20% of workers say they use AI for tasks that used to be given to colleagues, poll finds

TEXT START: A new poll has found that almost half of employed U.S. adults use AI for work, with over 20% of those using the tool for work every day the poll was conducted.

The Dissection

This is a normalization report disguised as a workforce update. Its headline contains the decisive fact: AI is taking work that previously justified hiring a colleague or contractor. The article then repeatedly narrows the meaning to “tasks,” “efficiency,” and “dividing work,” making labor substitution sound like collaboration.

The text records the first stage of obsolescence while treating the absence of whole-job automation as reassuring. Firms are learning how many human labor nodes can be removed before eliminating the job title itself.

The Core Fallacy

The article treats the disappearance of an entire job as the meaningful threshold. Under Discontinuity Thesis mechanics, a job is a bundle of economically purchased tasks. If AI absorbs enough of that bundle, employers need fewer workers, contractors lose demand, entry-level pathways collapse, and remaining workers cover more output with less bargaining power.

“No whole-job automation yet” is not a defense. It is the normal early mechanism of replacement. The distinction is semantic, not economic.

Hidden Assumptions

• Productivity gains will create enough new human work to offset displaced labor.
• Firms will preserve headcount rather than convert AI-enabled efficiency into labor cuts.
• Human oversight will remain scarce, valuable, and difficult to automate.
• Task-based contractors can reorganize into broader, defensible roles.
• Making work more “meaningful” will make it economically necessary.
• Worker testimony can shape the transition, despite competitive pressure pushing firms toward cheaper production.
• AI use will remain an optional productivity tool rather than become a cost-of-entry requirement.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The article is correct that task delegation does not automatically prove that an entire job has vanished. It is also useful that the poll measures worker behavior rather than relying solely on executive forecasts. But the framing makes active substitution psychologically acceptable before its consequences become politically unavoidable.

“Workers and AI are dividing work” sounds cooperative. Structurally, it can mean one worker and one model replacing several workers. The worker-voice framing humanizes the transition; it does not change the competitive constraint.

The Verdict

This is early evidence for P1 and the mechanism feeding P3. The supplied poll does not prove full cognitive automation dominance, coordination impossibility, or majority productive-participation collapse because it provides no data on wages, hiring, hours, layoffs, or output.

It does, however, destroy the comforting boundary between assistance and replacement. AI is already receiving tasks once assigned to other humans. That is labor substitution at the smallest economically useful unit. The job title can remain intact while the labor market beneath it is hollowed out. Under DT logic, this is not a workflow tweak. It is the opening cut in the wage-to-consumption circuit.