AI agents—autonomous systems that can pursue open-ended goals and complete tasks with minimal human input—are no longer science fiction. They’re quickly becoming a reality that could reshape how we work and do business.
While it’s still early, AI agents are showing increasing promise as major AI labs and ambitious startups begin to deploy them in areas like customer service and coding. And they have big implications for the work and the functions performed by human professionals.
What does that mean for your work, your business, and your industry?
I got the answers from Marketing AI Institute founder/CEO Paul Roetzer on Episode 105 of The Artificial Intelligence Show.
The AI agent explosion is coming—but not quite yet
While AI agents are garnering significant attention, Roetzer cautions that we’re still in the early stages.
In Episode 87 of The Artificial Intelligence Show, Roetzer outlined his anticipated timeline for AI innovation, and agents played a start role.
“What I said then was 2025 to 2027 is likely when you would see AI agents really become viable and change the way work was done,” he said.
For now, we’re seeing what Roetzer calls “GPT-1, GPT-2 level” agents—early versions that are unreliable and require significant human oversight.
They’re nowhere close to full autonomy yet. But they’re getting better—fast.
Where AI agents are showing promise
Despite being in their infancy, AI agents are already demonstrating potential in several areas.
Venture capitalist Rob Toews, in a recent Forbes article, highlighted some promising use cases:
- Customer support: Companies like Klarna are using AI agents to automate customer service, handling millions of conversations.
- Regulatory compliance: Startups like Norm AI and Greenlight AI are developing agents to handle structured, repeatable compliance workflows.
- Data science: Companies like Delfina are building agents to manage the full data science lifecycle.
- Personal assistants: Startups such as Mindy are creating AI-powered personal assistants to help with daily tasks.
The major AI labs at Google, OpenAI, and Microsoft are also all working on AI that displays agentic behavior.
The long game: World of Bits
To understand where AI agents are headed, Roetzer points to a concept called “World of Bits,” first explored in a 2017 research paper by famed AI researcher Andrej Karpathy.
“In 2017, they had the vision to give AI the ability to take action. But the technological capabilities weren’t there,” Roetzer explains.
The idea was to give AI access to a computer’s keyboard and mouse, allowing it to interact with digital interfaces just like humans do. While the concept was ahead of its time in 2017, recent advancements in language models have made it newly relevant.
Today, that idea is much closer to reality.
“This is what everyone’s been working on,” says Roetzer. “They know humans doing cognitive tasks spend billions of hours collectively doing them every day. And if they could build agents that do those tasks for humans, it will unlock all kinds of possibilities. So this is what they are all chasing.”
Roetzer fully acknowledges that, right now, agents are imprecise and error-prone. They have limited memories, lack common sense, and are largely black boxes. However, the assumption leading AI researchers are making, says Roetzer, “is that all of this is solvable.”
“And once they solve these things, then an intelligence explosion occurs.”
When that happens, businesses—and society as a whole—need to be ready, because AI agents that can perform cognitive work will have a massive impact on the economy.
“If economists and business leaders don’t start thinking about the possibilities now, then it’s going to get ugly in 18 or 24 months when reality hits,” Roetzer warns.