AI doesn't usually kill the task itself. It kills the premium the task used to command. The work still gets done - by someone cheaper, or by software, for free.
Most "AI threat" conversations focus on whether AI can do the task at all. Often that's the wrong threshold. The real question is whether customers will still pay extra for it once they have an alternative.
Drafting a memo is still done. Nobody pays the premium for the draft anymore. Once that premium goes, the role's economic logic goes with it.
Every task has three kinds of value. Intrinsic value - what it means to the person doing it. Economic value - what someone will pay for it. Contextual value - how essential it is within a specific workflow. AI affects these differently.
Wordsmithing a thoughtful email has high intrinsic value to the writer. Its economic value is zero - nobody pays extra for it. Issuing a building occupancy certificate is intrinsically trivial. Its economic and contextual value are high - only a licensed official can do it, and projects can't proceed without it.
When AI eats a task, intrinsic value often survives (you can still enjoy doing it). Economic value collapses first (nobody pays the premium). Contextual value can persist if the task remains gating to a workflow - but only until someone restructures the workflow to remove the gate.
A scarce skill that became automation, then a fond memory.
Through the early 20th century, ships at sea were tethered to the world by one role: the radio operator. Trained in Morse code, they could call for rescue, signal distress, coordinate across thousands of nautical miles. The scarcity of the skill plus the stakes of failure made the role highly paid and respected.
Communications infrastructure improved. Satellite, then GPS, then automated transponders. By the 1990s, the radio operator role was gone. The task of ship-to-shore communication never stopped - it became continuous and automatic. The premium collapsed because the scarcity that had earned it had gone.
The pattern repeats across knowledge work right now. Translation, legal research, copywriting, basic analysis - each one moving from "scarce skill worth paying for" to "free baseline capability included with the product."
The task didn't disappear. The price did. That's what "economic collapse of the task" means - and it's the leading indicator of role obsolescence.
Routine translation work - manuals, marketing copy, subtitles - used to bill hourly at meaningful rates. The task hasn't disappeared. What's collapsed is the price. Customers expect translation to be free with the authoring tool, paid only when accuracy is critical enough to need certification.
Translators who survive command premium for one of two things - certification (legal, medical) or literary craft. Everything in between has collapsed to free.
Logos, social posts, marketing templates - the task is still being done, often at higher volume than ever. The economic value to the customer has collapsed to "free with my Canva subscription" or "two prompts in Midjourney."
Premium has migrated to taste, brand systems, and the ability to translate a business vision into a visual identity executives trust. The middle layer of designer work - once where most of the income lived - is gone.
Pick a role. Drag the AI-capability slider. Watch tasks migrate from "still earns a premium" into the no-premium buckets below. What's left at the top is what the role's income comes from.
Pick whichever is closest to your work. The mechanism is the same for any knowledge role.
The wrong question is "what tasks does AI do." The right question is "what work still commands a premium five years from now."
An advisory engagement that redesigns the operating model around AI as a coordination capability - identifying which work is becoming a commodity, which is becoming the new premium, and how to re-architect roles and pricing around the new value distribution.