Agentic systems

AI will affect almost 40 percent of jobs around the world, replacing some and complementing others. We need a careful balance of policies to tap its potential said the recent IMF report on how AI will Transform the Global Economy. In the analyst-speak top down description of economic impact we can lose sight of how this will actually happen from the ground up and how it is already emerging all around us.

Consumer experience will likely transform ahead of our business experience due to the more distributed decision making that happens as we each decide to do something differently because of convenience and price, whereas companies must navigate long standing practices, regulatory restrictions, unions, and individual workers who weigh the risk of job loss against the advantages of new technology.

Not withstanding the regulatory issues that are faced as well in the consumer space such as those related to autonomous vehicles such as Waymo, the ability for us to transform our personal lives with physical and digital agents is already here (albeit with the Gibson caveat that it is unevenly distributed).

Companies such as EDAG are working to bring robots into our physical human environments over the next decade, companies are experimenting with land and sky based delivery drones and we will soon see autonomous robot workers mowing the grass amongst other tasks around our neighborhoods.

So too will our digital environments begin to fill with autonomous digital workers. This is going to happen in 2024, not a far off time. Just as with physical automation where we have had some clunky industrial machines for decades which are now being updated to sleek new human-friendly forms (such as this robot from Figure) we have also had "robotic process automation" (RPA) for decades which will now be replaced with large action models (LAM) from companies like Adept and Rabbit. We are seeing a fundamental innovation in how a computer system can interpret and interact with user interfaces originally designed for human interaction. The Rabbit website describes this in the following way:

We have developed a system that can infer and model human actions on computer applications, perform the actions reliably and quickly, and is well-suited for deployment in various AI assistants and operating systems. Our system is called the Large Action Model (LAM).

They have also provided a helpful video with an illustration of "training" Rabbit on a workflow - definitely worth watching to understand where this technology is today and sparking your imagination for where this is headed because this is just the first step towards autonomous agents - building a team of digital assistants, each with their own abilities, to work on different parts of a task.

A range of new platforms are being launched (two I've recently seen: Taskade and Newo.ai) which let you easily create and customize agents to help with different tasks, like research or content creation, without needing to know how to code. You use natural language to tell the agents what to do, and they use pre-programmed skills and knowledge to complete tasks. It's the digital equivalent of having a team of robot assistants that mow your grass and deliver your groceries.

Companies aim to have these agentic systems handle complex work, not just consumer use cases -- for example, manage supply chains, create marketing campaigns, or improve customer service by understanding and responding to people's needs in a more personalized way. While these systems will "sneak" into our lives as consumer affordances, making our lives easier, they will ultimately become embedded in our business world as well. How we decide to build, deploy, and manage these systems will result in massive changes in what work humans do. This working future will require that we refocus human effort on critical thinking, creativity, and interpersonal relationships.