In the middle of a technological shift, people tend to get stuck in tunnel vision — heads down, focused on whatever’s directly in front of them. It helps, every so often, to step back and get a broader read on the lay of the land.
Event Context
In 1969, the management thinker Peter Drucker, in The Age of Discontinuity, argued that the coming wave of technology would break apart the work structures everyone assumed were permanent. Two years later, Intel released the 4004 — the first commercially available microprocessor, a full central processing unit shrunk onto a single programmable chip.
Before it, computing power lived in the back rooms of large corporations and research labs. After it, computing got small enough to put on a desk, and eventually in a pocket.
Semiconductors did their part. So did software — UNIX, the C programming language — and together they built the scaffolding for what we came to call the knowledge economy.
Manual, repetitive industrial labor gave way to work that ran on judgment and information. A new class of worker emerged, one who effectively owned the means of production in their own head rather than on a factory floor.
The numbers back up how big that shift was. Global GDP grew from roughly $1.3 trillion in 1960 to about $33.5 trillion by 2000, then nearly doubled again to around $66.8 trillion within another decade, as trade barriers fell and emerging markets came online. Knowledge work didn’t just survive that expansion — it drove it.
Now we’re in the middle of another discontinuity, and once again it’s easy to miss while everyone’s arguing about the latest model release.
Knowledge work always had two parts: figuring out what the task actually was, and then doing it. A factory worker was handed a task. A knowledge worker had to work out what the right task even was before touching it.
That’s where the line is being redrawn. Agents are increasingly good at the second half — executing a well-defined task. They’re much shakier on complicated, ambiguous asks, and they’re genuinely bad at recognising when the task itself is the wrong one to be doing.
Someone still has to be willing to say “we’re optimising for the wrong thing.” For now, that someone is a person — and that willingness, not raw knowledge, is what won’t be easy to automate.
The agentic economy may reward the opposite. When an agent can already execute a well-scoped task inside a domain, the human’s value shifts toward defining the task well and checking the output — catching what’s wrong before it ships. And people are still hesitant to take AI output at face value; a lack of trust in reliability remains one of the biggest reasons people won’t treat an AI-generated answer as final.
The agentic economy won’t mean fewer decisions. It’ll mean more of them — faster, higher-stakes, more tangled together — than any person or organisation has had to navigate before.
The knowledge workers who come out ahead won’t be the ones who know the most. They’ll be the ones with the judgment to ask the right question, the skepticism to distrust a suspiciously confident answer, and the nerve to stop an agent mid-task when something looks off.
Player Focus
Agentic AI systems can pursue goals, sequence decisions, and operate over stretches of time without a human checking in at every step — work that, until recently, only people could do. These systems don’t have infinite capacity, but they can move through enormous volumes of information fast enough that the old premise of the knowledge economy — that scarce mental bandwidth is the valuable resource — no longer holds the way it used to.
So the real question isn’t whether AI is impressive. It’s what happens to the knowledge worker once knowledge itself stops being scarce.
Team Analysis
Drucker warned about something that reads as almost prophetic now: the knowledge worker’s biggest risk wasn’t ignorance, it was becoming a prisoner of their own speciality — brilliant in a narrow lane, lost outside it. The knowledge economy rewarded that narrowness, because depth was what execution required.
Match Outlook
That means the next class of essential worker won’t be defined just by the depth of their specialty, but by how well they can set context across domains, get a a grasp of the technology, the people involved, the institutional politics, and the ethics at stake to know which question is even worth handing to an agent in the first place.
That’s an uncomfortable finding for anyone who has spent a career building a narrow technical edge. But it fits Drucker’s deeper point about discontinuities: they don’t shrink the need for human judgment. They move it somewhere else.
We may need a new name for that kind of worker. But we’re going to need them more than ever.
