The AI Daily Brief: Artificial Intelligence News and Analysis - How One Company Saved 213,000 Hours with AI

Starting point is 00:00:00 Today on the AI Daily Brief, a case study of one company that made AI mandatory and saved 213,000 work hours in a single year. Before that in the headlines, a major jump in token usage that suggests a broader inflection point in AI growth. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, welcome back to another AI Daily Brief. Quick announcements before we dive in.

Starting point is 00:00:28 First of all, thank you to today's sponsors, KPMG, Blitzy, and Vanta. And to get an ad-free version of the show, go to Patreon. slash AI Daily Brief. Also, if you were interested in sponsoring the show, shoot me a note at NLW at Breakdown.network. But with that, let's talk about this big, big jump in token consumption and why it might be an even bigger deal than it seems. Welcome back to the AI Daily Brief Headlines edition. All the daily AI news you need in around five minutes. There is definitely a sense that even in an extremely fast-moving AI space, the last couple of months have seen another big phase up. we actually have some evidence that that's the case. On this week's earnings call, Google CEO

Starting point is 00:01:09 Sundarpe Chai revealed that the company is now processing 980 trillion monthly tokens across their products and APIs. And what makes that number impressive for more than just the fact that it's almost a quadrillion is how much it's grown since just May. Back at Google I.O. in May, they were processing 480 trillion tokens. That is 104% growth in just a couple of months. And one thing I think that's important to note about that, is that a huge amount of that usage is, of course, people who are building other types of AI experiences, which means that presumably all of this use builds on itself and will move even faster going forward. Now, unsurprising, the entire big theme of Google's earnings call was AI. Although, interestingly, it was a lot about analysts who

Starting point is 00:01:55 were concerned with AI cannibalizing various parts of Google's business, which I said in no uncertain terms, though, AI is positively impacting every part of the business. He said that features like AI overviews and AI mode are performing well, and that despite analysts' fears of AI disruption, Google's search by itself is bringing in $54 billion and still rising. Indeed, revenue jumped 14% overall, reaching a $96.4 billion quarterly pace, which makes their 10 billion or 13% CAPEX projection increase a little bit more tolerable to investors. Now, this was the first earnings call where we got solid user numbers for Google's AI search products. We all also got on this call some pretty actually solid user numbers. Gemini app users have grown to

Starting point is 00:02:35 250 million active users with daily requests increasing by 50% since Q1. And while they tried to position their CAPEX expansion as just keeping up with demand, one Forrester analysts said, Google's hand is forced by OpenAI to spend tremendously on AI's infrastructure and applications. I don't know that I think that they're being quote unquote forced in any way. I think that Google has for a very long time seen more or less its entire future in this AI space and has really hit its stride in the Gemini 2.5 era. More than 100% growth in token usage in just a couple of months tells a big part of the story right there. And speaking of OpenAI, Pichar actually kind of dropped a little bit of a bomb, disclosing a partnership with the company that had gone under the radar.

Starting point is 00:03:17 He told investors, with respect to Open AI, look, we are very excited to be partnering with them on Google Cloud. Google Cloud is an open platform and we have a strong history of supporting great companies, startups, AI labs, etc. So super excited about our partnership there on the cloud. outside, and we look forward to investing more in that relationship and growing that. It turns out that OpenAI models have been quietly added to Google Cloud earlier this month, making them the third provider alongside Oracle and Microsoft Azure. Basically, when you're playing this big with this many product lines, you can't be anything other than frenemies.

Starting point is 00:03:48 Speaking of earnings calls, Elon Musk was careful not to encourage discussion of an XAI investment during this week's Tesla earnings call. When asked about a potential XAI investment, he responded, shareholders are welcome to put forward any shareholder proposals that they'd like. Tesla's CFO added that this is, quote, not the forum to discuss the topic. Now, Musk has, of course, been winking at a Tesla deal recently, as XAI looks basically everywhere for investors. SpaceX has been tapped for $2 billion, and XAI is reportedly looking for another $5 billion in debt funding at the moment. Tesla has around $37 billion cash on hand, so could smooth over XAI's capital needs as they

Starting point is 00:04:24 build out more compute. However, as Musk doesn't have a controlling stake in the company, he doesn't get to make that decision. So he's been calling on shareholders to put together a proposal. Earlier this month, he posted, It's not up to me. If it was up to me, Tesla would have invested in XAI long ago. We will have a shareholder vote on the matter. And yes, while it's not up to him, the seed has definitely been planted among the Tesla faithful. Lastly today, AI coding platform Loveable has become the fastest software startup ever to hit $100 million in revenue. Founded just eight months ago, Lovable has beaten out Cursor and Whiz. And of course, while their rival Replit also reached 100 million in ARR over that same time space, Replett had fought as a smaller startup for eight long

Starting point is 00:05:04 years before becoming a overnight success. To many, Lovable is an iconic representation of what can be achieved with a lean team during the AI era. They have just 45 full-time employees in another 14 open positions, which makes for a pretty impressive revenue per employee ratio. They also seem to be monetizing customers extremely efficiently. Lovable claims 2.3 million active users, but only something like 180,000 paying customers, meaning that each customer is spending a good chunk of money over $500 in annual revenue. Now, growth is apparently continuing to be incredibly strong. They've reported a $75 million run rate in June, meaning they've tacked on another 30% in a month. Alongside the announcement, they also introduced a new agent design intended to be

Starting point is 00:05:45 much better at thinking and tool use. They said that the agent has 91% fewer errors, meaning that to quote CEO Antonosica, it should feel like you're now working with a scene. your developer. Now, one thing that's been really interesting is that I've seen a few people shade this announcement either directly or surreptitiously. I've seen a lot of people say, use caveats like reportedly or supposedly when re-quoting this $100 million number. And I also saw this post from Greg Eisenberg that read, I think within the next 24 months, we're going to witness an AI company that was one of the fastest growing companies of all time and realized that revenue was fake. Not predicting who that is, just think it's inevitable. Now, if Greg isn't talking

Starting point is 00:06:22 about lovable, he certainly timed that tweet awkwardly given their announcement. I don't know, man. As a lovable user, it's pretty easy for me to believe, given that about $99 million of those are probably mine. With that in mind, I will simply say congratulations. If you haven't tried lovable yet, go check them out. But that is going to do it for the headlines. Next up, the main episode. Today's episode is brought to you by KPMG. In today's fiercely competitive market, unlocking AI's potential could help give you a competitive edge, foster growth, and drive new value. But here's the key. You don't need an AI strategy. You need to embed AI into your overall business strategy to truly power it up. KPMG can show you how to integrate AI and

Starting point is 00:07:03 AI agents into your business strategy in a way that truly works and is built on trusted AI principles and platforms. Check out real stories from KPMG to hear how AI is driving success with its clients at www.kpmg.org.us slash AI. Again, that's www.kpmg.comg.com slash AI. This episode is brought to you by Blitzy, the Enterprise Autonomous Software Development Platform with Infinite Code Context. Blitzy is used alongside your favorite coding copilot as your batch software development platform for the Enterprise seeking dramatic development acceleration on large-scale code bases. While traditional co-pilots help with line-by-line completions, Blitzy works ahead of the IDE by first documenting your entire code base, then deploying

Starting point is 00:07:47 over 3,000 coordinated AI agents in parallel to batch build millions of lines of high-quality code. The scale difference is staggering. Copilot's might give you a few hundred lines of code in seconds, but Blitzy can generate up to 3 million lines of thoroughly vetted code. If your enterprise is looking to accelerate software development, contact us at blitzie.com to book a custom demo or press get started to begin using the product right away. As a founder, you're moving fast towards product market fit, your next round, or your first big

Starting point is 00:08:16 enterprise deal. But with AI accelerating how quickly startups build and ship, security expectations are higher earlier than ever. Getting security and compliance right can unlock growth or stall it if you wait too long. With deep integrations and automated workflows built for fast-moving teams, Vanta gets you audit-ready fast and keeps you secure with continuous monitoring as your models, infra, and customers evolved. Fast-growing customers like Langchain, writer and cursor, trusted Vanta to build a scalable foundation from the start. And look, as someone who lives in the world of enterprise procurement,

Starting point is 00:08:48 I love how Vanta makes it easy to get compliance right. The last thing you need when you're trying to win that big deal is to have it scuttled by something that Vanta has solved for over 10,000 companies. Go to Vanta.com slash NLW to save $1,000 today through the Vanta for Startups program and join over 10,000 ambitious companies already scaling with Vanta. That's V-A-N-T-A-com slash NLW to save $1,000 for a limited time. Welcome back to the AI Daily Brief. Today we are doing something just a little bit different.

Starting point is 00:09:19 Obviously on this show, we talk a lot about how people are actually getting value out of AI right now, or rather we try to put the theoretical in the context of real life as much as possible. And so I was really interested to see a case study recently from Norgasbank, which is Norway's sovereign wealth fund. What we're going to do today is use this case study as a jumping off point for sharing some other information around where enterprises are with AI adoption and what they're still struggling with. Norgas Bank Investment Management was set up in 1998 to manage Norway's oil wealth. At the time, the fund had around $14 billion entirely invested in bonds, but today the fund

Starting point is 00:09:56 is worth around $1.8 trillion. In an interview I saw recently, they said that that was about 70% equities and 30% fixed income in bonds, and whatever the case, it's enough that it has become the world's largest sovereign wealth fund. The fund represents a little over $300,000 for each Norwegian citizen. The goal of the fund is to attempt to maintain a portfolio that captures global asset exposure, which is no small feat given how many different options there are. And that's made all the trickier given the fact that it's just a 670-person team. Fund CEO Nikolai Tangan said that ever since 2022, he's been, quote, running around like a maniac trying to convince his staff to

Starting point is 00:10:32 adopt AI. Last year, however, the firm started approaching the question of adoption more systematically. And that's what makes them an interesting case study. The first lesson has to be about the buy-in of leadership. This was not something where the CEO had to be convinced. The CEO was the person pushing the policy. And yet, while obviously executives need to be bought into enterprise AI transformation and agentification, they need to do more than just talk about it to get people bought in. One of the big challenges for companies right now is that there is often a disparity between how leaders and how frontline employees see AI. Redder did an enterprise AI study in December, which surveyed 800 employees and 800 C-suite executives, and showed that there

Starting point is 00:11:13 was a big gap between how the two were seeing their company's AI efforts. For example, while 73% of the C-suite executives said that they thought their approach to AI was well-controlled and strategic, only 47% of employees did. The disparity was even higher when asked whether their company's AI adoption had been successful in the previous 12 months. 75% of C-suite executives thought it had, where only 45% of actual employees did. And this is not the only study to find something similar. In Microsoft's 2025 WorkTrend Index, they found this type of gap as well when it came to a number of criteria that Microsoft was using to try to understand who is in the mindset of

Starting point is 00:11:51 agentic change. While 67% of leaders they surveyed were familiar with agents, only 40% of employees were. while 69% of leaders regularly used AI, only 45% of employees did. On every question, trusting AI for high stakes work, using AI as a thought partner, seeing AI as a career accelerator, and a number of others, leaders were ahead of their employees. The TLDR is that it's not enough to have an AI strategy, and it's not even just enough to communicate it. You have to get people bought in. Or you just tell them they don't have a choice, apparently. That's what the Norway sovereign wealth fund did, said Tangan in an interview.

Starting point is 00:12:27 it can't be voluntary. It isn't voluntary to use AI or not. If you don't use it, you will never be promoted. You won't get a job. However, it wasn't just a mandate. Norges Bank also took the time to actually create structures that could help employees with this mandate. They created a six-person AI enabler team, 40 AI ambassadors across the organization, and had repeated seminars, conferences, and courses. Essentially, they made it as effortless as they could to access an AI leader within a particular team or find the right AI training to figure out the next step. A year in, Tangan is definitely convinced that the mandate was essential. He said my biggest surprise was that resistance when we started. People don't want change. There's always 10 to 20% who don't want to do things if it's voluntary,

Starting point is 00:13:10 but those are the ones who need it. Now, one big blocker for the bank was that workflows were already set in place and difficult to disrupt. Analysts were used to doing things a certain way and had no guarantee that automations would work. They made the determination then that putting the responsibility on individuals to reinvent their workflows piece by piece and on their own wasn't going to work. Instead, they needed an organization-wide effort. BcG recently published a study where one of the questions for people who worked in companies that were undergoing AI transformation was about exactly what their companies were actually doing. 72% said that their companies were deploying Gen. A.I. Tools basically rolling out things like co-pilot to increase productivity. 50% said that

Starting point is 00:13:50 their companies were redesigning end-to-end workflows and processes to reimagine functions, and just 22% said that they were building and innovating new business models and products to drive growth. You've frequently heard me talk about the difference between efficiency AI and opportunity AI, which is of course not an argument that companies shouldn't be excited about efficiency and productivity gains, just that they shouldn't see that as the be-all end-all. It sounds like what was going on inside the Norgas Bank was that they were not content to simply deploy AI. They wanted to get into this type of redesign of workflows. The organization partnered with Anthropic to power their AI transformation, with the first major goal of making their data more accessible. They integrated

Starting point is 00:14:28 clawed into their Snowflake Data Warehouse, and Tangan said, our portfolio managers and risk department can now seamlessly query our data warehouse and analyze earnings calls with unprecedented efficiency. Instead of needing SQL or SQL expertise, analysts could now query the database in natural language. Unsurprisingly, data remains one of the major challenges for enterprises undergoing agendic transformation. The Economist Impact recently found that only 22% of organizations said that their current architecture was fully capable of supporting the unique demands of AI workloads. They also found a huge number of challenges with data ranging from access control to privacy protection, to data silos, and much, much more. One really interesting theme that we're seeing

Starting point is 00:15:09 right now at Superintelligent as we talk to organizations is that the rise of model context protocol or MCP is actually making some of these questions about data feel more accessible even to non-technical populations. MCP is basically a way to pre-wire different data sources and make it accessible to agents and LLMs in a standard and unified way. This means, for example, that when an organization is designing or building an agent and trying to connect it to a particular data source, they don't have to do that work from scratch. They can simply connect it to an existing version of what's called an MCP server, allowing them to move much more quickly. Now, this is, of course, only one small part of the data work going into enterprise transformation, but it's a clear, important, and accessible piece that, as I said,

Starting point is 00:15:51 makes non-technical employees feel like they have a stake in an understanding of what's going on on the technical side. Another big automation project was in monitoring news and analyzing earnings calls. Norges Bank owns stock in thousands of companies across the globe, and they all report earnings every quarter, to say nothing of the news that happens in between. With Claude, they managed to automate all of the analysis. They generated transcripts from audio and used the chatbot to generate key insights, removing thousands of hours of tedious work. Claude also began to find patterns in the decision-making around earnings calls. By pointing these patterns out to analysts, they were able to recognize cognitive bias that caused suboptimal

Starting point is 00:16:24 decision-making. Another interesting use case was in analyzing executive compensation. Given that it's a large shareholder in many major companies, they often get a deciding vote on things like how CEOs should be compensated. In one high-profile example, the company opposed Elon Musk's $56 billion Tesla package in 2024 with Claude helping make the decision. In fact, after testing the results against human decision-making, Norges Bank found that Claude lined up with 95% accuracy. A year into the process, Tengen said that Claude has become indispensable. He says that the company has seen 20% productivity gains saving them some 213,000 hours a year. Now, for those looking to replicate this success, one thing to note is that some of the lessons learned have been embedded,

Starting point is 00:17:05 presumably into Anthropics recently launched Claude for Financial Services. This is the first verticalized application of Claude that Anthropic has launched, but it likely will not be the last. In the age of agents, all of the trends that are on display in this Norway sovereign wealth fund example are going to get nothing but louder. Every single study out there finds that people and companies just don't feel they have enough time to get all their work done, and increasingly, leaders are looking to agents as a way to fill that gap. Also, take a sample of any study on agent adoption in the enterprise, and you'll see a pace that mirrors or even exceeds the pace of general AI adoption.

Starting point is 00:17:40 If Norgas Bank shows what some of the near-term opportunities are, it's also very early, and we've barely scratched the surface on some of the challenges that we will come to face. First of all, most of these results were still largely in the co-pilot paradigm, where the agentic era brings entirely new types of challenges as human employees figure out how to interact with digital employees. An entirely new set of skills and a new way of thinking is going to be required to make the most of that era. And right now, I think upskilling programs are still largely stuck in the assistant paradigm, not focused on the agent paradigm. Cap Gemini recently asked executives what they thought the most important hard and soft skills were to harness the potential

Starting point is 00:18:15 of agents. And on the hard skills side, it was data management, programming and software development, troubleshooting and debugging. On the soft skills side, it was decision making, collaboration, and logical reasoning. What's clear from all the studies, that the more companies invest, the better their people do, and the more gains from AI they get. Going back to this BCG slide, they looked at the difference in companies who were just simply rolling out AI versus actively working to redesign workflows, and found just massive increases in the amount of time their employees saved, their employee's ability to shift to strategic tasks, and the percentage of their employees who believed that AI enabled better decisions.

Starting point is 00:18:52 I'm not sure that we have enough information to really call something the Norges Bank Playbook, But I think that if you're going to take away two big lessons, it's one, make AI usage mandatory, and two, support the hell out of people when you do so. I will obviously continue to watch these experiments and share with you guys here as people share their results. For now, that is going to do it for today's AI Daily Brief. Until next time, peace.

The AI Daily Brief: Artificial Intelligence News and Analysis - How One Company Saved 213,000 Hours with AI

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.