The Sunday When “Less Work” Died
On a quiet Sunday, a column cut across a month of layoff headlines like a cold front. Joe McKendrick’s piece didn’t just argue that AI won’t erase jobs; it suggested something more uncomfortable and more plausible for anyone inside a transforming company: we are on the verge of having too much to do. Days earlier, Jeff Bezos sat under studio lights and said the quiet part out loud—that the productivity surge from AI could pull the economy toward labor shortages and even deflation. The idea landed with the weight of a market signal. If the world’s most operationally minded billionaire thinks the constraint will be people, not tasks, you pay attention.
McKendrick’s framing is simple but not simplistic: AI shaves off the predictable slices of expert work, and what remains doesn’t shrink the job; it expands the orbit around it. When the easy parts move to machines, the center of gravity shifts to activities that are elastic rather than finite—defining the ambiguous, supervising the unreliable, handling the weird edge cases, and carrying the responsibility when everything rolls up to a signature. Each sliver of automation spawns an entourage of coordination: data curation, workflow design, quality assurance, compliance, user support, and the continuous tuning that living systems demand.
When Tasks Vanish, Coordination Multiplies
We have seen this movie before in other technologies. Remove the bottleneck in one place and you expose three new ones you had previously ignored. A sales team armed with AI assistants pushes more proposals, but that raises the bar for pricing governance, contract review, and customer success. A claims department gets straight-through processing; suddenly exception queues, appeal handling, and audit trails swell into full-time work. The paradox is structural: because AI is best at the narrow thing, it forces humans to specialize in the broad things—context, integration, and accountability.
This is not romanticism about human creativity. It’s an operations argument. Complex systems fail in complex ways. When you stitch models into products and products into businesses, you expand the surface area where reality can deviate from plan. That surface demands stewards. Some of those stewards will be domain experts retooled into workflow architects; others will be model risk managers, data reliability engineers, and line leaders who learn to ask adversarial questions of their tools. Their work is not the old work, but it is very much work.
The Macro Twist: Productivity That Hunts for People
Bezos’s deflation riff is a tell. If AI meaningfully reduces the unit cost of producing knowledge, organizations do not stop at their current output; they increase volume and pursue new lines they previously shelved as uneconomical. Demand is not a fixed pie. It expands to meet the lower cost of experimentation. But the constraints shift to skills and oversight, which don’t scale with a prompt. The result is a paradoxical labor market: fewer minutes on the rote tasks, more headcount pressure in the functions that channel, evaluate, and authorize the work AI makes possible.
That tension will show up unevenly. Within roles, the middle collapses. Steady, predictable work migrates into systems; judgment work aggregates into fewer hands, but those hands touch more situations. Wage dispersion widens inside occupations. Meanwhile, entire layers that sit between code and consequence grow: safety sign-offs, model governance, legal review, customer trust operations. Regulation won’t create this layer; risk will. Regulators will simply formalize what insurers and boards already demand—traceability and a human to answer the question, “Who approved this?”
The Full-Automation Mirage
McKendrick’s warning to leaders is pragmatic: if you chase total automation, you will build brittle operations. The returns from removing the last human degrade quickly, because the brittle edges of reality will cost you more than the salary you “saved.” A model that nails 95% of cases is a marvel; the remaining 5% is where revenue, safety, and brand live. That last mile is not a footpath; it’s terrain that shifts with markets, regulation, and customer behavior. Human-in-the-loop design isn’t a concession to nostalgia. It’s an operating principle for systems whose failure modes are social as much as technical.
Consider a mid-size insurer that automates initial claim triage. Cycle time drops. So does cost. But denials now require stronger explanations, escalation paths multiply, and plaintiffs’ attorneys probe the seams where automation meets judgment. The company doesn’t replace adjusters; it retrains them and hires model auditors and documentation specialists. Net headcount in the chain goes up, not down—because output went up and because accountability thickened.
What Smart Employers Will Actually Do
The implication is not to freeze hiring; it is to redesign work. Treat models like junior colleagues: fast, tireless, inconsistent, and in need of structure. Rewrite roles around orchestration—clear problem statements, robust interfaces, exception playbooks, measurable quality targets, and explicit decision rights. Upskill for supervision: how to test and probe systems, how to explain outcomes, how to intervene when the world looks unlike last quarter’s data. The firms that move first will compound their advantage, because organizational learning about when and how to trust machines is path-dependent. You cannot buy it off the shelf.
For workers, the shift is equally stark. The safe harbor isn’t “creative work” in the abstract; it’s accountable work—owning the definition of done, the fitness of the metric, the ethics of the edge case, and the moment you say no. The résumé line that matters will be less “built with AI” and more “operated with AI at scale without breaking.” That is not a vibe; it is a craft.
The New Default: Busier
McKendrick’s column doesn’t deny displacement. It reframes the baseline. Yes, some roles thin out. But the giant, practical fact emerging from boardrooms and broadcast studios alike is that AI expands the feasible frontier of projects. More products launched. More experiments run. More customers engaged. More logs to review, more models to retrain, more exceptions to adjudicate, more responsibility to shoulder. In other words: more human work, not less.
If you’ve been bracing for a vacuum, adjust. The coming scarcity is not jobs. It’s judgment.