Dario Amodei of Anthropic and Mustafa Suleyman of Microsoft AI warn that AI could automate up to half of entry-level white-collar roles—finance, consulting, law, tech—within a few years. Alex Imas argues that as AI cheapens cognitive labor, spending shifts toward services where human touch matters: teaching, care, hospitality, bespoke crafts. But that leaves millions of traditional office workers in limbo. The article’s authors counter that many jobs survive not because individual tasks can’t be automated but because firms buy bundles of tasks—and those bundles often resist being split apart cheaply. A weak bundle dissolves when AI handles isolated tasks—travel agents lost 60% of jobs once online booking went mainstream, even though surviving agents now command fees on high-end planning. In contrast, accountants remain in demand: rote bookkeeping clerks face a 6% decline over the next decade, while accountants and auditors grow by 5%, with median pay near $82,000. Three factors strengthen a bundle: unpredictable demand (customer service calls veer off script), production spillovers (a radiologist who sees the patient reads scans better), and measurement friction (who’s liable when AI drafts a report and a human signs it?). Beyond task bundles, organizations rely on managers to resolve internal conflicts and allocate resources—marketing budgets, engineering time, legal safeguards—roles that aren’t mere information processing. AI tools can boost productivity inside these bundles but not replace the human who balances competing goals, interprets tacit information, or bears responsibility when outcomes go wrong. That structure keeps many mid-level and professional roles intact even as basic tasks fall to algorithms.