Agentic AI Finally Gets to Work

Back in January, I predicted 2025 would be the year of agentic AI.

And there was plenty of hype coming from Silicon Valley to back me up.

In fact, Nvidia’s CEO, Jensen Huang, also insisted AI agents would “take off” this year and become a multi-trillion-dollar industry.

But by spring, it was starting to look like my prediction might be wrong.

Because what passed for an “agent” back then was mostly just a chatbot with a couple of extra tricks, like booking a trip or answering questions from the web.

When you peeled back the hype, most of these so-called “agents” weren’t really doing that much. They might have been useful, but they weren’t systems capable of taking on a job.

But now that we’re nine months into the year, the picture is starting to look different…

And I’m confident my prediction was correct. Because today’s AI agents are actually getting to work.

From Talk to Tasks

A turning point came on September 3, when OpenAI launched GPT-5 Codex.

Before this release, autocomplete tools like GitHub Copilot were good at filling in a few lines of code. But the grunt work of coding still had to be done by humans.

That’s why software engineers are calling Codex a gamechanger…

Because it’s able to do the stuff that slows coders down, like hunting down bugs and rewriting broken code.

In benchmark tests, Codex solved nearly three out of four real-world coding tasks. That’s more than double the hit rate of earlier models.

Codex also proved it could rewrite bad code with far greater accuracy. And in one trial, it kept working for seven straight hours, debugging, testing and pushing fixes without needing to be nudged along by a human.

In other words, Codex already acts like a junior engineer.

But OpenAI wasn’t the only one raising the bar. Just three days later, Amazon unveiled two new agents of its own.

The first, called Seller Assistant, automates the grind of running a storefront on Amazon. With this new agent, things like inventory management, product listings and customer returns can all be handled by AI.

Amazon also dropped an agent called Ads Creative Studio that can design entire ad campaigns. Now, instead of hiring an agency, small sellers can spin up copy, images and targeting suggestions instantly.

Codex, Seller Assistant and Ads Creative Studio are examples of how AI agents are starting to prove their value in the workplace. We can now take tedious things that typically eat up hours of a human’s time and hand them off to an agent that never gets tired or distracted.

Meanwhile, Anthropic is pushing agentic AI forward from a different direction.

It added persistent team memory to its Claude AI, which enables organizations to give their AI context that sticks across conversations. This means users don’t have to start every new chat by re-explaining their project from scratch.

Anthropic also quietly pushed Claude into Apple’s Xcode, where millions of developers already build apps.

And the importance of this move can’t be overstated. Instead of forcing workers to learn a new tool, Anthropic is embedding agents into the software they already use.

That’s how adoption spreads.

It’s why I wasn’t surprised when, on September 10, Accenture announced it would train its entire global workforce on agentic AI.

All 700,000 employees.

This means every client Accenture touches, from banks to manufacturers to governments, will now see agentic AI woven into their workflows.

But the workplace isn’t the only place where AI agents are proving their value.

This month, Coca-Cola announced a partnership with MIT to use AI to help battle a disease called “citrus greening” that has devastated Florida’s orange groves.

According to USDA data, the U.S. is producing only 2.2 million tons of oranges in 2025, down from 5.5 million in 2015.

Orange production in the United States from 2000 to 2024(in 1,000 tons)*

Source: Statista

If this trend continues, oranges are in danger of going extinct. But Coke is betting that AI can help save them.

At the same time, NYU Abu Dhabi researchers revealed an AI model that can forecast dangerous solar storms up to four days in advance. These storms threaten satellites, GPS and even the power grid.

Predicting them buys time to protect critical infrastructure.

This shows you how AI agents aren’t just tools for boosting productivity.

Just like Codex works through bug fixes without human intervention, these systems are working on problems without waiting for a prompt.

In labs and observatories, they’re acting like junior researchers. The AI is crunching decades of agricultural data or scanning solar activity, spotting patterns and coming up with solutions in ways that could take humans decades.

That’s what makes them agents, not just algorithms.

Here’s My Take

We are now two-thirds of the way through 2025, and we have a much clearer idea of where agentic AI really stands.

AI Agents are reviewing code. They’re running ad campaigns and managing small businesses. They’re being trained into the daily routines of hundreds of thousands of workers. And they’re helping us solve complex global issues.

Put it all together, and September seems like a tipping point. It feels like the moment AI moved from the demo phase to the deployment phase.

But the story isn’t finished. In fact, this recent leap in agentic AI is all part of a bigger pattern.

You see, most people think tech breakthroughs happen one at a time…

A new chip here, better software there, faster internet somewhere else.

But that’s not how the biggest wealth creation actually works.

George Gilder has spent 50 years tracking these patterns, and the math is simple:

The largest fortunes are always made when multiple technologies converge at the exact same moment.

When the railroad met the telegraph, it created America’s first connected economy — and built the Vanderbilt fortune, worth $185 billion in today’s dollars.

When the microchip met the internet, it launched the digital economy — and created Microsoft, Amazon and Google.

But what’s happening right now is completely different.

For the first time in history we’re watching eight breakthrough technologies all accelerating together.

Any one of these technologies on its own would be enough to create massive wealth.

But when they start feeding off each other — making each other smarter and faster in ways we’ve never seen before — that’s when you get what we’re calling “Convergence X.”

My first message with George on this rare “super convergence” has already reached over 2 million people in just the last month.

But since then, I’ve realized our prediction was wrong.

It’s actually happening MUCH FASTER than we initially predicted.

Which is why I’m hopping on a last-minute flight with my team to the Berkshire Mountains of Massachusetts.

Because the smart money already sees this super convergence coming.

Mark Zuckerberg just committed $65 billion to convergence infrastructure.
Larry Ellison landed a $100 billion contract for something called Project Stargate.
Jeff Bezos liquidated billions from Amazon to position in smaller convergence plays.

They understand what most people miss…

The companies making the convergence possible often outperform household names.

During the internet boom, everyone chased Microsoft and Amazon. But Cisco quietly gained 75,000% over the same decade.

Today, everyone is piling into Nvidia at $4 trillion. But the real winners will likely be the small companies enabling Nvidia’s success.

And in just a few days, a bombshell announcement could flip the switch on this convergence once and for all.

I want YOU to be prepared… while there’s still time.

Please be sure to clear your schedule on Monday, September 29 at 1 p.m. Eastern time as I join George at his home for an urgent update.

You can’t afford to miss it.

Click Here Now To Reserve Your Spot.

Regards,

Ian KingChief Strategist, Banyan Hill Publishing

Editor’s Note: We’d love to hear from you!

If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to [email protected].

Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!