What's going on with AI Agents?

In the next few years, we’ll be onboarding billions of skilled workers into our workforce, and they won’t be humans. They are Autonomous Agents.

Who are these guys, what do they do, and would this really happen? I’d be trying to answer these questions and more here.

Thanks for reading, please subscribe to get simple explanations of the most important stories in technology, economics and more directly in your inbox.

Autonomous Agents: What they are

If you’re reading this, you probably use generative AI (gen AI) through ChatGPT, Meta, Claude, or other foundation model providers. And you most likely use them for research, content creation and augmentation. Or, if you like weird images, you use them to generate images, too. Most of them are question-in-answer-out types of systems.

We’re zooming past those. Now, we’ve got the rise of gen AI products that can not only create texts, images and so on but act on your behalf.

AI is combining reasoning and action. Imagine an invisible co-worker who’s incredibly intelligent; they not only follow instructions but also learn, adapt, and become more effective over time. They can handle large amounts of information at incredible speeds, make decisions independently, and are available 24/7, ready to scale up their efforts at a moment’s notice.

How does it work?

Agentic systems aren’t new. Rules-based systems like “if-then logic” have been widely used in applications where tasks follow a predictable, structured pattern. Think Zapier.

Gen AI agents, however, can accept natural language instructions (specific goals and objectives), create rules to execute, and adapt to new challenges without constant human oversight.

In an ideal world, you have an army of specialized agents that collaborate to achieve this overarching goal.

So for example, you are going for a vacation. You simply chat with a general AI agent about what you’re looking to do on your vacation. This agent would ask clarifying questions to completely understand your request and then delegate specific tasks to other specialized agents.

Goal: Plan a Vacation to Hawaii with Flights, Hotel, Activities, and Budget Management

Agents Involved:

Budgeting Agent: Works with you to determine the budget and allocates funds to various parts of the trip.

Flight & Accommodation Booking Agent: Searches for flights, considers travel preferences (non-stop, morning flights), and compares costs. Also finds lodging options based on preferences like location, amenities, and price. The agent will coordinate with you for final approval.

Activity Planner Agent: Suggests activities based on interests and optimizes scheduling.

Itinerary Agent: Coordinates all bookings, manages travel times, and provides a cohesive schedule.

There are possibly even more specialized agents involved in this specific loop. With this example, you can project the potential in business processes. An agentic system, for instance, can execute a digital marketing campaign via different distribution channels based on specific objectives that you set.

Where are we now?