Productivity Without Paychecks
Yesterday’s most consequential argument didn’t come wrapped in a glossy report or a demo reel. It slipped in through the Letters page of the Financial Times and quietly rearranged the furniture in the room. The writer took aim at the soft consensus that AI-led productivity will do the heavy lifting for growth and debt, and asked a more uncomfortable question: what happens when the productivity arrives but the payrolls don’t?
That framing matters. We’re used to adding up job gains and losses like a scoreboard, debating net effects while the economy’s plumbing hums out of view. The letter pulled us under the floorboards, back to the transmission mechanism that turned earlier productivity waves into broad prosperity. The machinery was deceptively simple: firms got more efficient, but crucially hired armies of people to run and expand the new systems. Wages flowed, households spent, and demand validated the supply. Productivity didn’t just raise potential; it financed its own realization through mass paid work.
The Missing Transmission
This time is different in a way that isn’t techno-theatrical but macro-mechanical. Automation, IT, robotics, and now AI are absorbing tasks in precisely the places where modern economies stash career on-ramps. Services were the shock absorber of the last 30 years—call centers, back offices, retail floors, junior analysts, entry-level creatives. Those roles didn’t just pay rent; they trained workers, created ladders, and kept household demand wide-based. When AI compresses those rungs into a handful of specialized supervisors and a lattice of bots, it doesn’t merely lower the wage bill at a firm. It narrows the aperture through which productivity diffuses into the real economy.
Strip out breadth of employment and you strip out the income mass that makes demand resilient. You can then have measured productivity rising—more output per hour for the people still on payroll—while actual and potential GDP sag because fewer hours are paid at all. The paradox is uncomfortable but coherent: an economy can become technically better at producing while becoming worse at circulating the gains.
Why Services Make This Different
In manufacturing-led booms, adding capacity usually required physical expansion and lots of workers. In services, the cost structure is more intangible and more amenable to substitution. A model that drafts the brief, edits the code stub, negotiates the shipment, and triages the claim can take out not one job but a chain of training experiences that used to justify a career ladder. The organization chart hollows out, the middle thins, and the surviving roles become narrower and more specialized. That concentration is toxic for demand because it concentrates income. Households don’t buy groceries with productivity statistics; they buy them with wages that arrive consistently across a broad base.
The Macro Math That Scares Bond Desks
Here’s the growth arithmetic the letter forces back onto the table. If AI is a powerful supply shock that flattens the firm-level cost curve, it should, in theory, be disinflationary and boost potential output. But if it simultaneously erodes paid hours and compresses the wage share, aggregate demand underperforms. Central banks can’t fix a distributional shortfall with basis points, and fiscal policy that relies on buoyant tax receipts from mass employment suddenly meets a weaker revenue base just as societies face aging and climate costs. You solve for productivity and accidentally unsolve for solvency.
There’s a second-order effect that CFOs already sense in their dashboards. Efficiency raises margins, then softens volumes when customers’ incomes stagnate. The spreadsheet looks glorious for a quarter or two. Lifetime value assumptions, however, don’t survive a customer base that’s shedding hours. The economy is still an ecosystem; drain the wage pond and even the apex predators go hungry.
If You Build Productivity and No One Can Buy It
The letter’s novelty isn’t that it predicts job losses or claims job gains can’t happen. It rotates the lens from counting positions to evaluating the employment-to-demand loop. Will this wave preserve the very channel through which earlier revolutions paid for themselves? Or are we about to discover that productivity without paychecks is like horsepower without fuel lines—the machine revs in place, impressive and stuck?
Designing Circulation, Not Just Innovation
If that diagnosis holds, the policy conversation changes tone. Retraining still matters, but as an alone strategy it misfires when the rungs themselves are gone. The design problem becomes creating institutions that keep earned income broad and recurrent while adoption proceeds. That may mean binding automation to wage-linked diffusion through bargaining frameworks and procurement rules, sharing gains through profit-sharing or data dividends, tilting tax and subsidy toward labor-absorbing sectors like care and climate adaptation, and shortening workweeks while protecting pay to spread hours without shrinking incomes. The point isn’t to freeze technology; it’s to engineer circulation so the new productivity finds wallets, not just earnings per share.
Signals to Watch in 2026
If this thesis is right, you won’t confirm it in headcount releases alone. You’ll see it in the employment-to-population ratio plateauing despite strong efficiency metrics, in the wage share’s reluctance to rise even as profits do, in the collapse of apprenticeships and junior roles that once served as feeders for middle-income work. You’ll see services prices cooling while volumes disappoint. You’ll see governments quietly revising growth and revenue assumptions even as firms brag about cost per task. These are not vibes; they’re diagnostics of a broken transmission.
The Wager in Front of Us
Yesterday’s letter deserves more than a polite nod because it reframes the stakes. The optimistic claim that “AI will lift growth” is incomplete without an answer to “through which paychecks?” Without that answer, we risk mistaking a redistribution from wages to margins for progress. With it, we can take AI’s very real efficiency and wire it into the economy we actually live in—one where prosperity is not an abstract ratio but the sum of households able to spend, save, and plan. Productivity is the engine. Employment is the drive shaft. Remove it, and the wheels don’t turn.
