed, and Millennium is said to have pulled $1.5 billion in leveraged capital from two outside managers. Bloomberg's Nishant Kumar wrote, I've covered hedge funds for nearly 16 years, and I can't remember a day like today. Nicholas Mugali from TMT Research wrote, If you're wondering why this sell-off is so aggressive, this is what forced selling looks like. It isn't happening in one sector. It's literally everywhere you look at once, from financials to AI, infrastructure, and the best businesses are falling hardest because they're the easiest things to sell when a funds need cash. The fundamentals are intact, but every chart looks broken right now and has been broken for a couple of months. In short, rather than investors waking up on Thursday and deciding to sell everything, it looks like Wednesday's drawdown in semiconductors was enough to trigger a broad deleveraging across the hedge fund industry. That could mark the end of the AI bull market, or it could be a temporary blip like we saw when tariffs were unveiled last year. Now, with markets in freefall, rumors are swirling that the Anthropic IPO might get pulled. Previous reports said that Anthropic was aiming to get out before Thanksgiving, which means their roadshow would need to begin next week. Obviously, a frontier AI lab is going to be a harder sell under these market conditions, and prediction markets are beginning to price in a delay or even a cancellation. On Thursday, odds of the Anthropic IPO before Thanksgiving fell from around 85% to 60%, and the odds they get out by the end of the year have fallen from 95% to 66% over the past week. Thinly traded secondary markets are still pricing the AI bull market, pricing the company at around $1.6 trillion. But according to the Wall Street Journal, institutional investors are increasingly skeptical of the price. They wrote that three major investors who plan to participate in the IPO say the valuation will need to come down, with one arguing it needs to be closer to $1 trillion. Either that, or Anthropic might need to trim their reported $100 billion target for the fundraising. This is obviously a big shift from where they were in August, when investors were leaking a $2 trillion target in the press. Still, that same article in the WSJ reported an $80 billion run rate for October, up from $65 billion reported for the end of July. So can Anthropic slough off the market turmoil? I think we'll have a better sense once the hedge fund blow-up stops ricocheting around the market, but for now, something that was idle chatter is getting a lot more attention. Now, speaking of Anthropic, outside of markets, maybe the biggest discussed story right now is Meter's release of their first independent safety report for Anthropic, which discloses a major acceleration in AI-led research. Last month, Anthropic announced that they'd onboarded Meter as their first independent with employee-level access. The idea is, of course, to give the public better visibility into Anthropic's safety practices in the wake of so-called rogue agent incidents. The key finding of the report was that the pace of improvement for Anthropic's models has roughly doubled over the past year. Meter concluded that the acceleration was due to AI being used in research efforts. They wrote, The report stated, This could lead to a sudden emergence of extreme and risky capabilities. Meter noted that they were given access to internal information that surpassed their expectations. This included analysis of three unreleased models trained since July. The most recent model was completed in September or October and continued the trend of accelerated improvement observed earlier in the year. Anthropic co-founder and responsible scaling officer Sam McCandlish signed up for X this week to discuss the report. He commented, The speed of progress in model capabilities is one of the most important questions in AI safety, and Meter's work here provides the public with an independent, external view into how this is happening inside Anthropic. Sam acknowledged criticism that Anthropic redacted some of the most important evidence that Meter used to come to their conclusion, but said that Anthropic intends to provide a supplemental report on their own analysis in the coming weeks. Now, in terms of reactions, the AI safety folks were rattled by the report, claiming it accelerated their timelines. Ryan Greenblatt from Meter wrote, The automation is a big update towards ASI within one to two years. Nathan Calvin, the general counsel for Encode AI, called this a, quote, extremely unnerving update. New York Assemblyman Alex Boras argued that there's no time to wait, writing, Congress should ban RSI this year if possible. But others were a little skeptical of the methodology, with open model researcher Nathan Lambert noting, A lack of good enough benchmarks is a bottleneck, at least for the type of RSI described as an intelligence explosion. Now, some rehashed the criticism that Meter is closely affiliated with Anthropic, at least in ideology, if not through actual financial and personnel ties. And whatever you think of those arguments, the report does highlight the issue with presenting Meter as an external evaluator. One of the key points of appointing Meter was to address public concerns about cybersecurity breaches from the labs. Meter's report, though, only mentions cybersecurity in terms of the models getting more advanced, and doesn't touch on sandboxing or other safety measures. It's essentially non-responsive to concerns raised over recent months. Instead, assuming that RSI is the only risk that matters. Now, it's perfectly reasonable for Meter to decide that that is the only concern that they really care about. But if that's the case, having only Meter here means a very incomplete independent auditor suite. Now, closing out with some feature and product updates, Anthropic has relaunched Claude for personal finance with the ability to connect to your bank account. This is read-only access via Plaid, which is common across the personal finance industry, so theoretically shouldn't be a huge concern to the big banks that are worried about agents making transactions. Claude won't even give investment advice, instead giving users help with budgeting by analyzing their spending data. Giving a sense of the current state of how they talk about AI's potential disruption of professional services, they wrote, Personal finance is a first step. For life's bigger money decisions, we think people are best served by Claude working alongside the licensed professionals and regulated firms they trust. Over in Meta land, many are rejoicing that Muse now has a voice. Meta's breakout personal agent now has a feature that allows users to communicate with it through voice and video chats. To start, Muse has leaned into the idea of personalizing your agent by giving it a name in an avatar. This feature takes that idea to the next level, fully animating the avatars for voice calls and letting users choose from a range of different voice options. AI CEO Alexander Wang showed off the new feature, demonstrating his agent Euler, coaching him through problems at work related to posting too many memes. The feature uses Meta's live voice models, which are top-notch, as well as their real-time avatar technology. Meta says the avatars will use subtle expressions and gestures to match the tone of the conversation, but they don't actually have visual inputs to pick up on non-verbal cues. The feature is now live for mobile, web, and Mac. use it to power the forthcoming Muse Charm device. It also works with the Muse Gadget SDK, so you can now talk to your homebrewed AI devices as well. And lastly, OpenAI has rolled out a new version of their image model that now supports animated GIFs. Users can now prompt a short animation, edit the speed, and modify individual frames before they share it online. OpenAI says that their users are now creating 4 billion images per week, up from 3 billion a couple of months ago. Images 2.6, which is the name of the new version of the model, is 50% faster and able to generate an image in as little as 6 seconds. It can also now accept reference images, meaning users can upload a rough doodle or easily turn it into a high-quality image. And so, if you didn't already have plans for the weekend, maybe it's time to experiment with making your own GIFs. For now, that is going to do it for the headlines. Next up, the main episode. Hello, everyone. One big change around AI, is we've shifted our thinking from how we rank our pages to how do we become the source that AI trusts enough to answer with. At KPMG, they're seeing this firsthand. AI-generated results now surface answers directly, often without a single click. That's why they are increasingly focused on Generative Engine Optimization, or GEO, structuring content so AI systems can retrieve it, understand it, and cite it as trusted authority. This is not just an SEO evolution, but a visibility mandate. And indeed, the GEO mandate from KPMG is simple. If AI is shaping the world, it's going to be a reality. And if it's not, then it's going to be a reality. If AI is shaping decisions, your expertise needs to show up inside the answer. Read all about it at kpmg.com slash US slash GEO. Again, that is kpmg.com slash US slash GEO. This episode is brought to you by CDATA. Companies rolling out AI keep hitting the same wall. Token spend is climbing, security is holding agents back, and no one knows how AI is truly being used. The root cause? AI is cut off from the systems and context your business runs on. That's why CDATA built Connect AI Gateway, one control point between your AI and your systems. Built on the connectivity layer powering Palantir, Google, Salesforce, and hundreds more, each agent gets exactly the data and tools it needs. It governs down to the record and lets you finally see your org's complete AI usage in one place. Accurate answers, a fraction of the token cost, and a gateway that learns your business with every interaction. See how at cdata.ai. You're in back-to-back meetings all day. You're trying to stay present, but you're also worried you'll forget the decision, the action item, the important next step. This is a scenario that we absolutely live here at the AI Daily Brief, given how fast things move and change and how quickly we're evolving everything that we do. That's where Granola comes in. Granola is an AI-powered notepad for meetings. You jot down rough notes like you always do, and in the background, Granola transcribes and turns them into clear, useful notes when the meeting ends. No bots joining your calls, no distractions, just a clean notepad that helps you focus. During or after the call, you can chat with your notes. Ask Granola to pull out action items, help you negotiate, write a follow-up email, or even coach you using recipes, which are pre-made prompts. Once you try it on a first meeting, it is very hard to go without. Head to granola.ai.ai.ai daily and use code AI daily to get 100% off for three months for new users. That's granola.ai.ai.ai daily and code AI daily to get 100% off for new users for three months. Blitzy's understanding of massive code bases unlocks autonomous security fixes, modernization, and new features. So, what happens when there's no legacy code at all? Greenfield is supposed to be the easy part. Clean slate, no technical debt. But even Greenfield moves at human speed one sprint at a time. Blitzy changes the unit of work from the developer to the project, autonomously planning, building, testing, and validating entire applications from scratch. Hundreds of thousands of lines of production-ready code. One Blitzy customer stood up a brand new application, 534,000 lines of code, compressing a 65-week roadmap into two weeks. Another shipped an entire application with no front-end engineer. Legacy or Greenfield, the answer is the same. Software at the speed of compute. Build what's next at Blitzy.com. That's B-L-I-T-Z-Y dot com. Welcome back to the AI Daily Brief. Today, we are digging into Menlo's latest enterprise AI report. They've been doing these state of enterprise AI reports for three years now, and they provide a really good snapshot into how businesses are using AI. Now, the data for this survey comes from a survey of around 400 U.S. enterprise decision makers that was conducted by Morning Consult in mid-September. So this is pretty fresh data. Those enterprise decision makers, by the way, are responsible for about $4 billion in AI spend. So it's a pretty meaningful-sized group. Now, in terms of caveats, of course, Menlo is not some unbiased observer. They are heavily invested in the AI field, including in Anthropic and many other companies. But as we'll see, this data all directionally backs up, I think, a lot of common sentiment about business AI spend right now. What we're going to do on today's show is not go through every single number in the deck. Instead, I've pulled out the 10 most interesting, each of which I think has something else to tell us about the AI market. The first set of stats are all about how much business AI spend has grown. And what's interesting about 2026 is that not only has spending gone up, the rate at which it is growing has also gone up. Across the four years, excuse me, I think I said three years before, but the four years that Menlo has been doing the survey, estimates for enterprise AI spending have jumped from $1.7 billion annually in 2023 to $8.5 billion in 2024 to $33 billion last year to almost $177 billion in 2026. Now, from a growth rate perspective, that means that the growth rate from 23 to 24 was 5x, but the growth rate from 24 to 25 was down to 3.9x. Now, this is kind of what you'd expect as a market gets more mature, more saturated, and the total size of the numbers go up, while the sheer amount of spend can continue to increase. The multiple of the previous year's spend at some point has to start coming down. That's sort of what it looked like happened between 24 and 25. Except now in 26, we've had a jump not only ahead of the 25 number, but a faster growth rate than even the 23 to 24 number. And obviously, the explanation is agents. This year, the big story of 2026 from Opus 4-5 in November of last year on is that AI spend stopped being about seats and started being about tokens and everything that intelligence could do. And that has totally transformed our expectations of how big the market can get. This number also shows up in our next stat, where breadth of AI spend was up only incrementally, but depth of AI spend was up significantly. So the share of enterprises using AI jumped from 55 to 65% between 25 and 26, which is insane that that still means that 35% of enterprises report not using AI. But here we are. If you've ever doubted the inertia of the business world, there you go. However, the average AI spend per enterprise using 2025 as 1x is up almost 5x, 4.7 times more per enterprise on average this year. Again, the answer is agents. Agents that can work independently, that can run in parallel, that can keep working after people step away. Spending then scales with work, not with headcount. Now, one thing that we've been debating on this show, and in markets, frankly, is whether the concentration and top heaviness of AI spend is a bearish or bullish sign, the bearish sign, at least from a market's perspective, is that when big chunks of revenue are concentrated in a small handful of customers, if any one of those customers stops spending or changes their spending habits significantly, it can have an outsized effect on the entire market, which could then have knock-on effects that drag the whole market down. This is why concentration is a risk. The other way to look at it, however, is that the concentration is not a natural state of the market where there are only a handful of companies that can use that those handful of companies that are using that much AI are leading indicators for where the rest of the market will eventually go. That is, of course, the bullish assessment that takes what those companies at the very top are spending and multiplies that across all the other companies that could possibly spend that much as well. Menlo certainly reads this 5x growth in average spend per enterprise as a sign of future demand. They believe that the long tail basically is barely getting started. And one more on growth, and this is something that certainly we've been seeing companies reconcile with all year. Enterprises spent about 2.5 times what they had budgeted this year. Basically, the average company planned to roughly double spend between 25 and 26. They had planned to grow spend about 1.9x. So if you spent $10 million in 2025, you were anticipating that you'd spend $19 million in 2026. Those numbers were, of course, calculated by most firms on the basis of increased adoption on a per seat basis across more parts of their companies. In actuality, spend has jumped 4.7x around 2.5 times more growth than they had budgeted for. And the challenge, of course, for CFOs is that this isn't necessarily some out-of-control spend. It's people flexing and understanding what they can actually do with all this raw intelligence power. The whiplash that companies have had between token maxing on the one hand and token restrictions on the other are all symptomatic of the fact that we planned for one paradigm, a per seat paradigm, and instead got a per token paradigm, which needs a totally different type of strategy. Budgeting for AI is starting to look more like budgeting for cloud compute, but even that analogy falls down as well. It really is this new category of thing that doesn't cleanly fit in past buckets, and so needs to be considered on its own new terms. For our next section, we'll look at what Menlo found about the competition and where those enterprise dollars were going. One story that kind of flies in the face of some of the conventional wisdom, or at least some of the narratives for the past few months, is that overall, buyers moved to more expensive models. The most popular models a year ago were CloudSonnet 4.5, which cost $3 per million input tokens and $15 per million and GPT-5, which was $1.25 and $10 respectively. The most popular models today are, unsurprisingly, the latest generation from each of those companies as well, Claude Opus 55, which is $4 input and $20 output, and GPT-6 Astra, which is $10 and $60. The frontier model's spend of share went from 35% to 63%. Meanwhile, every other category, mid-tier, small, and older, went down as a percentage of total spend. Now, part of this is also small model price wars. Labs cut small model prices by 80% plus. But the story that this survey is finding is that buyers still, despite all of the narrative about cost and cost efficiency, are choosing capability over cost. So is the narrative that we've had about cost wrong, or is there room for both of these stories simultaneously? I actually think that these two things are not mutually exclusive. I believe that the continued emphasis on frontier models reflects the fact that starting with Opus 45, we really have had a differentiated set of capabilities that came online, and since the entire year has been businesses racing to catch up with those capabilities and implement new agent decoding, they've sort of been priced in, at least initially, to taking advantage of the frontier models. Opus 55 feels like another one of those jumps where we're uncovering a bunch of use cases that really are significantly different with Opus 55 and aren't as easily translated to other models. I don't think that that changes the fact that there is still going to be fierce and extremely important competition. I don't think that that changes the competition for those cheaper and more efficient models. I think we're still in an early adoption phase where adoption is necessarily going to be more heavily concentrated in the state of the art, but as that adoption matures and the stack matures along with it, you will see those smaller, more efficient models start to come in and form a more complex stack. If anything can be said, it's that perhaps our narratives and storytelling are getting out ahead of where businesses actually are, and in that way, this is a good reminder that the things that we're discussing on the AI Daily Brief, and the things that we're discussing on the AI Daily Brief, are going to be more within heavily enfranchised AI circles, don't necessarily reflect the lived reality right now of the average business buying and using AI. Now when it comes to the competition, Anthropic has continued its ascendancy. Back in 2023, Anthropic had about 12% of business API spend. In 24, it jumped to 24%, and by last year it was up to 40%. Now it has crested over the halfway mark at 55%. And if you are once again seeing the fingerprints of agents here, you are not wrong. The Opus models, every single one of them, are going to be more and more different. Ever since Opus 4.5 have been perhaps the default for long-horizon agentic encoding tasks, and that's now showing up in the numbers. Now OpenAI has throughout the same time, in general, seen a decline. They went from 50% of spend in 23% to 34% in 24% to 27% in 25%. Last year, in other words, was the first year that Anthropic actually flipped OpenAI. This year though, we did see a slight recovery. OpenAI is now up to 31% of enterprise API spend, and Menlo's read is that this represents a strong second-hand half-of-the-year recovery. The GPT 5.6 series of models, including Sol, Terra, and Luna, coming in July, as well as GPT-6 in September, and just in general, OpenAI's emphasis throughout 2026 on enterprise and business use cases is clearly bearing fruit. Now the main company that suffered and allowed Anthropic and OpenAI to both win more market share is Google. Google was down from 21% of enterprise spend last year to just 9% this year. And again, this is just the are. While Gemini 3 Pro was at the frontier in November before Opus 4.5 came out and agentic coding took over everything, that was the last time Google could reasonably be considered in a three-way race for the top of the heap when it comes to model capability. Perhaps that is changing now with Gemini 4 Carbon. Early indications look good, although it remains to be seen and needs a little bit more battle testing for us to claim that. But at the moment, when it comes to enterprise AI spend, this is a completely two-horse race between Anthropic and OpenAI. The near-frontier models are the ones that are going to win. They're going to win. Still haven't really broken through for as much attention as they get. Grok 4.7, Kimi K3, DeepSeek V4, GLM 5.3, all together still account for under 5% of API spend. Although, as we'll see when we get to open source, when it comes to understanding open and cheaper models, it's worth looking not only at percentage of spend, but percentage of tokens. Because it turns out that while open source only gets 4% of spending, it represents about 19% of tokens. Interestingly enough, as a percentage of spend, open source is actually down. It was at 6% in 2025 and is at 4% this year. However, as a percentage of tokens, it has jumped from 11% to 19%. I would argue that in the medium to long term, the token share numbers are more significant than the spending numbers. Like I argued before, I think right now we're still in an early adoption phase where companies are concentrating around the most expensive models to bring these new capabilities online. But the fact that you're looking at open source models, you're looking at open source models, you're looking at open source models, you're looking at open source models, you're looking at open source models. And I don't think that that's a trend that's reached its peak. Now, what this says for the total addressable market value of all tokens is an interesting debate and one that happens on Wall Street all the time. There are plenty of smart people who think that in the long run, open source models represent the vast majority of tokens used, but still a relatively small percentage of overall spend, which is concentrated among the most important and elite use cases driven by the state-of-the-art closed models. By the way, when it comes to open source spend, Kimi was the dominant force that was used to build open source models. So, I don't think that's a trend that's reached its peak this year, representing 61% of open source spending. Next up, we move to different categories of usage within the enterprise. One of the most staggering numbers, not unexpected, but still staggering in its magnitude, was the revenue growth of the lab coding agents, Codex and CloudCode. That revenue jumped from $1 billion at the end of 2025 to $24 billion at the end of 2026. Like I said, not surprising given how much CloudCode and Codex have defined the experience of AI users throughout the year, but still, 24x growth in revenue in a single year is just astronomical. Now, what's interesting is that this didn't necessarily come at the cost of other coding platforms like Cursor, Cognition, Factory, Lovable, etc. Their revenue still grew almost 3x this year as well, from $2.6 billion last year to $7.8 billion now. It's just that the lab coding apps were by far the more significant. Now, when you start to go category by category, you might expect then that codecoding is going to be the most important thing in the world. Codecoding grew faster than any other category when it came to app spending. And you'd be almost right. Agentic coding apps grew at a 7.7x rate, ahead of everything except financial services, which grew at an 8x rate. Now, the total spend on financial services apps was a tiny fraction of coding apps. Overall, Menlo estimates financial services apps saw about $800 million in spending compared to the $34 billion for coding apps. But still, their speed of growth is an interesting trend. Some of the other notable areas of growth include customer success, which grew 4.3x, sales apps, which grew 3.6x, legal, which grew 2.8x, and down at the bottom, marketing, which grew just 1.8x. It's tempting, once again, I think, to read this chart as, at least in part, emphasizing the fact that applications in verifiable domains where the power of agentic coding could be brought to bear, even if it wasn't for building apps, were the ones that benefited the most from the developments of 2026. Now, when it comes to where codecoding comes from, and which companies are being chosen to deliver the AI, there is a really strong divide between AI-native companies, by which, of course, I mean companies like Anthropic and OpenAI, and the incumbents who are fast bringing AI into their services, such as ServiceNow. When it comes to those coding applications, they are dominated by AI-native companies. AI-native companies, in fact, represent 96% of that spend. Finance and operations is also up there, where AI-native companies represent 95% of spend, and AI-natives also rule the design of AI. So, as a result, AI-native companies are dominated by AI-native companies. So, as a result, AI-native companies representing 81% of spend. Sales is nearly split in the middle, with 54% of the spend going to AI-natives versus 46% to incumbents. But then every other area, marketing, customer success, HR, data science, and IT, all see the incumbents getting more than 50% of market share. Basically, big chunks of enterprise spend has wanted to flow towards the solutions they already use, and already have embedded, and already have connected to their systems of record, or which are their systems of record, rather than ripping that out and building something else. So, as a result, AI-native companies are dominated by AI-native companies completely new. It makes sense, then, on the other end of the spectrum, that agentic coding, which is a net new capability, doesn't yet have an incumbent solution that's plugged into all the systems, so it has room for a new solution to come in. Compare that to IT, for example, where many companies have been using the same vendor or partner for years, if not decades, and it makes sense, perhaps, then, that only 9% of AI app spend for IT is coming from AI-native companies, compared to 91% that's going to incumbents. So, what does this all mean for software? Another question that lurks, particularly in the minds of investors who are trying to figure out how AI plays out, is whether AI just eats SaaS. One of the big stories of 2026 was, of course, the SaaSpocalypse, where for a while, at the beginning of the year, as agentic coding came online, SaaS stocks were just absolutely decimated, as people operated on the idea that the ability to roll your own apps or spin your own apps was going to fundamentally undermine the use of those SaaS tools. That's not exactly the story that Menlo found, however. While Menlo'd did find some diminished growth of the SaaS application market, from 13% growth in 2025 to 11% growth in 2026, they found that the majority of new AI spending actually came from net new budgets, not existing software budgets. Of the AI spending that could be accounted for by Menlo, 21% came from existing software budgets. That compared to 55%, which was net new budgets. Now, that's obviously positive news for SaaS investors and SaaS companies, although I do think we've gotten started on the next narrative which threatens SaaS, which is the rise of agents using software and headless software. and all it takes is reading some comments from Salesforce to understand how significant that disruption could be as well. For now, however, it looks like a lot of AI spend is net new, not just cannibalizing spend that already existed. So that is Menlo's state of enterprise AI. Like I said at the top, I think that this broadly reflects a lot of what we've been discussing and seeing throughout the show, although it is a good reminder that there remains a gap between our narratives that we explore every day and the trends that we see on the horizon and the lived experience that companies operate within every day. For now, that is going to do it for today's AI Daily Brief. Appreciate you listening or watching as always, and until next time, peace.
PodcastTranscript AI