Dominic Jainy is a seasoned IT professional whose expertise spans the critical intersections of artificial intelligence, machine learning, and blockchain technology. With a career dedicated to understanding how nascent technologies transition from theoretical models to practical industry applications, he has become a vital voice in evaluating the rapid shifts in the digital landscape. In this discussion, we explore the transition from the era of simple chatbots to the “Agentic Age,” where AI functions as an autonomous, self-correcting force. We delve into the concept of the “jagged frontier” of capability, the psychological strain of exponential growth on society, and the enduring value of human touch in an increasingly automated world.
How does the shift from interactive chatbots to autonomous, self-correcting agents change the daily workflow for experts?
The transformation we are witnessing is a fundamental move from being a passenger to becoming a conductor of complex digital systems. In the earlier stages of this technology, such as the hot summer of July when many first began experimenting, the dominant method was using AI as a co-intelligence where you had to prompt, check, and prompt again. Now, we are entering the era of the “harness,” a term used to describe the infrastructure that boosts model capability and autonomy, allowing AI to run longer and more complex tasks without constant human hand-holding. For an expert, this means your daily routine shifts from manual micro-management to a high-level managerial role where you oversee self-correcting systems that can lurch forward on their own. It is no longer about filling in small gaps with a chatbot; it is about deploying an agent that can navigate a “harbor game” of variables and produce valuable work while you focus on strategic direction.
Mollick mentions a “jagged frontier” where AI excels in some areas but fails in others. How should professionals navigate this uneven and often unpredictable landscape?
Navigating the jagged frontier requires a high degree of discernment because the technology often lunges ahead in one domain while remaining surprisingly stagnant in another. We see this in the benchmarking tools like METR, which is a standard used by AI safety research organizations to measure real-world capabilities and risks to ensure models are deployed responsibly. A professional might find that an AI can recall ten top ideas from a decade of blog posts with perfect accuracy, yet it might struggle with a task that requires simple physical intuition or nuanced social grace. To handle this, you must treat the AI as a powerful but uneven tool, constantly testing its boundaries rather than assuming its competence is uniform across all tasks. It’s like watching a dramatic orange sky at sunset; it looks spectacular and broad, but the light doesn’t reach every corner, and you have to know where the shadows are to avoid being caught off guard.
We often hear that humans are “bad at feeling exponentials.” Why does the steady doubling of AI capability feel like a series of sudden shocks rather than a smooth curve?
Even though the progress of these models can be plotted as a smooth, exponential curve on a graph, our human experience is defined by “turbulence” and sudden jolts that disturb our sense of normalcy. We saw this on June 30 when new updates dropped, reminding us that while the doubling of capability is constant, the impact on our institutions feels like a series of improvised policy changes at the highest levels of government. One day, a technology is not a cybersecurity threat, and then suddenly, with a single leap in capability, it undermines an entire business model and leads to massive swings in the stock market. We are effectively the frog in the water, but the water isn’t just warming—it is reaching a boiling point through a series of “shocks” that we are psychologically unprepared to process as a steady trend. This leads to a profound sense of angst, as people feel the “exponential” is outrunning their ability to adapt to the tool they only just learned to use last quarter.
There is a growing anxiety among the younger generation and traditional workers regarding obsolescence. How can we address the gap between what a tool can do today and what a person learned just last quarter?
The gap between current tool capability and past training is the new form of obsolescence, and it is leaving many teens and professionals feeling incredibly “glum” about the future. This angst is real because the technology doesn’t just outrun our institutions; it outruns the very people who are trying their hardest to keep up with the latest open-weight and closed-weight systems. To address this, we have to move the conversation beyond the “brain damage” of simplistic claims that AI either destroys or enhances cognition and instead focus on the “harness” of human expertise. We need to accept that the “real work” of the future involves managing these agents rather than competing with them on tasks they can now do exponentially faster. If you spent last quarter learning to prompt a chatbot, this quarter you must learn to manage an agentic system, or you will find yourself on the wrong side of the jagged frontier.
While AI is mastering coding and data recall, some argue it lacks the human touch for tasks like gardening or nursing. Where do you see the boundary between digital efficiency and physical reality?
There is a powerful argument, often voiced by those who value “mammalian nurture,” that real work involves physical acts like planting, pulling weeds, or breastfeeding a child—things a robot simply cannot “enjoy” or “taste.” While AI is excellent at “recall” and can find 15 times to use AI and 5 times not to from a December 2024 archive, it cannot yet wipe up dog piddle or put leftover cake in the fridge with genuine domestic care. However, we should be cautious about where we draw that line, as tasks like “pulling weeds” or “attending committee meetings” are already being targeted by agentic AI and robotics. The boundary currently lies in the emotional and sensory experience—the ability to actually “enjoy the cake” rather than just analyzing its ingredients—but as robots begin to take over more physical chores, the definition of “real work” will continue to shrink toward the purely biological and emotional.
What is your forecast for the Agentic Age?
My forecast for the remainder of 2024 and heading into July 2025 is that we will see a decisive shift away from “chatting” with computers and toward “tasking” autonomous agents. We will see the sunset of the simple chatbot as models become more integrated into “harnesses” that allow them to perform long-term, self-correcting projects without human intervention. This will lead to a period of intense institutional “shocks” as markets and governments realize that the “jagged frontier” has moved much further than they anticipated, particularly in areas like coding and cybersecurity. Ultimately, the successful professionals will be those who stop trying to outwork the AI and instead learn to be the managers of a digital workforce, focusing on the tasks that require a human heart while letting the agents handle the lurching complexity of the exponential curve.