The three rebundle positions.
Tasks get absorbed by AI. Positions don't. The rebundled role stabilizes around three structural positions AI cannot occupy - liability, customer + IP ownership, and authority to ratify.
From tasks to positions.
Most defenses of a profession's future treat the role as a list of tasks and ask which the model can't yet do. That defense loses ground every year. Better models close the task list one item at a time.
A durable rebundle isn't built around which tasks the human keeps. It's built around which structural positions only a human can occupy. A position is a seat in a value chain or institutional system - defined by legal personhood, commercial ownership, or conferred authority. None of those reduce to skill. All three are durable against better models because the binder is non-technical.
Risk owner - new accountability infrastructure.
The role bears institutional, legal, or professional responsibility when AI deployment drives a regulated outcome. The position exists because someone has to be liable, and AI has no legal personhood.
The distinction from pre-existing risk-bound roles (court interpreters, IMF translators) is that this position is built around accountability infrastructure that did not exist pre-AI. The risk it owns was created by AI deployment.
Why AI cannot occupy it: liability requires a human or institutional entity. AI cannot be sued, sworn in, censured by a professional body, or named on a regulatory filing.
Translator example: Ariel Koren's Respond Crisis Translation - the institutional accountability layer for MT deployment in asylum contexts, an infrastructure that scaled because MT deployment scaled.
Customer + IP owner.
The role owns the offering itself - the buyer relationship, the workflow, and the intellectual property. AI sits inside the offering as rented infrastructure; the practitioner sells the offering, not the AI.
Why AI cannot occupy it: AI cannot sign contracts, hold a brand, or own a customer relationship. Commercial ownership is a structural seat. Better models don't compound into it.
Translator example: Gabriel Fairman, founder and CEO of Bureau Works - a translator-founded translation platform where AI capabilities are embedded under the founder's brand and customer relationships. Also: Corinne McKay and Damián Santilli's productized educator businesses, where AI is subject matter inside a course product the practitioner owns.
Constraint-definition authority.
The role holds the social, professional, or institutional authority to ratify what counts as faithful, safe, or acceptable in a domain - and to refuse AI deployments that fall short.
Why AI cannot occupy it: authority is relational and conferred. It depends on the standing the role holds in an institutional / professional ecosystem. AI can suggest definitions and draft style guides; it cannot ratify them in a recognized institutional context.
Translator examples: Ariel Koren (asylum-MT accountability authority); FIT - the International Federation of Translators - and its 2025–2028 Council under President Guillaume Deneufbourg; ATA's President Andy Benzo. Each carries the professional / institutional authority to set field-level positions on AI translation.
How a rebundle qualifies.
A practitioner's rebundle qualifies as durable to the extent it occupies at least one of these three positions, and bundles AI as ingredient inside the offering (the leverage test). A position without AI as ingredient is durable but parallel to the algorithm; a position plus AI as ingredient is durable above the algorithm.
The pay differential is the empirical fingerprint. Positions earn 3–5× the Human-in-the-loop gap-filler floor. The premium isn't compensation for rare skill (skill collapses); it's compensation for the structural seat AI cannot occupy.
The translator synthesis.
In the translator case study, the Rebundled Role section maps the five named genuine-rebundle practitioners across the three positions. The pay-differential marquee inside that section shows the empirical signature: $5–$18/hour Human-in-the-loop floor vs $40+/hour positional work, with ATA-certified premium adding 21% on top.