Image Credit & Source: UNLEASH World 2025
The loudest signal from Susskind’s recent keynote: stop framing AI’s impact as a binary of “jobs destroyed vs jobs created.” Look at tasks. Every role is a bundle of activities; AI displaces some tasks, amplifies others, and creates new ones. That is where the real shifts and opportunities sit.
What humans keep doing
Susskind’s latest paper argues there are three durable limits that keep meaningful human work in the loop:
- General equilibrium limits, even if machines are better at everything on paper, comparative advantage can still make it efficient for people to do specific tasks.
- Preference limits customers sometimes simply prefer a human touch.
- Moral limits society will insist on human oversight for tasks with normative stakes. These boundaries define tomorrow’s human task set.
The white collar wake up call
We have long assumed creativity, judgment, and empathy insulated office work. Generative AI has punctured that myth by tackling such outputs differently, not by “thinking like us,” but by handling uncertainty with scale and speed. The practical move for leaders: re-specify roles as task portfolios, then decide which tasks are automated, which are augmented, and which remain resolutely human.
Strategy: from job descriptions to task architectures
- Map the task graph. Decompose priority roles into tasks; label each as automate, augment, or assign to humans. Rebuild roles around the augmented core.
- Invest in targeted training. Near-term demand rises for tasks not yet automatable. Upskilling should be sequenced against that map, not generic catalogs.
- Codify human in the loop. Where moral or preference limits apply (credit denials, clinical triage, hiring adjudication), write clear human oversight standards and audit them.
Metrics that matter
Retire “jobs saved” dashboards. Track:
- Task half-life: time before an augmented task becomes automatable.
- Augmentation lift: output and quality gains when humans and AI pair on a task family.
- Human value density: share of a role’s time spent on tasks within the three limits above.
The takeaway
Susskind’s point is not comfort; it is direction. The next decade is not a cliff of mass unemployment so much as a reconfiguration of what people do at work. Leaders who move from roles to tasks, and fund training, governance, and redesign accordingly, will not just protect jobs. They will build better ones.
