In short
Podcast Notes: Leveraging AI - Episode 184
Episode Overview
- Title: AI “Frontier Companies” Replacing Traditional Companies
- Date: Week Ending April 26, 2025
- Host: Isar Meitis
- Focus: The transformative impact of AI on business structures, workforce changes, educational requirements, and operational risks associated with AI adoption.
---
Key Topics Discussed
- AI Adoption Trends
- Microsoft’s Work Trend Index for 2025:
- Highlights a shift from grassroots AI adoption to top-down strategies.
- Introduction of "Frontier Firms" that rethink organizational structures using AI.
- Statistics:
- 82% of leaders see 2025 as pivotal for AI strategy.
- 81% expect extensive integration of AI agents in their strategies within 12-18 months.
- Only 24% of leaders report organization-wide AI deployment, up from 12%.
- The Hybrid Workforce
- Definition: A blend of human and AI agents focusing on specific projects rather than traditional roles.
- Concept of "Work Chart": New organizational structure focused on agile and project-oriented teams.
- Productivity Challenges:
- 53% leaders demand increased productivity.
- 80% of the workforce feels time-constrained.
- Implications for Education and Skills
- AI literacy is becoming more critical than formal degrees as a career differentiator.
- Over half of Gen Z job seekers feel their degrees are devalued due to AI.
- Companies are increasingly prioritizing skills over educational qualifications.
- Risks of AI Deployment
- Operational Risks:
- Potential for unsustainable business models if AI is not integrated thoughtfully.
- Issues such as "hallucinating" chatbots could damage brand trust.
- Transparency in AI usage is critical to avoid backlash from users and stakeholders.
- Future Workforce Dynamics
- Predictions about job displacement versus new roles in managing AI resources.
- Discussion about potential future roles, such as a Chief Resource Officer, to balance human and AI capabilities.
- AI in Action
- Examples of AI's current applications:
- Customer service, marketing, and product development.
- AI companies exploring outsourcing to AI-assisted contractors.
---
Key Takeaways
- Immediate Action Required: Businesses must act now to integrate AI or risk being outpaced by competitors.
- Rethinking Work: Companies should rethink traditional work structures and embrace agile and project-based team formations.
- Focus on Training: Organizations need to prioritize AI literacy and hands-on training to prepare their workforce for the changes ahead.
- Understanding Risks: Leaders must be aware of the risks associated with AI, including operational risks and the need for transparency.
---
Conclusion The episode emphasizes the critical moment businesses are facing regarding AI integration. Companies need to adapt quickly to leverage AI's potential effectively, balancing human and digital workforce dynamics while managing associated risks. The conversation also highlights the necessity of AI education and the evolving definition of job qualifications in this new landscape.
---
Additional Resources
- Listener Survey: [Fill out the listener survey](https://services.multiplai.ai/lai-survey)
- AI Business Transformation Course: [Learn more about the course](http://multiplai.ai/ai-course/)
- Connect with Isar Meitis: [LinkedIn Profile](https://www.linkedin.com/in/isarmeitis/)
- YouTube Full Episodes: [YouTube Channel](https://www.youtube.com/@Multiplai_AI/)
---
Feel free to utilize this structured summary to better understand the key insights from this episode, as well as to share with colleagues or team members interested in the implications of AI in the business landscape.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Transcript
Automatic transcript. May contain errors.0:00Hello, and welcome to a weekend news episode of the Leveraging AI podcast, the podcast that shares practical ethical ways to improve efficiency, grow your business, and advance your career. This is Isar Maitis, your host, and we have a really fascinating news episode for you today. We are going to start with Microsoft's updated Work Trend Index annual report, and they call the report The Year the Frontier Firm is Born. We're going to explain more in a minute, and we're going to dive a lot more into the impact of AI on the global workforce and job seekers and education. So there's a lot even on just this initial topic that we're going to unpack.
0:37And then we're going to also talk about the acceleration of AI models and then lots of rapid fire items, including some very interesting new updates from OpenAI, as well as new releases from Adobe and a lot of other good stuff. So let's get started.
0:56So Microsoft just released their Work Trend Index Annual Report for 2025. the 2024 report was released just less than a year ago on May of 2024. If you remember, the key findings back then were that AI is getting more and more grassroots adoption, and many employees are bringing their own AI to work. They coined that phrase, and that enterprises were just in the beginning of AI adoption from a company-wide perspective. Well, this report is exactly the other way around, showing a very strong top-down investment in many organizations and a complete change and transformation of how organizations view AI and its implementation, both now and in the future.
1:34This is a huge survey. They surveyed 31 ,000 workers linked in labor market trends and their own internal Microsoft 365 productivity signals, basically them monitoring how people use Microsoft 365. And the interviews were done across industries and levels. So from AI native startups to academy to economists, scientists, thought leaders, et cetera. So it's a huge survey that really does a very thorough job of trying to understand where AI is in 2025 and how it's impacting the global workforce. So they defined a new type of company and they call it the Frontier Firm, which is a company that adopts AI, not just in replacing people or specific processes, but actually rethinking how the company is actually structured.
2:20When their main concept is that instead of building around specific current structure of companies, building agile teams that are target-oriented and project-oriented and goal-oriented, that is using on-demand intelligence to power a hybrid type of teams of humans and agents that are going to complete specific tasks and specific projects versus relying on marketing, accounting, finance, sales, and so on as the structure for companies today. And the report is claiming that in the next two to five years, most companies will go through some level of transition in that direction. Now, I mentioned that from 2024 and now obviously in the beginning of 2025, that 2025 is going to be the big year of AI in business, and it's going to go beyond the buzz in specific individuals to company-wide implementation and agent building.
3:10And this is exactly what this report shows. 82 % of leaders see this year as pivotal for rethinking strategy and operations, and 81 % of leaders expect agents to be moderately to extensively integrated into their strategy within the next 12 to 18 months. So not necessarily implementation yet, but definitely very strong signals of implementing AI agents across multiple aspects of the business. Now, again, another big shift from last year, 24 % of leaders report organization-wide AI deployment, while only 12 % remain in only pilot mode and not actually deploying any AI for actual tasks. Now, maybe the biggest claim of this report is that it argues that intelligence, though previously was limited and expensive asset, is becoming a durable good, abundant, affordable, and available on demand.
3:59And the idea is that agents can provide digital labor that can scale capacity basically without limit. And the interesting thing is in addition to growth, a lot of leaders and a lot of people in the survey basically saying that will help close the capacity gap. So there's growing business demand and not enough human capacity and energy in humans to comply with that demand, and AI agents will allow to close that gap. Two interesting parameters that relate to that is that 53 % of leaders say productivity must increase, but 80 % of global workforce feels they lack enough time or energy. Basically, most people in the survey are saying they don't have enough hours in the day to complete the tasks that they need to complete.
4:40In the Microsoft 365 telemetry, basically data from people using Microsoft 365, they found some profound information about the way we currently work. They found, as an example, that on average, every employee that uses Microsoft 365 is interrupted every two minutes in their nine to five job. That's 275 interruptions per day. Now, that may mean many different things for many different reasons, but this is actual data from actual usage of actual apps basically tells us that we became really, really bad in actually starting and completing tasks. And the interesting thing is that AI agents and AI tools may be the thing that can help us close that gap, meaning if we don't have to do the actual work.
5:22And as the report says, if we can separate knowledge workers from knowledge work, that interruption stops being an interruption. It's us managing agents doing the actual work. So instead of humans performing the actual tasks, agents will perform the tasks and humans will be involved in orchestrating, managing, and leading strategic initiatives and the overall operation versus doing the actual end tasks. And here's an interesting quote that relates to that. To keep up, companies must do more than add AI to existing workflows. They must rethink the very nature of knowledge work. This starts with separation of knowledge workers from knowledge work.
5:59So they are claiming that the traditional org chart will change to a hybrid model with a new concept that they called work chart, basically hybrid teams of human and agents that will be assembled for specific goals, for specific projects, completing specific aspects of the needs of the organization and can be disassembled and reassembled as needed for specific aspects of the company. This obviously requires a complete rethinking of how companies are structured. Now, an important fact that they share that I actually find not reasonable based on all my communications with multiple CEOs across multiple things that I do.
6:36But this report is claiming that 46 % of leaders say their companies are currently using agents to fully automate workflows and processes. I find that not reasonable, but I think there's a lot of confusion around the concept of what agents are. And some people may call a custom GPT an agent, or maybe even just using Gemini or Copilot. Some people would consider agentic work, which is not really, but that's still a very high number if 46 % of leaders saying that their companies are regularly using even AI just to complete regular tasks, that's higher than what I'm seeing, but that's definitely showing you that this wave is not slowing down, it's just accelerating.
7:13The main areas in which companies are saying that AI is being used is not surprising. Customer service, marketing, and product development are the top three, but that's going to grow to more or less every aspect of the business in the next 12 to 18 months. Now, an interesting new part of this paradigm shift is the new ratio between humans and agents for different roles, functions, and projects. So basically when companies will need to think these teams, each task, each goal, each company, each department will need to think about the ratio required for that specific aspect between humans and agents.
7:45And what's the right ratio is going to be a critical aspect because it will define what resources are required, which most likely will lead to new roles in the company that will manage these resources. So they're assuming that there's going to be this new kind of role that's going to be a chief resource officer. So think something like a cross between IT and HR that will be able to identify how many resources a company has across human and digital employees and build the right teams with the right ratios for the right goals. In general, this report envisions many new roles around AI capacity, AI infrastructure, AI deployment, AI optimization, and so on, which may or may not offset the amount of people that may lose their jobs due to AI doing a lot of the day to day tasks.
8:31But in general, a very positive report showing that many leaders do not see this as replacing people, but actually growing capacity and capabilities and allowing most employees to become different kinds of managers, many of them managing agents instead of doing the actual tasks themselves and using their own brain capacity for more strategic work. Now, when I read this, On one hand, this is very positive, maybe one of the most positive, real serious reports I've read recently. But my question is, how many people we actually need to do strategic work in a company? Do you actually need so many employees like you have right now just to do strategic work?
9:10I think the answer is no. So the question is what happens with these people. And I don't think still anybody has an answer for that. I think in the short term, that's what we're going to see. We're going to see companies who are early adopters, who are these frontier companies moving faster into that direction and really gaining significant market share. But over time, there is limited growth because there's limited demand. And because there's limited demand, then that means that you won't be able to scale indefinitely. And then what do you do then with the extra capacity that you can continue to generate at a faster and faster speed?
9:41We're going to talk about the faster speed in a minute. So one of two things will happen. Either a lot of people are going to be unemployed, or we're going to see this huge wave of entrepreneurship because smaller teams with single individuals or a few individuals will be able to do a lot more and create stuff because they will be able to manage a much larger workforce, most of it digital. That still doesn't answer my previous question of there's finite demand, meaning even if it is entrepreneurship, even if it is smaller companies, even if a lot of people create a lot of stuff, there's still a limit to what the demand is.
10:13There's going to be new types of demands, As I mentioned before, many of them we can't even imagine. The one thing that I'm certain of is that the global workforce, as we know it today, is going to cease to exist in the next five years or so, maybe even faster because of this trend. So what is this report suggesting are the next steps? The first thing is hire your first digital employees, or in other words, define roles for automation and AI in your organization and figure out exactly how to deploy them in an effective way, which people needs to be involved. what is the performance criteria, what are the KPIs, what is the ratio, et cetera, et cetera.
10:49How do you basically deploy the first AI capabilities in your organization? The second is to really identify that ratio between humans and automation for multiple aspects and processes in your organization. And the third one is how do you scale that? How do you move beyond just the pilots and adoption level at the team or individual and deliver a company-wide restructuring and rethinking of the entire organization. So a quick recap of all of that, what is the most important stuff here is AI literacy, knowledge, skills, and hands-on experience, right? To do all these things, the leadership team and every leader in the organization, and eventually every person in the organization needs to know how AI works, how to use it, how to apply it, what to be afraid of, what to be excited about, and so on and so forth.
11:34And many companies are struggling with how to get started. And the way to get started is with company training, actually getting your people to get the basic knowledge of how all these things work, how it can impact their day-to-day job. So then you can go into the step of rethinking how you're doing it with AI. This is exactly what we have been doing at Multiply, my company, for the past two years or more. We've been working with organization-wide training for multiple organizations from 20 people all the way to 30 ,000 people and providing specific training to specific teams, including a focus on leadership teams in order to enable them to have these conversations and plan and strategize the next variation of a company.
12:17In addition, we've been running the AI Business Transformation course, and this is for people who do not feel that their organization is providing them this kind of training. And the next course, by the way, starts on May 12th. There's a link in the show notes. I'm not going to dive into that, but it's a course that we've been successfully delivering for over two years and hundreds and probably thousands of business leaders and business people has been through it and completely transformed their careers and their businesses based on the knowledge they acquire during that course. If there's one thing that this report tells you is that you have to learn how to use AI, otherwise you and or your business will stay behind.
12:49Now, from that to a article that talks about how CEOs embrace AI and slash jobs. So this article shows examples from multiple different companies. As an example, 80 % of PayPal's customer service tickets are now handled by AI chatbots. 80%. That means they need significantly less customer service people. Other companies like United Wholesale Mortgage and Shopify, as we mentioned last week, are already prioritizing digital workers over human hires. If you remember in an interview that I shared with you last week, Shopify CEO, Tobias Latke, basically told their teams that nobody gets to hire anybody unless they can prove that AI cannot do that job that they want to do.
13:30Ernest & Young automates HR and finance functions with chatbots, and some companies are even exploring outsourcing some of the work to AI-assisted contractors. So basically, AI agents for hire that will be managed by third parties helping you doing work in your business in order to reduce the amount of employees that they need to have in the company. So again, that goes back to prove as a leader, you have to figure this out. And as an employee, you must have AI skills or you will be one of the first people on the chopping board. Now, I know if you're in a leadership position, you might be thinking, I don't have time for this right now.
14:03There's other bigger things to take care of before I take care of this. Well, you got to remember, your competition either will or already doing this, meaning their cost structure that is probably right now similar to yours is going to dramatically change, meaning they'll be able to provide the same products or the same services for significantly less money while having higher margins, which means you will not be able to compete with them. And the flip side is also true. Obviously, if you move first, you'll be able to have a first mover advantage and be able to be more competitive and grow your market share.
14:33And to again, show you how across the board this is, Fiverr's CEO, so those of you who don't know Fiverr, it's one of the most successful companies in the world today to allow individuals with skills to offer their services to anybody on the planet, similar to Upwork. And their CEO, Micha Kaufman, said, it does not make sense to hire more people before we learn how to do more with what we have. And that is with relation to doing it with AI. Now to show you this is just the beginning, a new startup called Mechanize is aiming to replace all human work. Obviously starting with white collar jobs, they announced that on X on April 17th.
15:09Their mission is to provide data evaluations and digital environments to enable AI agent bots to replace workers in any job. That's their mission statement. So their goal, as they're claiming, and I'm quoting, is to completely automate labor that could generate vast abundance, much higher standards of living, and new goods and services we can't even imagine today. This sounds awesome, right? It's fantastic. We're going to get everything that we need, and it's all going to be abundant, and we can get everything that we need, and everything is going to be awesome. But my problem with this mindset is, let's assume for a second it is correct, and we can generate everything we want at almost no cost.
15:47We still need the consumption power to be able to consume this. And if people are not going to have jobs, they're not going to have money. If they're not going to have money, they won't be able to purchase any of these goods and services that are going to be fantastic and really, really cheap. And nobody knows exactly how that equation is going to work. I think the concept of UBI, universal basic income, is not deeply thought into. I don't see how that can even work. I don't think we have any infrastructure to provide that. I don't think that universal basic income for a person living in Beverly Hills is the same as somebody living in a low income area.
16:18And they will both need to sustain their life one way or another. There's so many questions that nobody knows how to answer about UBI that I don't think that's the right solution. And I also don't think we'll have any kind of setup of UBI in time to meet the speed in which this revolution is moving. Now, how does that impact the workforce? As an example, in a new research, 49 % of Gen Z job seekers say that AI devalued their college degrees. So what they're claiming is that the rapid AI adoption in automating multiple entry-level tasks, he's basically eliminating the tasks that they were supposed to do as their first job out of college.
16:55That leads to one-third of millennials regretting getting a formal degree. Lindsay Fagan, Indeed's Senior Talent Strategy Advisor, said, for any organization to succeed with AI, every single employee needs to have a basic understanding of AI and how their company uses it. So she's basically saying that understanding how to use AI in your aspect of the business is as, and maybe more important than having a degree in that topic. As a CEO of a company myself, actually two companies, because I also run a software company, I highly agree. Because if you go back to the report from Microsoft, the ability to do stuff and the access to knowledge is going to be democratized.
17:34Everybody will have access to that, whether you have a degree and don't have a degree. And your ability to use and operate in an AI-centric environment is going to be the biggest differentiator of human capabilities in the future. And to be fair, in the present as well, going back to you need AI training for yourself and for your company right now. And to push more on the point of where the workforce is going from requirements, the share of job posting requiring bachelor's degree or higher dropped from 20.4 % to 17.8 % over the last five years based on Indeed opening opportunities for 64 % of US adults that do not have a bachelor's degree.
18:14Now, while that's the statement from Indeed, I actually don't see that true. I just think the requirements are going to be different. So instead of needing a degree, you will need AI capabilities and deep knowledge in a specific topic because that combination is going to create the magic, right? If you deeply understand how a specific thing works, a process, a system, a product, a demand, something, and you know how to apply AI to that area of knowledge, this is going to be the most required combination when hiring new people. Corey Stahl, who is an economist at Indeed Hiring Lab, said, while educational requirements are unlikely to vanish from the job posting, growing support of skill-first hiring approaches is a clear sign for workers to invest in skills now.
18:59And he's specifically talking about AI skills and industry-specific skills. And if you want to take that to the next level, Anthropic is warning that AI virtual employees can become a common thing on companies' networks by 2026. And what they mean by virtual employees, they mean far beyond using ChatGPT or Gemini or Copilot and even beyond agents. They're talking actual quote-unquote employees that have memories and roles and login credentials and can actually participate in multiple aspects of how organization operates. A digital employee. So not task-specific like most agents are, but a much broader, with much more autonomy capability that will potentially handle much more complex roles and tasks than we can see or even imagine right now.
19:49And one of the reasons Anthropic is raising this concern is the fact that this will open the network to a whole different level of risks. And now, in the specific quote from Jason Clinton, Anthropic's chief information security officer, he said, in that world, there are so many problems that we haven't solved yet from a security perspective that we need to solve. And they're talking about next year of having non-humans be part of companies' networks across more or less everything that we're doing. And that's obviously just going to accelerate over time. So a quick summary of this entire first segment, AI education and practical training hold the key for future careers as well as the future of companies.
20:27Without having that knowledge, you cannot start having the more complex discussions in your organization, the strategic discussions on how your organization needs to transform in order to be ready for this new world. And that is true for any white collar type of job. And in any industry, there's obviously ones that are going to move faster. There's going to be ones that are going to move slower. But the trend is very, very clear. This training cannot be theoretical. It needs to be skill based and addressing actual real world use cases. Again, that's what I've been doing with companies in their facilities or online for the past two years.
21:02And I can see dramatic changes and shifts in companies' ability to A, be more efficient, but B, think differently on how their company should operate in the future. And whether you do this with me or with somebody else, you as an individual or you as a business leader have to do that process. You need to research and find what would be best for your company, what would be best for your career, and go and do that. And the sooner, the better, because this thing is coming, it's coming fast, and it's going to have a profound implication on everything we know on how work is done and how companies operate.
21:33If you want to explore what we offer, there's a link in the show notes. You can just click on that, go and figure that out, or even set up a meeting with me so we can explore this together. Now, let's talk about the negative side of really fast deployment of these AI capabilities in your company. So Cursor, the most successful AI development platform that's been growing like crazy in the past year and a half, just had their support chatbot, Sam, falsely make up a policy of the company, causing a huge outrage in a lot of developers who were threatening to cancel their Cursor subscription. So the whole thing started with just a simple bug that was causing developers to be logged out of their accounts when they were trying to switch devices.
22:14When people went to the chatbot to ask what's happening, the chatbot claimed that the company's policy is that you have the license only for a single device. And if you want to use multiple devices, you need multiple licenses. That created a huge outrage across social media and Reddit. And very, very quickly, Cursor's co-founder and CEO, Michael Truel, basically came out on Reddit and said, we have no such policy. This is an exact quote, basically saying this is a AI chatbot hallucination that gave that wrong policy. If you remember a few months ago, there was a similar incident with an AI chatbot from Air Canada inventing a refund policy that was incorrect.
22:51People bought the tickets, they had to cancel, and they couldn't get their money back because Air Canada said they don't have such a policy. And yet the court made them pay that back because the fact that the chatbot said it still makes them liable because it was said on behalf of Air Canada. So these tools are not perfect yet, to be fair. They're far from perfect on many different aspects, but they're definitely good enough for many other aspects. To be even more fair, many human customer service people make mistakes as well. So looking for the one time that AI chatbot makes a mistake in saying that this is the most critical thing, and that's why you shouldn't apply AI for that task, is not necessarily the right approach.
23:25You need to compare it to the current level of operation and to the current level of human operators. But that being said, it is important to know that these tools hallucinate, that these tools may provide different questions to the same answer as several times. And when you're deploying these kind of tools, you need to think very carefully how to A, A, reduce that number, and B, how to make sure that it's mitigated to the right places and doesn't create a catastrophic impact on your business. Continuing on the same line of what might happen if you deploy AI without sharing it with the universe, the State Bar of California admitted that its contractor, ACS Ventures, used AI to craft 23 multiple choice questions for the February 2025 bar exam.
24:05Now, by itself, I don't see that as a problem. The problem was that they were not disclosing it and were not being transparent about it. And now there's a whole issue around it. That's not the only issue. They also copied questions from other exams. They did a lot of other bad stuff when creating this exam, but using AI definitely was rolled into that as one of the things they should have shared. And now the California Supreme Court has ordered the bar to explain why and how they're using AI to ensure question reliability. So in general, the point that I'm trying to make here is you need to be very transparent with your employees, with your customers, with your ecosystem, suppliers, et cetera, that you are using AI, how you're using it, what you're using it for.
24:42So people are aware of that. And then I think there's going to be significantly less backlash versus people finding after the fact, that's what you were doing, especially in cases that things go wrong. And the last point in this deep dive, in today's deep dive that has to do with how this will impact us in the next few years before we switch into the rapid fire is that the speed in which AI and agents are performing tasks is accelerating faster than we thought. So if you remember a couple of weeks ago, I shared with you a report that is showing that AI agents are doubling their capability every seven months.
Read the full transcript
25:14If you remember in the same report, I told you that recently it became three to four months. Now in a new report done by the AI Digest, they are confirming that recent models are indeed doubling the capability of doing tasks every three to four months and not seven. If you remember the process of doing this survey was looking at what types of tasks and how long they take humans to do that AI agents can replace at a 50 % success rate. Again, this is not a perfect measurement, but it's a measurement to try to figure out are AI agents making progress or not. And they found that, again, it went from seconds to minutes to now days, and that pace is doubling every three to four months.
25:59What this report is basically saying is that if this continues, we are very quickly going to get into a flywheel acceleration effect where every generation of agents will be able to faster develop the next generation of agents that will be able to then faster develop the next generation of agents going into an explosion of capability of AI and agents in general. Now, nobody knows if this is going to hold. They're themselves claiming that they're only looking at one year of data, which may not be enough. but using the same argument, we also don't know if that's going to accelerate even faster moving forward.
26:30So whether this space is going to continue, whether it's going to slow down or whether it's going to accelerate is unclear. What is clear is that the acceleration is there. And even if it slows down by an order of magnitude, it's going to change everything we know very, very quickly. And now two rapid fire items. The first one has to do with investments in new data centers. We shared with you several times that Microsoft have been slowing down their investments in data centers, well, now AWS is doing the same. So two of the world's three largest hyperscalers of data centers, data, and compute are slowing down their future data center leasing and development.
27:06Does that mean that there's already less demand? I don't think so. I think that means they're expecting the demand to slow down in the future and they're afraid of overcapacity. It's always better to operate in a scenario where there's more demand than supply. And I think that's what they're trying to aim for. I don't think that has any implications on the immediate future. If you look at everybody in the AI race are looking for more and more capacity at this point, I think this looks more on whether that capacity is going to slow down as things starts to level off and more and more companies already have this in place and models become more efficient and they are.
27:41And so I think this is the combination that's driving this slowdown. And now to a bunch of news from OpenAI. A very interesting piece of news from OpenAI comes from Nick Turley, who is OpenAI's head of product, and he testified on April 22nd that OpenAI would be interested in buying Google's Chrome browser if the court forces Google to sell it as a remedy for its illegal search monopoly. So if you remember, the Department of Justice has ruled in 2024 that Google's search dominance is illegal, and now they're looking for different ways to stop that. One of them might be forcing Google to sell Chrome.
28:14Chrome currently holds 64 % share in the global browser market and a much bigger share in the Western hemisphere, which is obviously very, very significant. And OpenAI would definitely be one of the companies that would love that asset under its belt. And for OpenAI specifically, it makes sense from multiple different reasons. OpenAI tried to partner with Google previously and get access to their search capabilities in order to integrate it of how ChatGPT works, but Google did not want to partner with them, claiming that they're competing with them, which they are. And that's why they went with Microsoft using Bing, but they're claiming that Bing was never good enough.
28:53And they're also claiming related to Chrome, that this will obviously provide them huge distribution and their planning, if this ever happens, to show the world how an AI first browser should actually work. I must admit personally that taking Chrome from one company that controls everything that we know and do to a different company that is the future potentially of everything that we know and do is not necessarily a good solution, but I'm not the one that's going to make that decision. Staying on OpenAI, we talked many times before on the lawsuit from Sam Altman trying to stop OpenAI from transitioning from a nonprofit to a for-profit company.
29:28I also shared with you that multiple past employees of OpenAI have joined that lawsuit. Now there's a different phase in that where a group of ex-OpenAI employees, as well as Nobel Prize winner Jeffrey Hinton and law professors and a lot of other people has sent a letter to the California and Delaware attorney generals urging them to stop OpenAI's restructuring, citing safety risks and betrayal of its nonprofit mission. The letter that was also delivered to OpenAI's board on April 22nd argues that becoming a for-profit entity would, and I'm quoting, subvert OpenAI's charitable purpose and remove nonprofit control and eliminate critical governance safeguards, basically saying that they're going to prefer profits over safety and benefits to humanity, which was their actual original goal.
30:18How this will evolve, I don't know, but there's a lot of money involved in that process as a part of their recent raise of$40 billion, mostly from SoftBank. $10 billion depends on that transition. That's a lot of money. There's a lot of power and a lot of money and a lot of interests across every aspect of this transition. It will be very interesting to see whether that happens. Again, the deadline for this$10 billion is the end of this year, December 31st, 2025. Will SoftBank and the other investors actually pull that money out if it's not successful? I don't know, but this is what it feels like right now.
30:51Speaking of a lot of money and chat GPT, in an interesting exchange on X this week, Sam Altman shared how much politeness costs OpenAI. So he said that tens of millions of dollars on electricity is being invested in people saying please and thank you to Chachupiti, but he also said that is tens of millions of dollars well spent. I've shared with you multiple times in the past that there is ample research that is showing that being polite to these models actually give you better results. I know some people saying it's a machine, why the hell does it care you being polite? Well, that machine was trained on millions of human interactions, and in most cases, when one person is polite, they're getting better answers compared to when they're being rude, and hence the models follow the same kind of behavior.
31:32I shared with you in the past two weeks that OpenAI are planning to release an open source model for the first time since GPT-2, making them, for the first time again, OpenAI. Well, this is obviously done to a lot of competition, but there's new information about this new model. They're claiming it's going to be close to a frontier model, meaning they're planning to release something that is going to be highly capable, and they're planning to release it with significantly less commercial restrictions compared to models like Llama from Meta and Gemma from Google. So that I believe is good news to the open source universe.
32:02We're going to get another very capable model that is going to be open source. What does that mean to the way OpenAI is going to operate in the future? How it's going to compete with its existing models? And a lot of other questions stays unanswered, but it's very obvious that the competition from Llama, from Gemma, from DeepSeek and others is definitely affecting and impacting how OpenAI are going to operate in the future. Good piece of news from OpenAI. OpenAI just introduced a cost-effective version of its highly useful deep research tool. The way it works is instead of being powered by a high-capacity, high-cost model, it's actually running all for mini in the background.
32:39And what's going to happen is once you run out of your capacity with the full model, it is going to switch you behind the scenes to this lightweight version of deep research. OpenAI is claiming that it's very close in its ability to deliver high quality research results. And as I mentioned, the benefit is that you're going to get a lot more queries than you could before. So the plus and team and education members are going to receive 25 of these in addition to the 10 in the full capability. And the pro users are going to get 250 and three users are going to get five additional searches that they didn't get any before.
33:14So these are again, good news. How close Is it to the actual full model? I don't know, but I'm sure we will learn in the next few weeks as people start using it and we'll be able to compare the two. Another new small change from OpenAI that could be profound in its implications is that it's now using the memory of everything that you did with ChatGPT when it's providing you search results. So when you are searching the web through ChatGPT for different answers, it will look at what you've shared with it in the past and will provide you personalized answers based on that. So on one hand, this is very helpful.
33:44If you're a vegan and you're just asking for nearby restaurants, you will find vegan restaurants for you without you even saying that because you will make that assumption. And this is obviously a cool differentiator for Chachupiti as far as providing this kind of personalized data. The problem that I see with that is it's just another way to allow each and every one of us to live in its own echo chamber, just like social media. So if you have a specific political opinion, if you have a specific standing on a specific thing, these are the results that you're going to see compared with Google when you search and you get random results that may or may not align with what you want, but you at least see that there are other options.
34:19And if all we're going to see in the future is stuff that is aligned with what we've seen before, I don't think that's a good thing for humanity and society. Another piece of good news from OpenAI is that OpenAI 03, their new model, their new full model, so not 03 mini, just doubled its quota to people using it to 50. And it also introduced the task scheduling capability with 03 and 04 mini. Previously, it was only done by 04. So new capabilities that is available to everyone right now, allowing us to do more with these new models. Another profound piece of news from ChatGPT from its implications on the future is a code string in the back end of ChatGPT was located this week showing BuyNow and Shopify checkout capabilities in the back end of ChatGPT, signaling that there might be a real true partnership brewing between Shopify and ChatGPT, allowing ChatGPT users to search, find, and purchase straight from the ChatGPT interface without having to go to ChatGPT.
35:18Shopify's platform. In the long run, I'm sure they will integrate other suppliers as well. Now for Shopify, that obviously makes a lot of sense. We know that ChatGPT is growing like crazy. The user base is anything between 500 million and 800 million weekly users right now. And if they can now search and buy things straight from Shopify, that's obviously beneficial to Shopify. In the long run, it's very clear that this approach of using agents or AI tools in order to search and find what we want to buy is going to be the norm. And that means that anybody who's selling anything online has to change the way they work.
35:51Think about all the millions of people who run Shopify stores right now, and their stores are going to get significantly less traffic regardless of everything they've done so far for SEO and so on. And they will have to optimize for something else. And what that something else is, nobody exactly knows. So take that to the bigger scale of companies like Amazon and Target, and anybody who's selling stuff online will have to make significant adjustments in order to make sure that traffic is coming to them. Take that even to other aspects like buying travel, and you understand that the online buying world as we know it is going to change dramatically, but nobody knows exactly how.
36:27And staying on OpenAI, but now on the lawsuit front, another company in this particular case, Ziff Davis, the owner of IGN and Synet and PC Magazine and 45 other media outlets has filed a lawsuit in Delaware Federal a court against OpenAI for illegally, and now I'm quoting, intentionally and relentlessly copying their work, ignoring robot TXT files instructions, and stripping copyright information. This is obviously not the first lawsuit, but they are seeking hundreds of millions of dollars in damages. How will they prove these damages? That's a whole different question, but that's the lawsuit right now that joins other similar outlets such as the New York Times and the Intercept and others that are suing OpenAI.
37:09Other companies, such as the Washington Post, has chose a different avenue. So on April 22nd, the Washington Post signed a licensing deal with OpenAI, integrating its content into ChatGPT to provide answers and links from ChatGPT to the Washington Post content. And they're joining 20 other news outlets and publishers, including News Corp and The Atlantic and The Financial Times that already have these kinds of licensing deals with OpenAI. We mentioned that time and time again before. this is going to end up at the Supreme Court and they will decide whatever they're going to decide, most likely making OpenAI and their investors pay and compensate for damages.
37:43And that's probably going to be the end of it. I'm not going to want deciding how that's going to turn out. But from my perspective, the two most exciting piece of news from OpenAI this week, one is that ChatGPT image generator, the amazing new tool that they just released, is now available through the API, meaning you can now build or integrate it into other applications. Now, this new tool, if you haven't played with it yet, go and check our previous episode that was just released on Tuesday of this week. It's called 15 Mind-Blowing Use Cases for the New ChatGPT Image Generation. If you haven't tested it or used it, or even tried to use it for business use cases, go check out that episode, especially on our YouTube channel, where you can actually see all the images and see what we're doing.
38:26From a cost perspective, they're defining exactly what's going to be the cost per tokens and text and so on, but it translates into two to 19 cents per image, depending on the quality of the image that you need. You can decide for yourself if that's a lot or little, it just depends on how many images you're going to generate. But if you're integrating into a product that you're going to charge for, then it doesn't really matter. And you now have access to this amazing capability via API. The other piece of news, which is not available to me yet, but I've seen several people share, is that this new image generation capability is going to be embedded and integrated into custom GPTs.
39:00So those of you who don't know what custom GPTs are, is these automations that you can build in ChatGPT and can use to do multiple tasks that are tedious tasks that are happening in your company or in your personal life right now. It's a magical tool. And I'm now more or less addicted to both capabilities, to the image generation and to custom GPTs. And being able to use the new image generation capabilities in the custom GPTs will be incredible. Right now, for me, it's still using DALI 3, which is not even close. Now, if you want to see why I'm so excited about this whole image generation capability and you haven't been following what's happening, I released a very fun post on Friday, April 26 on LinkedIn, doing a quiz game that you can try to guess different images and what they are.
39:39Just go and check my profile, Isar Matis on LinkedIn. Go and find my post from yesterday, Friday 26, and go and try to play the game. But it will show you some of the capabilities of this new model and how amazing it can be both for business and for leisure. Now, staying on image generation, Adobe held its Max event in London this past week, and they shared a lot of interesting news on what they are doing. The first interesting thing is they have released what they called created without generative AI tag on their Adobe Stock Photo website, basically allowing users to identify images, videos, and audio that were created by humans without any AI assistance.
40:17They're doing that because they're claiming 70%, 7-0, of the recent submissions to Adobe stock have been AI generated. So they want to be as transparent as possible and allow users and end users to know what was created with AI versus what was not created with AI. And obviously, the naked eye cannot tell anymore. And so they're adding this tag that will tell people how is that done. Now, my question is, does it matter? And when does it matter? And I think it will matter in very specific use cases. If you're buying art, then I think it will matter. Even then, I assume some people won't care because they would just like the art.
40:50But I don't think that stock photos for websites, blogs, ads, or other digital media matters. I think what does matter is that photo going to achieve the goal that I needed to achieve in that particular use case because it's a business use case. And whether it was done by humans or not makes absolutely no difference to me. And I think it won't make any difference to anybody else. I also think that the concept of stock photos will disappear because it makes absolutely no sense. I stopped using stock photos about two years ago when these tools were not even close to what they are right now. And I can generate the exact image that I want way faster than I can find an image that will most likely be a compromise of what I need.
41:31And so I don't understand the concept of stock photos websites anymore. I think videos are going to be next and music is already there. So I think this whole concept of tagging stuff in a stock photo website as human generated is interesting as a statement, but I think it's completely irrelevant to where the world is going. Adobe also released their Firefly Model 4 and 4 Ultra as part of their announcement in the conference, and it can generate lifelike images of up to 2K resolution with enhanced control over style and camera and zooms and so on. Just another very powerful tool to generate images like many others.
42:05The interesting part of this is that they're going to release a Firefly-specific mobile app for iOS and Android that will allow people to use that capability to generate images on the fly that is already available on several different apps, including the ChatGPT app with its amazing capability. I will say something about the difference between this tool and the ChatGPT tool. While I don't have a clear comparison which one generates better images, if there is such a thing, I can tell you that because ChatGPT understand the context of what you're trying to do, it is a lot easier to generate exactly the images that you want in ChatGPT than in any of the other tools.
42:41Now, previously, there was a compromise between the image generation capabilities in DALI 3 versus your ability to generate the images you needed because it was easier to generate. now the new image generation tool in Chatubit is incredible. And so you don't compromise anything. And yet it's a lot easier to generate exactly the images that you want because it understands what you are trying to do. Now, Adobe is claiming that Firefly users has generated 22 billion assets already with its different tools. First of all, that's an incredible number. But I think the vast majority of it was probably generated within the Adobe suite.
43:12So people are already Adobe users, Adobe users who are generating AI assets for what they're creating versus users like me who do not have access to Adobe Suite and is not using it and probably won't ever use it unless it somehow becomes better than my daily tools. From image generation to hardware on April 23rd, Meta rolled out a live translation feature to all of its Ray-Ban Meta smart glasses users, allowing you to do live translation between English, French, Italian, and Spanish. I think this is an awesome capability that will connect people around the world. And if you're going on a trip in companies who speak any of these languages and you speak one of these languages, you will be able to communicate back and forth with other people without being able to speak the same native tongue.
43:53There are also recently added capabilities to send and receive Instagram DMs, photos, and phone calls via the glasses, and WhatsApp Messenger is coming next. So it's becoming a hub for communication in general. There is very small doubt in my mind that this is the form factor of future devices, especially once they learn how to project on them in high resolution. It is very obvious why Meta is pushing this as aggressively as they are. And it's very clear why OpenAI is trying to partner with Johnny Ive and Apple is also pushing in that direction because this is going to be the device that will most likely replace cell phones sometime in the next few years.
44:33Two quick pieces of updates from Anthropic. Anthropic made their first startup investment. And that investment is in a company called Goodfire, which allows training interpreter models to identify concepts inside existing models that aligns perfectly with what Anthropic has been trying to do anyway. So this investment makes perfect sense. Anthropic also released a comprehensive best practice guide for its Claude Code CLI that they released a few weeks back. So if you want to learn how to use it and you want to learn how to use it in the best way, and millions of developers around the world do, you can now go and check out their guide.
45:09Switching from that to safety, two interesting reports show the dark side of using AI tools. A new report by Appnox, which is a company that monitors apps and their behavior, uncovers 10 critical vulnerabilities in Perplexity's Android app, exposing over 10 million users of the app to risks like data theft and account take. Now, I'm not sure if that's unique to Perplexity, but it is very obvious that as you allow these tools more and more access, so Perplexity now has its personal assistant that can replace the Google Assistant on Android phones. It gives it a lot more access to a lot more aspects of your phone, making it significantly more dangerous.
45:46From a timing perspective, that's really bad for Perplexity as they're seeking to double their valuation and raise more money. now at an$18 billion valuation, and they're negotiating partnerships with Samsung and Motorola to integrate their assistance into their phones. So that's not good news, especially at this time. Staying on this topic of security, Hidden Layer, which is a company that develops a solution that monitors what's happening in AIs to reduce security risks, has shared that they have found a universal prompt injection bypass that they have dubbed policy puppetry that allows to compromise all of the major language models like GPT-4, Claude, Gemini, and so on and so forth, allowing harmful content generation, including violating other policies, meaning creating biological weapons, atomic weapons, violence, and so on with minimal effort.
46:37They obviously have a very clear interest to explain that risk because they are the ones who are providing the remedy for that. But it's very important to know that this is what's happening right now. And it's becoming more and more fierce. And we're learning that these companies are releasing models with less and less safety scrutiny because of the competition. Now to a few interesting raises of capital in the past week. ManyChat has announced$140 million Series B. ManyChat has been around before AI and they're serving over 1.5 million customers across 170 countries. But now they released obviously their AI capabilities and this raise will allow them to increase their AI communication with customers across multiple platforms, such as TikTok, WhatsApp, and Instagram, and big companies are already using them like Nike and New York Times and Yahoo.
47:25So it's a company that develops a really great chat product that now is developing really advanced AI capabilities into their chat interface. Another big raise comes from Supabase, which is an online database hosting company. And they just raised$200 million in a Series D, valuing them at$2 billion dollars only months after their previous round of Series C that valued them at only $900 million. But to be fair, they have seen a huge, insane growth fueled by Vibe coding capabilities because they provide a lot of the backend databases for platforms such as Lovable, Bolt, Cursor, and other tools like that.
48:04So if you are developing or you want to develop on these platforms, you're most likely using Supabase databases in the backend without even knowing that, which is driving a huge growth for them. The next interesting raise come from a company called Lace AI. They just raised$40 million and their goal is to transform the house service industry like HVAC, plumbing, roofing, et cetera. What they do is they listen and analyze 100 % of the calls coming into businesses in that industry and they find hidden revenue opportunities they are claiming that are driving between 35 % and 95 % increase in revenue to these companies.
48:40The interesting parameter that I didn't know is that 80 % of revenue in this business comes from phone calls. So if you can identify opportunities in these phone calls, you can obviously drive a lot more business. Very interesting and very specific use case that probably is worth more than the$14 million they raised, but that's a step in that direction. Now, speaking of small companies doing interesting things, one of the most interesting pieces of news this week was that a two-people startup, both of them undergrads from Korea, has released Dia. It's a 1.6 billion parameter open source text-to-speech model that they're claiming, and based on a lot of examples that I've seen, these claims are correct, that it outperforms industry giants like 11 Labs, OpenAIs, text-to-speech, and Google's Notebook LM podcast features.
49:23So it's two people with no funding that has developed a fully open sourced, amazing voice agent capability. This obviously raises a lot of excitement in the world of people who want to use agents, and their tool, very different than others can mimic laughs and sniffs and multi-turn emotional conversations and complex rhythms and makes it significantly more human than most of the tool outs there, including Eleven Labs and the recent fanboy Sesame. So two kids basically developed that. You can go and check it out. It's available on Hugging Face and GitHub under an Apache license, meaning you can use it for more or less anything you want.
50:02Switching back to how these tools are going to most likely impact the future of our world, MIT Sloan Report identifies generative AI as the new way to develop applications for mobile and applications in general. And they're claiming that non-coders, or what they call citizen developers, now create apps using generative AI natural language interfaces. I've been doing this in the past few weeks, with 60 % of new business apps expected to be low-code or no code by 2026. That's next year. So I said many times before on this podcast that I think that the concept of App Store is going to either disappear or erode dramatically because in the near future, we'll be able to create on the fly the specific tools to do the specific tasks that we will need.
50:47And it will be as fast and sometimes faster than actually searching for the right tool to do that task. So that's very similar to what I mentioned before with regards to image generation. If Previously, we had to search through stock photos to find an image that were more or less what we were looking for, but good enough for what we need. The same thing happening with apps right now, right? We spend hours researching and comparing different tools for a task that we need. But what if in less time I can generate the application or the agent that can do the stuff that I need instead of waiting or searching and just creating it on the fly?
51:19And I think that's what's going to happen in the near future. When I say near future, probably 18 to 24 months. And to finish a fun and surprising AI use case, Coca-Cola Freestyle, their machines that are in 50 ,000 locations around the world that enables you to mix all the different Coke products to whatever weird flavor you want, just added AI capability to it. So now consumers can scan a QR code and answer a short question and use digital stickers such as funny or creative and so on. And the AI will generate unique recipes on the fly for them that the machine will pour into your cup. I find this to be a really cool, unique, and interesting use case.
51:57I myself don't drink these kind of drinks. I think they have way too much sugar and other bad things for you, but there are 14 million daily beverages being poured by these machines. So I'm sure there's enough other people who will appreciate this new cool AI use case. That's it. Before we say goodbye for today, I want to ask for a favor. If you enjoy this podcast, please share it with other people who can benefit from it. AI education is becoming one of the most critical aspects of the future of our society and definitely our jobs in the next few years. And all I'm trying to do is to share this with as many people as possible.
52:31And you can help by clicking the share button on your phone right now and sharing it with a few people who can benefit from this podcast. And while you're there, if you're on Spotify or Apple podcast, please leave us a review. I read these reviews. I really care about these reviews and it's your ability to provide us feedback. And speaking of feedback, there's a survey that you can share what you like and don't like about this podcast, helping us to make it better. It's available in the show notes. It will take you a minute to fill out and I would really appreciate if you do that. And one last thing, if you're looking for proper AI training that can accelerate your career and can dramatically change your business, go and check out the AI Business Transformation course.
53:07The next public cohort starts on May 12th. It's right around the corner and the seats are limited and we're almost fully booked. So if you want to take that course, come join us and do it right now. And with that, I will say have an awesome rest of your weekend.
From the publisher
👉 Fill out the listener survey - https://services.multiplai.ai/lai-survey
👉 Learn more about the AI Business Transformation Course starting May 12 — spots are limited - http://multiplai.ai/ai-course/
Are you ready for the AI-powered business revolution — or will you be left managing a ghost town of outdated workflows?
In this weekend episode, host Isar Metis unpacks how AI is reshaping company structures, crushing traditional org charts, and creating a hybrid workforce of humans and digital agents — faster than anyone anticipated.
If you're still thinking AI is a side project, this episode is your wake-up call. The time to act is now — before you’re left behind by companies who are already deploying digital employees.
- What Microsoft’s 2025 Work Trend Index says about the urgent AI shift happening right now.
- Why AI literacy is overtaking college degrees as the #1 career differentiator.
- How “Frontier Companies” are restructuring — and why the traditional org chart is dying.
- The real risk: not just losing jobs to AI, but creating unsustainable business models.
- Critical first steps every leader must take today: hiring digital employees and balancing human/AI ratios.
- Why companies who wait will lose market share to first movers embracing AI at scale.
- Practical examples of where AI is already slashing costs and increasing margins across industries.
- The dangers of hallucinating chatbots and how transparency can save your brand.
About Leveraging AI
- The Ultimate AI Course for Business People: https://multiplai.ai/ai-course/
- YouTube Full Episodes: https://www.youtube.com/@Multiplai_AI/
- Connect with Isar Meitis: https://www.linkedin.com/in/isarmeitis/
- Join our Live Sessions, AI Hangouts and newsletter: https://services.multiplai.ai/events
If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!
