Transcription
Jensen Hang has just revealed that AI has changed forever. He says, "We have moved from AI that answers your questions to agentic systems that can actually do the work."
Jensen believes that ignoring this shift is like ignoring the internet in 1998. A big mistake. He says businesses that adapt quickly can become up to 100 times more profitable. And the ones that don't won't just fall behind, they'll get out competed to extinction.
And so I studied hundreds of hours of Jensen's interviews talking about agentic systems. So I can break down for you Jensen's exact theory on AI and the future of business. This massive agentic system shift that you must understand to stay competitive and most importantly, how you can set up Jensen style AI agents to work for your business before your head hits that soft pillow tonight.
All I ask in return is you throw in your cleanest black leather jacket, hit that like button down below, and let's jump straight in.
So, to truly understand how we can make the shift from generative AI helping us here and there with our business to an agentic Jensen Hongang style system running our business for us, we must first zoom out and understand this.
All the value in the world that has historically been created has been created through this human flow. Humans first consume calories in our food. We use the energy from the calories to build neurons. These neurons organize themselves in a scalable structure called a brain. And then our brain interprets decades of life information and experiences to build mental models.
Now, these mental models can then look at a new experience, and based on the structure of them and the pattern in them, can generate new thoughts. And finally, these thoughts that we're having can be directed through our hands, through our mouths, through whatever tools that we have, into actions.
Now, what's really important to note is it's this action part of the stack that has created all of the value in the world. You can imagine if we're just humans rolling around having the energy, the neurons, the brain, and the thoughts, but never being able to actually act on these thoughts, no value would be created. But it's because we have access to all sorts of technology and tools that we can actually turn our ideas into valuable things in the physical realm. And in fact, over the thousands of years that historically humans have been creating all the value, we've created $900 trillion worth of value for each other.
Okay, so here comes the fun part. Jensen Hang often talks about AI as a five layer cake. Roll the clip.
"AI invented a whole new industry at the energy level. Um, of course, at the chip level, at the infrastructure level, then it's the model layer. Most of the time we're talking about models, but the fact of the matter is without this underlying part, there is no models. That's useful. And then most importantly is the application layer."
So this is where things get really interesting. Just like humans need energy from food, so does the AI. But the AI obviously doesn't eat calories. They get their energy from nuclear, from solar, or from fossil fuels.
That energy is then transformed into computational power by Nvidia's chips that can convert energy into mathematics. They are the individual neurons of an AI.
Then you have infrastructure, which is everything that allows these chips to scale. Things like warehouses, racks, cooling mechanisms, or even chip coordination. This infrastructure or data center is AI's version of their own brain.
Now, if we pause here for just a second, it is fascinating to note that this stack makes a storage and retrieval-based computer that we've had since the 1950s. Jensen refers to them as retrieval style computers. Roll the clip.
"The way we used to do computing is essentially retrieval-based computing. We would pre-record information, take a picture, record a video, uh write a story, do a brief, so on so forth, and we put it online. That's why we would put it in a data center. Then once you select, it would go and retrieve that file for you. It's all retrieval-based computing."
But the problem, as Jensen highlights, with this computer is that they can never create anything new. They are only storage vessels for data that we've previously put in them that can retrieve them when we ask the computer to go and get that file. And therefore, Jensen says they cannot think yet. They do not have their own thoughts.
But this all changed in 2022 at an event that I'm sure you remember, the announcement from OpenAI of an app called Chat GPT, which was, as Jensen says, the world's first generative computer, now known as generative AI.
"AI is a new computing paradigm. The way that AI operates is so fundamentally different than the way computers operate. It no longer uh stores the information exactly as you described it and retrieve it from storage. The way that we do uh computing now with artificial intelligence is based on your context. It would then generate from some initial seed of information. It could be grounded in some fact. It could be grounded in a piece of report and it generates information for you for the very first time. And so it's generative computing."
And as Jensen says, generative AI was the first time ever that a computer could think of something new. And the big breakthrough is now these computers are trained on massive data sets that become models. Just like the mental models we have in our brain from our life experiences, we can now give the AI these data sets to have their own models that they can index into complex vectors and allow those models to be able to predict what data would come next in a sequence.
And for the first time, the actual prediction of the data that might come next is the first time that AI is creating new thoughts. And so, just like when your mental model in your brain sees the sentence "I turn the handle and open the," it will likely output the thought of a door. Well, a generative AI can now do the same. They look at this sentence and they see statistically the most likely next word in this sentence is a door.
But most people don't call these thoughts. They call them tokens. Which, according to Jensen, who collapses these ideas into one, which is his brilliance, says that a token is an AI's equivalent to a human thought.
"The ability to generate has two profound uh capabilities. One, in order to think, you have to generate tokens in your mind inside."
"What do you mean mo tokens?"
"Tokens. Oh, uh you have to generate thoughts. And so the fact that our ability to think and reason requires us to generate thoughts. So the moment we got AI to generate, we realized you can now think."
Now, according to Jensen, it took another 2 years before AI had the next breakthrough, which took us from generative AI to something called reasoning AI, which in hindsight seems like a pretty simple step to take because reasoning AI happened in 2024 when AI researchers actually used these thoughts or tokens that the AI was having to turn them inward, to start to plan, start to reflect, and start to workshop its own ideas before spitting them back out, which allowed us to get even smarter.
Other AIs outputting more impressive answers, and you can see when Chat GPT or Claude is loading now, it'll actually show you the thoughts that they're having and reasoning with as they are working through those thoughts, improving the potential output that it's going to give you back. Now, generative AI, as you know, generates tokens for internal consumption as well as external consumption. Internal consumption is thinking, which led to reasoning.
Now, this is where most business owners are currently stuck, where they are using something like Chat GPT, Claude, or Grock, which is a reasoning AI that is helping them come up with ideas or doing small things in their business. But the actual action taking is then redirected back to the human.
But recently, the latest breakthrough just occurred this year in AI, once we equipped these reasoning models with the tools and the applications to actually take action in the real world on their thoughts. And with the right tools being equipped to them, AI agents can now do anything a human can do on a computer.
"We now have for the very first time useful AI agentic systems. The AI has to understand, has to reason, has to think, it has to plan, use tools, look at the results of the tools, think about it some more, come up with maybe an improved plan. And so it iterates until it can get the job done using a whole bunch of different tools."
And if you think, okay, that's amazing. AI has now caught up to humans and now can do everything they can do. Sort of like I just said in this video, you and I would both be wrong.
Because unlike humans, AI don't have to go to university and study for years to get the mental models and the intelligence that they need in order to have the right thoughts and actions in a situation. In fact, you can take as much data or as much intelligence as you want and copy them between agents instantly, making them 10,000 times more powerful than an actual human.
And more than that, humans can only take action for about six to eight productive hours a day before they fatigue and get tired and want to go and watch Tik Tok. And we're also limited by our hands, our thought speed, or even our typing speed. But AI agents will work 24/7 around the clock and even currently with the current computational capacity that we have, still be able to type at a 150 words per second.
So this is why Jensen is saying it is critical that we pivot to an agentic style business before our competitors do.
So, how does Jensen recommend that we implement an agentic system for our personal businesses? Well, in my hundreds of hours of research, I put together four Jensen golden rules for implementing AI agents into your business that are going to reliably work for you and save you time and energy so you could wave goodbye to your competitors as we lift off together in this rocket ship.
Starting with the first golden rule, which is kind of an awesome one. Jensen says, "You must first refuse the constraints of what used to be possible."
Because according to Jensen, all of business history and all of building anything was limited by two things, capital and employees. But now AI agents have removed this limitation. It's no longer about people or money, but it's about your creativity, understanding, and your ability to point tokens through agents at the most productive work possible.
And he infers that the trap that most business owners are falling into is being constrained by the old world's thinking. They're asking questions like, "How do I do what I already do but put less time into it?" Instead, Jensen urges you to use it as an expander, asking, "What can I build now that was impossible before?"
"What would you build now if you had $30 million of capital and 50 individual experts by your side working 24/7?" You're going to start to think much bigger.
"Whatever ambition that you had in the past, it just not high enough. The fundamental change that we have to make and I have to make is whatever expectations I have for the company, you've got to increase it by about 100x. If people tell me they can do something, I've got a 100x in my head now."
And now, if I haven't made it clear yet, these AI agents can literally be plugged into any area of your business. From email to CRM to Photoshop to video editing to finances to research to accounting to calls or even social media. The rule is if a human can do it on a computer, an AI agent can now do it for you, which I know can be probably more overwhelming than putting your mind at ease.
Because this leads to the question, where the hell do we start? which would take us to Jensen's second golden rule.
Jensen recommends that you first start by finding what he calls the essence of your business. This is the unique core value that you offer that agents will never be able to touch.
"But what I would do is I would go find out what is the essence of my company. What's the most impactful work that we do in our company? Don't don't mess around. Don't mess around with with peripheral stuff. So once we identify the essence of your business, which is what is the value that you're actually offering, not just in the old paradigm, but going into the new one of having agents do most of the work for you, what area of the business will AI agents never touch? And then once we have that essence, we mark out everything that's not in that essence that has repeatable tasks in them. These will all be categorized and grouped into workflows that we will build these agents to do for you."
Now, I've also created you a completely free tool. In fact, it's free for you. It's not free for me cuz I plug my own API in here. So, you could use it for free. But if you go through this process, I've set you up the Jensen style questions that you can answer in plain English in order to find the essence of your business and then mark out everything that you're either repeatedly doing or taking a lot of your time that AI agents could be built around and deployed straight away. The link to that is in the description below.
But whether you're doing it manually or using my free tool, the next step is to start prioritizing. Because you'll probably find you'll have five to 10 agents that you'd like to immediately build. Jensen talks about how he wants every one of his employees to have a 100 agents working for them. It can once again be a little bit overwhelming. So, let's go to Jensen's golden rule number three.
You must aim your AI agents at your most productive tasks first.
So once you've used my tool to figure out the exact priority of the AI agents that you're going to start implementing, or maybe you've worked with your own cord or Chat GPT to get your own manual list sorted, we now have a crystallized step-by-step road map of deployment in order to take you from a generative AI business into an agentic system.
Now, I've also seen a lot of business owners out there try to implement AI agents completely wrong. And this leads to, of course, more time consuming, more energy, and more headaches than never even touching agents in the first place. No point deploying an agent if it's just going to give you back a lot of technical bugs or it's going to be down half the time. You don't know why, and you have to get someone in to try and fix it. In fact, I tried to outsource my email list at one point and did it completely wrong, and it just started sending out some very cringy emails.
Which is why I personally highly recommend you only deploy one agent at a time and you work on this agent until its workflow is perfected so that you never have to touch that agent again. Because if you give them the right guard rails, the right context, and the right configuration, you can ensure that they will do a phenomenal and a consistent job that won't require more of your time and energy in the future.
And how do you know when you've done this? Well, you'll literally feel an excitement and a bubbling and a sense of freedom inside because you know you've actually claimed back more of your time and energy.
And so the components that you actually need in order to set up that first agent are an LLM to power them, somewhere to host it, your business context and what your business does and what it stands for. Then the agents' individual workflow context, a set of instructions that it is to follow every single day. And then finally, a self-learning loop so that every time it actually runs that task, you can either give it feedback or it can take feedback from its environment and improve on itself so that every time it runs a task, it actually runs it better than the last time.
Now, lastly, I've seen a lot of business owners try to outsource this AI agent implementation to either a company or even just a young teen consultant, which I personally think is a massive mistake. Not only because they don't know your business as well as you do, but you are forever in the future dependent on them and their energy to come back, upgrade your bots as technology advances, or even inform you of future opportunities. All of these things which you would have complete control over if you were the one who was competent with AI.
It's like having to hire Harry Potter to come and do magic every time you need some magic done instead of realizing that you are actually a wizard.
"You're a wizard, Harry."
And learning this magic and learning AI is so much simpler than you think. And I'm a massive believer, just like Jensen, in the next three, five, 10 years, businesses that are winning are going to be led by AI competent founders and owners, not AI ignorant owners that are having to hire people to come in every time they need a small upgrade or a tweak. It just will not make sense. And I'm guessing you must agree if you're here watching this video.
Now, if you do need help setting up this LLM, somewhere to host the agent, making a business context, making the agent context file, and having them be self-learning agents, as Jensen recommends, in this free AI tool I recommend, we are going to help you get started.
If you would like even more help building AI agents who actually go out there and can start taking on large, repetitive tasks in your business better than you could possibly do it. Set up with the right guard rails, the right context, and the right consistent configuration personalized to your business. Use the link in the description below to join our AI agent workshop where we will guide you through the entire process, troubleshooting any problems you have so that you have your first AI agent working in your business in literal minutes.
We've worked really hard on this system to make sure we're hiding all the technical jargon that you don't need while still giving you the education that you would like in order to deploy the second, third, fifth, 10th, or even hundth agent for your business once we've done the first together.
And if you do join our AI agent workshop, I guarantee we will shift you from a generative AI style business into a Jensen style agentic system business that you can be confident in, or we will personally work with you until you have it to join our AI agent workshop and use our software to set all this up as simply and as effectively as possible. The link is in the YouTube description below. Go ahead and click that and come and join us. If you have any questions or things that we need to troubleshoot along the way, we'll be happy to assist you at every step.
Now, whether you use our workshop or you go and do this manually using the free tool that we are giving you once you have your first AI agent set up and working for your business, we are now up to Jensen's golden rule number four.
Burn tokens abundantly against your targets.
"Let's say you have a software engineer or AI researcher and you pay them $500,000 a year. At the end of the year, I'm going to ask them, 'How much did you spend in tokens?' And that person said, ' $5,000. I will go ape something else. If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed."
So Jensen says pretty clearly, once you have these agents set up, you must run them to their maximum capacity in order to get the most out of them. Don't be precious about it. Don't try to conserve money on tokens. If they are pointed at productive tasks, they should far outearn the money that you will spend on tokens. In fact, Jensen says that each of his engineers should be burning a quarter of a million dollars in tokens every single year.
And honestly, depending on the size of your business, like even my business, which we do seven figures a year and we use a decent amount of AI agents in it, we're not really going further than Chat GPT's pro plan. So half a million of tokens a year will be a crazy amount of silicon being grinded.
And so Jensen says the same logic will apply to you. Every token that you spend is replacing a fraction of a human hour somewhere in your business. So, the maths is wildly in your favor every single time. Max out and crank that token lever.
So, to recap, make sure you're not being restrained and constrained by the old thinking of your business before you have access to literally the equivalent of millions of dollars of capital and thousands of employees. Then, define the essence of your business. This is the thing that is going to stay true no matter how many agents you have working for you. Then outsource any repetitive or time-consuming task outside this by building your first AI agents. Whether that's with us in our workshop, once again link in the description down below, or if you want to do it by yourself with your AI model. And then fourthly, and finally, crank that token lever and get the most out of your AI agents in order to accelerate your business far ahead of your competitors.