There is a long, quiet trend underneath the last seventy years of work: the unit keeps shrinking. Our grandparents had careers — one institution, one arc, one pension at the end. Their children had jobs — portable, swappable, defined by a title rather than a lifetime. We have roles and gigs, defined less by where we sit than by what we are responsible for this quarter.
Now the unit is shrinking again, below the job, below the role, down to the task. Software has spent two decades unbundling the job into discrete, assignable, increasingly automatable pieces. AI is the accelerant: it can now do the tasks at the bottom of the stack faster and cheaper than scheduling a human to do them.
Decomposition is not the same as disappearance
It is tempting to read this as a story of loss — the dignified career dissolving into a feed of micro-tasks. But decomposition is not destruction. When you break a job into its tasks, you discover that they are not all equal. Most are routine. A few are decisive. The routine ones flow to automation. The decisive ones become more concentrated, more valuable, and more clearly attributable to the specific person who can do them.
When work is unbundled, value does not vanish. It pools in the few tasks that resist being unbundled — the ones that require judgment, context, and accountability.
Three layers fall out of the decomposition
- Commodity tasks — searching, summarizing, formatting, first-pass drafting. These are already migrating to machines, and that migration will not reverse.
- Coordinated tasks — assembling, routing, and checking the commodity work. This is where most software is being built right now.
- Judgment tasks — deciding what is actually true, what to do about it, and who is accountable when it is wrong. This is the layer that does not decompose cleanly, because it depends on lived experience.
The mistake is to assume your worth lives in the first two layers because that is where your hours currently go. The hours are not the value. As the bottom layers automate, the only durable position is to be the person the judgment task routes to.
The task economy needs a market for judgment
If work is becoming a stream of tasks, then the missing infrastructure is a way to summon judgment on demand — to pull a specific person into a specific decision for exactly as long as the decision requires, and no longer. Not a 40-hour hire. Not a 6-week consulting engagement. A 20-minute call, precisely matched to the gap, precisely when the gap appears.
That is the logical endpoint of the trajectory. The career compressed into a job, the job into a role, the role into tasks — and the most valuable task, judgment, gets its own liquid market. The expert stops being an employee or even a contractor. They become a node you can reach the instant the work demands them.
The unit of work got smaller. The value of being genuinely good at the irreducible part got larger. Both things are true, and they are the same trend.