Show Notes

https://manus.im/ https://www.anthropic.com/ https://www.anthropic.com/news/model-context-protocol https://mvara.ai/ https://codeium.com/windsurf

About Jack Crawford

https://www.linkedin.com/in/jackcrawford/

Transcript

Agents AI Tom Koulopoulos [00.00.02] Welcome back to Foresight Radio, where we dive deep into the technologies shaping our world and explore how they're redefining the way we work, live and lead. I'm Tom Koulopoulos, and today we're going to talk about digital workers, not just digital assistants or chatbots, but full on a gigantic AI, the kind of AI that can coordinate projects, make decisions, and even negotiate contracts. If you've been following AI lately, there's been a lot of news about agent or agenda based AI. At the heart of AI is this notion of giving AI agency, giving it the ability to make decisions within certain defined parameters. OpenAI is reportedly planning to offer specialized AI agents to businesses that could run up to $20,000 per month. So why the hefty price tag? Because these agents are meant to handle everything from sorting sales leads to software development and engineering to even PhD level research. SoftBank one of OpenAI's investors, is planning to spend up to $3 billion on a gigantic AI products this year alone. Also in the news are developments out of China, where Magnus AI is reportedly taking a somewhat different approach here, not unlike what we saw with Deep Seek R1 mantises using an open source model for its agenda guy. The reason this is important is because a lot of the power in a genetic. I could in very much the same way as cloud based approaches, allowed smaller and medium sized businesses to deploy technologies that otherwise were inaccessible to them. The accessibility to agent based technology is a real leveling factor in creating competitive advantage for smaller and midsize businesses. It sort of supercharges smaller enterprises and startups. But back to a gigantic AI. This goes well beyond the basic Q&A chatbot that we're all used to with ChatGPT and other forms of current AI that we're using on the day to day. To give you a sense here, you'd give an agent a big picture goal the same way you want to an individual that you've just hired to take on part of your business. So let's say the goal would be boost sales conversions by 10% this quarter. It would then figure out, as would a very smart individual, what are the tasks, the data, the resources that I need to able to make that happen. So think of it sort of like a project manager. It operates within defined autonomy. You set the guardrails and then it does the rest. And this is not just theoretical. This is happening today. Salesforce is doing this with Einstein. GPT, which analyzes large volumes of customer data and automatically drafts personalized sales emails and marketing messages. IBM Watson Orchestrate is also doing this. It's an AI system that automates repetitive business processes, like scheduling meetings or approving invoices or triaging emails. And instead of constantly pestering a manager with and is this okay? It just takes the action again within these predefined parameters that you've provided. With a jet ski is also being used for code development. For example, Amazon's Code Whisperer, which is effectively a software engineer that you can use to help make suggestions on existing code or to create code. If you really want to get a sense for what autonomous code development looks like, check out Windsurf by Coding. It's an AI development environment that allows you to code by using an interactive developer agent. You communicate back and forth using prompting and standard English dialogue, not unlike how you would if you were talking to an IT professional trying to describe to him or her what your application should do, what it should look like, and what its requirements are. Now, of course, we can't talk about any of this without also addressing the issue of what will all this do? What will a tech technology do to knowledge worker jobs in the overall job market? To get some insight that I talked to Jack Crawford, who has been deep into AI for some time now. Jack is the founder and CEO of a company called Kinvara X, Wells Fargo Exec and an Air Force veteran, which is actually going to play a role in the conversation that we're about to have. He's also a bit of a windsurf maven, and he's the one who introduced me to windsurf. Here's some of what he had to say. Jack Crawford [00.04.10] Every technological revolution, there is a disruption or constructed disruption of the economy. This is the basic economic principle that you don't get something new without getting rid of something old. The word agent is being used because it is such a radical step from passive large language models. ChatGPT is a passive ball in a business. Chat doesn't work just like the talk is not. The action is everything and an agent acts when automation is given agency, it doesn't need a human to prod it along. It doesn't need a human to check its progress and doesn't need a human to evaluate its outcome. All of that can be automated. This is Tom Koulopoulos [00.04.52] where my conversation with Jack took an interesting turn. I asked him about the advantage that agent technology can give businesses, and his response brought up one of the key figures in aviation warfare, Colonel John Boyd, who created what's called the OODA loop. OODA stands for observe, Orient, decide and act. It's a method of dogfighting in which the fighter pilot realistic, continuously outpacing opponent, and it's become part of the lexicon in many cases where speed of innovation is the key to success. While OODA is in part about changing behavior so that you're constantly re-evaluating data and making decisions quickly in this OODA loop, it's also about applying technology. And if you think about aviation, clearly, there's a lot more technology involved in today's fighter pilot. And there was when we were moving from propeller to jet airplanes. Same thing applies to business. And this is where I plays a crucial role. I accelerates that OODA loop not just by helping you to make better decisions, but by actually taking over some of those decisions. Colonel Jack Crawford [00.05.52] John Boyd was an Air Force officer who was a pilot, and he was one of the key principles that took us from. Propeller driven aircraft to jet aircraft in a operational way. Why? Because it's just like Top Gun. You get behind the opponent before they see you. Then you can shoot the tail. That's what business needs to do with AI. They need to shorten that loop. So we're going to see the businesses who leverage our intelligence of today identify or outperform the competitors. And they'll do it so quickly the competitors will fall. Tom Koulopoulos [00.06.27] The point here was that the ability to observe, orient, decide and act is often something done in the trenches, minute to minute, hour to hour. A genetic AI gives us that ability at the periphery of the organization, where we constantly need to identify new patterns and make real time adjustments to the organization's behavior. Think of this as an autonomic nervous system for the organization, where the organization is given agency to act without the need for centralized authority to make those decisions. The issue here, of course, is how fast will all this happen? According to Jack, faster than anything we've experienced before. Jack Crawford [00.07.02] What is happening, from my vantage point is that AI is advancing daily, not weekly, not monthly, not yearly. There are not. We don't have five years. In five years, much more will be accomplished. But we are months to a year away from an inflection point where you won't be able to do your job successfully without AI. But I will be able to do your job successfully. You Tom Koulopoulos [00.07.31] did you catch that last comment? You will not be able to do your job without AI, but I will be able to do your job successfully without you. And this is where things get really interesting and perhaps a bit scary. One of the areas that is not often talked about, at least it hasn't been up until now, is how agents will work with each other. And this is where we'll see the greatest acceleration of AI, in large part because as AI Asians start to communicate with each other, they'll be able to accomplish tasks at a much faster speed, a much greater accuracy than when humans are involved in the mix. And if this sounds a bit frightening, I'd ask you to stop for a minute and consider how you work today. I'll bet that there are countless tasks that are done by specialized services that you subscribe to, by individuals or teams that you work with who are far better and faster doing that work than you could ever be. That is the very nature of specialization. You can't do everything. No one can. So we build organizations that allow us to tap into the core competencies and experiences of each individual. The challenge we have is trying to stitch all of these decisions with these numerous specialists, this network of professionals with different core competencies into an organization that is flawless in how it communicates. Over time, agents will become highly specialized in individual tasks, and then higher order agents will coordinate the work of these specialized agents, forming a sort of organizational mesh of. AI agents. There will even be agents that oversee and provide governance for these networks. Much of what I just described is going to happen over the next 12 to 24 months, which means you can't really wait this out. In the Jack Crawford [00.09.12] past, workers had plenty of time to adapt. You can spend the 5 to 10 years resisting change, so contributing in a good way and then retire. It's not going to happen this way. I see no crisis. If you're able to leverage, I could use ten times and that'll work. You can do. For now, you've moved to the top 20% of productivity in the company rather than in the middle. And we know that roughly 10 to 20% of people do 90% of the work in accounting. Every single company, every employee who joins the company doesn't become the CEO. Only if you did. And that's natural selection. It just happens. And we haven't objected to this since the Industrial Revolution, so let's just move on with it. But now we're dealing with technology where the AI itself is competing for jobs. It doesn't have to be like a human. It just has to do the work And we already see that happen. You're more like what you're doing now is difficult to automate, but not impossible. Tom Koulopoulos [00.10.06] So where does this leave us as humans? What work are we left to do? One of the things that I'm always amazed at is how outstanding we are at underestimating the new value that we can create as humans. We've seen it in every prior economic revolution, from agriculture to industrialization to computerization. In each case, we fail marvellously at seeing the value that we can create. Value creation doesn't come to a grinding halt because of a new way to automate human involvement. Instead, we just move up onto new ways that we can create greater value. Still, I get us challenging to fully understand what those new ways will look like. But if you really want to stretch your imagination, consider a scenario where AI creates so much national wealth on a scale we've never before experienced. I find Jack's take on this to be a truly fascinating scenario. Jack Crawford [00.10.56] If the gross domestic product of a country is radically increased as a result of automation that everybody gets rich as far as like it's like being in, uh, what is it, Norway that has oil revenues? Everybody gets a check every year, and in Alaska, everybody has to check every because they have a natural resource. They can give it back to the citizens. So you'll want to be a citizen of the United States because the United States is the lead in this area. Tom Koulopoulos [00.11.22] What do you think? Far fetched, perhaps in the context of what we know. But AI is going to change so many of the foundational elements of the way economies operate and the way that we work and create value. The bottom line with all of these sorts of agent models is that digital coworkers will free us, free human beings, to do what we do best, to tackle big picture thinking, to strategize new markets, to create new value. Think of all the day to day tasks that grind us down. Energetic AI thrives on these. There's no doubt that agenda guy will require. Guardrails, compliance standards, data usage policies, ethical frameworks. All of this is very much part of the conversation, and you definitely need fail safes you don't want. I misinterpret a slash expenses. As for half of the organization, it's about responsible autonomy. You define the sandbox the AI plays within that sandbox. But we will see over time. That is, AI takes on more and more of these rudimentary decision making processes within the organization. We will be able to move humans up into higher value work. We are at the brink of where digital workers are no longer niche experiments. They're mainstream ready. In the coming years, you'll see entire teams run by AI that delegates tasks to both human and digital teammates, and they've done right. It'll be a powerhouse. Things like managers entrance into the space are also going to speed up the adoption curve with open source models, smaller businesses and even individual developers will be able to explore energetic AI without paying eye popping fees. And that could democratize access and fuel a new wave of innovation from unexpected corners of the world. And that's one of the reasons I find agenda AI to be so promising. It is the ability to democratize AI in a way that we simply have no precedent for giving individuals the capacity and the ability to do things that would otherwise take an entire organization, what that sort of innovation will look like. Honestly, I can't wait to see. Thanks for listening. If you're enjoying Foresight Radio, be sure to subscribe and share it with friends and colleagues. The best way to navigate the future is to keep asking questions, embrace change, and seek out new perspectives. Until next time, I'm Tom Koulopoulos. As always, stay curious.