AI Agents: Get Ready For A New Era of AI

Did you know that ChatGPT was launched in 2022, just two years ago? By making generative AI (GenAI) models accessible to everyone, it ignited a digital revolution. Today, it’s hard to imagine a world without it. But in the fast-paced digital landscape, there’s always a new innovation around the corner, and what’s next promises to drastically change our world. Forget AI chatbots — AI Agents take AI’s role in our society to the next level.

What are AI Agents?  

AI Agents are independent tools that use Large Language Models (LLMs) to understand given goals and generate tasks to achieve those goals. They make decisions autonomously and learn continually from their experiences and mistakes. So, unlike traditional current models, they don’t just respond to prompts but act on them.

The potential of these AI Agents is immense. You can configure one with a prompt and it could, for example, take over your mailbox. It will then be able to open your emails, analyze their content and respond to them professionally, all while still keeping you in the loop. Even filling out forms can be automated. How is it able to do this? AI Agents continuously take screenshots, so it knows where to find what it needs to complete a task.

This level of autonomy and utility sets AI Agents apart from the current, and simpler, models like ChatGPT, which only respond to prompts. While impressive, AI Agents take this much further and do not require manual input and guidance, but take initiative, manage complex workflows and are continuously refining their performance.

Several organizations are in the race to become a frontrunner in this field. Microsoft is working on Copilots and Google is developing Jarvis, an AI Agent that will be able to book vacations for you. OpenAI will release Operator in January. This already shows that AI Agents will transform how we plan and organize our lives in the future. While these developments signal that 2025 will be the year of AI Agents, Anthropic has launched Claude 3.5 only 1,5 months ago. This model is already showing us what will be possible with AI Agents in the future.

Beyond Generative AI  

Imagine a tool that could do anything a human can, whether this is writing a story, fixing a bug in code, answering mails, and much more. This is the idea behind Artificial General Intelligence (AGI). It does not only follow instructions but will be able to handle the work of an entire organization, fully autonomous. AGI represents a groundbreaking milestone in AI development and OpenAI has laid out 5 steps that will lead us to this goal.

The first step towards AGI is conversational AI, for example chatbots, which can answer questions and hold conversations in natural language. These, however, rely on specific instructions and lack any kind of autonomy.

The second step concerns “reasoning” AI systems. These can solve complex problems and are designed for analytical thinking but are still task specific. An example of this is OpenAI’s own o1 model, which can solve mathematical and scientific problems by executing each complex problem step-by-step.

The following, third step takes another leap forward, to AI Agents which can act autonomously for an extended period of time without supervision. Even more fascinating, they can communicate with each other and work together to tackle even more complex tasks. For example, think about an Agent that analyzes data, while another drafts up a report and yet another one organizes the logistics of it all. They all work together in harmony to work towards one, shared, goal. In other words, this is called Agentic AI.

Step four concerns AI that can develop new ideas that will expand human knowledge in a meaningful way. These can be new ideas, content, products or solutions. These will drive innovation to a new level by, for example, developing groundbreaking or revolutionary technologies in various industries.

The last and fifth step is AGI itself. As mentioned before, this will be a kind of AI that will be able to do the work of an entire organization by itself. This will include strategizing, managing resources and handling complex operations all on its own.

OpenAI believes that we are currently nearing the third step with AI Agents already demonstrating remarkable potential. This signals a future where AI will transform our daily lives and industries in ways that are even more impactful than they already do now.

Embacing AI Agents - A First Test with Claude 3.5  

AI Agents are taking automation to a whole new level, and at Squadra we have recognized this potential early on. We have already started experimenting with ways to use AI Agents for scraping product information from a webshop.

Giving Claude 3.5 a goal, which is to scrape product data from a webshop, it went to work on its own. The model can autonomously search for the right coding libraries to use, how to open and use a webbrowser and where to type in the address of the website. It then analyzed the website and its layout, finding where the relevant data was located and started to retrieve this.

Of course, the process was not entirely flawless at the first try. The agent failed a few times at first, but what is impressive is how it learned from these mistakes. It reviewed what it did wrong and then adjusted and improved its approach. After around 10 minutes the agent had a fully functioning script and reached its goal to scrape data from a website.

This level of autonomy is a real game-changer, as automating these kinds of processes can also reduce costs and improve efficiency in a way we have not seen before. Up until now, this kind of automation was not possible. This also shows that we have a future ahead of us with an immense amount of opportunities, and this right here is only the beginning.

The Future of AI Agents  

As mentioned before, we are rapidly advancing towards Artificial General Intelligence. This progress is so quick that we will achieve this milestone within the next five to ten years. Of course, there are still some challenges to overcome. We have also seen this in our own experiment. However, what sets these models apart is that they learn from their own mistakes and improve on them, making these models better and better with each iteration.

Nevertheless, it is also important to keep caution and listen to experts in technology. For example, former CEO of Google, Eric Smidt, has emphasized that if AI Agents reach a point where they operate together without human oversight, abandoning the human-in-the-loop principle, we must be prepared to intervene. To quote him: “At some point, people believe, these agents will develop their own language, and that’s the point when we don’t understand what we’re doing . You know what we should do? Pull the plug.”

With responsible development and considering the ethical implications as well, the future promised by these AI Agents is incredible. Current roles and functions will be changed by it in an apparent way. For example, customer service will be improved by being able to remember customer history and preferences to deliver tailored answers and solutions to clients 24/7. Factories will run more efficiently as agents will oversee supply chains and monitor quality control to optimize workflows and cut costs. Administrative work like scheduling and data entry will all be fully automated which will free professionals to focus on innovation and strategy.

The possibilities go on and on, as these agents are improving daily as well. The key to success lies in implementing these advancements responsibly, ensuring AI keeps working alongside humans to build a better and more efficient future. The next few years will be crucial and will show us how AI Agents are going to influence and shape our future.

AI Agents are revolutionizing industries—are you ready to harness their potential? Contact us to explore how AI Agents can transform your business!