Entropy Rising

Why unchecked complexity and a vanishing talent reservoir are stalling America’s longpromised productivity revival The first time I met Emilio Torres he was leaning into the guts of a thirty-year-old blow-molder, one hand resting on the side panel, the other tracing invisible layers of logic that lived somewhere deep inside the relay ladder. He spoke of the machine the way a vintner talks about soil, every nuance mattered, every vibration carried a signal. That intimacy with equipment once defined North American manufacturing culture. Today Emilio is an outlier: a sixty-seven-yearold who delayed retirement because “no one left in maintenance speaks this machine’s language.” When he finally hangs up his tool belt, the plant manager admits they will either decommission the line or replace it outright. It is cheaper to buy a new asset than to replicate Emilio’s tacit model of how pressure, polymer viscosity, and ambient humidity conspire to create perfect pharmaceutical bottles. The paradox is glaring: capital is abundant, but the expertise that turns capital into throughput is evaporating.

That tension, between rising technological possibility and shrinking operational mastery, lies at the heart of the productivity puzzle visible in the federal data. Labor output in U.S. factories climbed briskly from the late 1980s until roughly the eve of the Great Recession, then leveled off and has since hugged a flat plateau, never recapturing its pre-crisis momentum Liberty Street Economics. Popular commentary blames everything from globalization to cheap capital, but the root driver hides in plain sight: complexity accumulated faster than human capacity could absorb it. By mispricing the cost of intricacy, and by allowing accountants to overrule the systems intuition of engineers, companies engineered their own stall. To understand why, one must rewind to the post-OPEC modernization wave. Plants originally built for single-product economies were asked to flex: shorter runs, greater mix, tighter tolerances. The easiest answer was incremental automation. A retrofitted servo here, a vision system there, a batch of upgraded PLC I/O to accommodate new fail-safe rules. Each microproject cleared the hurdle of payback spreadsheets, yet no one tallied the compounding cognitive strain. What began as elegant engineering mutated into what system theorists call “detail complexity”: innumerable pieces whose interactions defy intuitive control. Scholarly work confirms the penalty, variety-induced complexity raises unit cost, erodes reliability, and saps quality consistency ScienceDirect. By the 2000s a second layer surfaced: informational bloat. Cheap sensors and even cheaper storage encouraged companies to collect everything. Dashboards multiplied, each championed by a functional silo, production, maintenance, quality, procurement, until a single shift supervisor might juggle forty tiles of KPIs, alerts, and charts. In principle the deluge promised omniscience; in practice it fostered paralysis. Recent reviews in ergonomics and decision science document what floor leaders feel anecdotally each night: past a threshold, additional data slows reaction time and boosts error rates PMCXMPRO. The dashboard became a mirror image of the mechanical sprawl beneath it: elaborate, fragile, and opaque. Still, for a while, complexity paid. The generation that built the machinery could still decipher its quirks; deep-tenure engineers like Emilio acted as living middleware, translating new code into old physics. Then demographics shifted. Baby boomers who had entered apprenticeship programs in the late 1970s began to retire en masse. The Bureau of Labor Statistics now reports

that nearly a quarter of the industrial workforce is older than fifty-five; each year tens of thousands cross the retirement line, taking with them experiential algorithms never captured in any enterprise resource planner AutodeskForbes. Managers who once relied on on-the-job osmosis to train replacements discovered there were no shadow shifts left to carry the lore forward. The red line in our complexity-tenure graphic charts a fifty-year erosion of that buffer, a downdraft that accelerates whenever economic turbulence prompts voluntary exits. Complexity thus metastasized precisely as orchestration capacity withered.

At first the imbalance registered as longer change-over windows, nagging quality escapes, or unexpected micro-stoppages that never made the quarterly deck. Over time it ate into macrometrics: overall equipment effectiveness sagged, scrap budgets ballooned, and energy intensity crept upward even as variable frequency drives proliferated. In short, productivity’s denominator grimly outpaced the numerator. Analysts at the New York Fed calculate that manufacturing totalfactor productivity, which had grown at 1.4 percent annually for two decades, decelerated to near-zero after 2010 Liberty Street Economics. A separate macro assessment notes that factories now produce virtually the same output per hour they did in 2007 despite a torrent of “Industry 4.0” investment Apricitas Economics. The 2010 Breakpoint – Where the Curve Snapped and Stayed Flat Look closely at the FRED labor-productivity series and you will see a single kink—2010. That is the first calendar year in which five structural brakes operated simultaneously, locking the line into its lower trajectory.

1. Capital lungs collapsed. Real equipment spending had fallen more than a third during 2008-09. When the usual two-year replacement cycle should have been pumping fresh capacity into plants, nothing new arrived, so output per worker never regained its earlier thrust. 2. The high-octane sectors left the mix. Domestic production of computers, telecom gear, and advanced electronics—industries that had supplied almost forty percent of pre-2007 productivity gains—shifted offshore. Their absence removed the fastest-growing piston from the national engine. 3. Tribal knowledge evaporated. Early-retirement packages and cost-cutting in 2009 took tens of thousands of veteran technicians off the floor. Plants entered 2010 without the orchestration talent that had previously absorbed complexity shocks. 4. Complexity overshot human bandwidth. A fourteen-year binge of dashboard add-ons and PLC patches reached its tipping-point just as crews thinned. The marginal sensor flipped from asset to liability, burying operators in alarms they could no longer triage. 5. Credit stayed frozen after demand thawed. Orders came back, but risk-averse boards stretched aged machinery rather than approve holistic rebuilds. Deferred maintenance ossified into chronic downtime. In 2009 both the numerator (output) and the denominator (hours worked) collapsed together, so the ratio merely dipped. In 2010 hours remained subdued while output rebounded only halfway, revealing the new normal in a single, unambiguous data point. From that moment on, complexity, capital scarcity, and talent drain forged an iron ceiling the curve has yet to pierce. Yet charts alone cannot convey the lived texture of this entropic turn. Consider an aluminum tube mill in Tennessee that added a laser-based seam tracker to shave microscopic weld variance. The retrofit worked, until seasonal voltage sag induced by a decades-old substation pushed the tracker’s control board out of tolerance. The vendor’s fix required firmware reflashing, but the only technician qualified had shifted to a different division. Operators reverted to manual eyeballing, lowering yields. Finance asked why the plant missed its ROI targets; engineering replied that no budget line existed for “latency between vendor capability and staffing reality.” Multiply that vignette across thousands of plants and the national plateau feels less mysterious. Why did leadership misjudge the tipping point? Partly because balance-sheet optics reward asset accumulation. An upgrade appears as a capital asset with a defined depreciation schedule; headcount, by contrast, is expensed immediately. The asymmetry encourages what one manufacturing CFO once bragged to me as “capex over opex arbitrage.” Each wave of marginal automation thus looks efficient on paper while the supporting human mesh is quietly starved. But arbitrage is not free. By externalizing orchestration cost, shifting it onto an invisible ledger of cognitive load, firms sow future volatility. The Great Recession, a synchronized demand contraction, exposed the fragility at scale: as line-level staffing thinned, recovery lagged behind material demand. The productivity line in the FRED series plunges in 2009 and never climbs back to its previous trajectory, like a sprinter who tears a tendon and thereafter settles for jogging. It could have been different. In the late-1990s a handful of plants adopted a contrasting doctrine: structural simplicity. Rather than layering software atop legacy mechanisms, they redesigned

flow from first principles. At a paperboard mill in Québec engineers collapsed five discreet trim stations into one servo-coordinated unit, eliminating hundreds of feet of conveyor and fifty sensors. The up-front capital was larger than piecemeal fixes, but the resulting control architecture was simpler, documented, and less maintenance intensive. Twenty years on, that mill still posts uptime figures in the ninety-seventh percentile. Case-studies like these, quiet, local, often family-run, reveal an overlooked truth: complexity is a choice, not an inevitability. The question, then, is how to engineer a systemic turn toward simplicity before the demographic wave crests further. A first pillar is valuation reform. Boards must insist that every capital request disclose not only direct ROI but also its incremental coordination coefficient, how many new tags, control loops, learning curves, and vendor dependencies the project injects. Scholars studying complexity costs show that beyond a modest threshold each additional variant can degrade quality linearly and cost exponentially PMC. Translating that science into a standardized surcharge, call it an “entropy tax”, would tilt investment decisions toward elegant designs that embed fewer moving parts per unit of yield. A second pillar is talent reconstitution. No one can conjure a quarter-century of tacit knowledge overnight, but companies can restore apprenticeship as a strategic asset. Pairing retiring masters with cohorts of digital-native engineers shortens the handoff cycle: the veteran explains the physics of a heat exchanger; the graduate captures the logic in a causal model that will outlive them both. Government and industry trade groups already trial such knowledge-capture frameworks, yet uptake remains sluggish because productivity gains accrue over horizons longer than most corporate bonus plans. Realigning incentive contracts, tying executive compensation to five-year resilience indices rather than one-year EBITDA, would close that temporal mismatch. Third, technology itself must mature from data spray to decision scaffolding. Most plant dashboards remain glorified scoreboards: they reproduce all the telemetry that operators ignore manually, simply relocating it onto a luminous screen. What operators need is selective inference: an agentic layer that absorbs raw signals, reasons about causality, and surfaces only what demands human intervention. Prototype causal-AI engines already achieve this for discrete tasks, predicting compressor surging or recommending resin swaps, but integration lags because senior teams mistake advanced math for magic. The real breakthrough is psychological: trusting a system to suppress noise so that humans can focus on levers that matter. When complexity is masked from cognition, effective capacity rises, buying time for the tenure gap to narrow. Skeptics will counter that simplification and reskilling invite their own costs. They do. But the status quo is not free either; it merely invoices society later in the form of shuttered plants, dislocated workers, and supply-chain brittleness. The pandemic amplified that invoice: when oxygen lines failed in ventilator plants or when chip fabs stalled over a single valve misspecification, the national economy paid for decades of under-appreciated orchestration fragility. What looked cheaper on the ledger was in fact a leveraged bet against entropy. One might ask whether reshoring incentives, federal subsidies for semiconductor or electric vehicle capacity, will compound or alleviate the dilemma. Evidence so far is mixed. Green-field megaprojects funded by recent industrial policy often import the same sprawling vendor mosaics

that handicap legacy sites. Yet they also arrive at a historical juncture where AI-native orchestration is feasible. If these projects bake in causal frameworks from day one, they can bypass dashboard bloat. If they import yesterday’s MES logic, they risk locking complexity into concrete for another half-century. Policy makers therefore have a stake. Subsidy packages could mandate entropy audits: a quantitative assessment of coordination burden per unit of throughput. Europe’s aviation regulators already require such risk models for fly-by-wire systems; factories handling volatile chemicals arguably deserve the same philosophical rigor. Public procurement standards, too, can reward suppliers whose equipment ships with open, self-describing data schemas, reducing vendor lock-in and easing cognitive overhead. The cultural dimension may be hardest. Western business lore celebrates scale, optionality, and the audacity of more. Consultants echo that mantra: digital twins on every pump, KPIs on every wrist. Simplicity sounds quaint by comparison, as though advocating for fewer SKUs or fewer alerts jeopardizes modernity. Yet history is unequivocal: civilizations that outrun their coordination bandwidth collapse. The Roman road network was a marvel until taxation and administration costs exceeded the center’s ability to govern; the empire fragmented under the weight of its own infrastructure. Factories, though smaller in territorial span, obey the same thermodynamic logic. If flows grow more entangled than the brainpower that stewards them, entropy wins. The alternative is not minimalism but intentional architecture. In music, complexity delights precisely because it is choreographed, Bach’s fugues weave five voices yet never sound cluttered. Manufacturing can aspire to similar harmony: multiple technologies converging under a unifying score. Achieving that harmony requires elevating system composers, engineers, process scientists, cognitively literate AI designers, to podium positions. Finance is indispensable, but it should be the orchestra’s patron, not its conductor. Let us imagine, for a moment, an industrial Renaissance where every capital review begins with two questions: Does this initiative lower the cognitive burden on our future workforce? If not, does it at least pay the full cost of the burden it creates? Posed earnestly, those questions would recast investment priorities from gadget acquisition to flow mastery. They would privilege intuitive interfaces over redundant HMI screens, modular code over one-off patches, and purposeful data over exhaustive logging. They would treat Emilio’s decades of heuristic wisdom not as a sentimental relic but as an algorithm with concrete asset value. Such a world may sound romantic, yet the economics align. Scholars charting the hidden cost of complexity estimate that eliminating a single menu step from a high-mix assembly line can cut defect rates by seven percent and operator fatigue by double digits PMC. Multiply that across global production and the gains rival the efficiency leap promised by the first wave of automation. In other words, simplicity is not retreat; it is the next frontier of competitive edge. We stand at a crossroads. The demographic exit ramp grows steeper each quarter, and every dashboard we add widens the gulf between signal and comprehension. By continuing on the current trajectory, we consign productivity to the horizontal line it has already drawn, an empire

content with zero sum. By pivoting, re-valuing orchestration, re-investing in human expertise, and reimagining technology as cognitive ally, we can bend the curve upward once again. When Emilio finally retires later this year, the bottle line he shepherds will either be decommissioned or reborn under a new philosophy. His plant manager, newly aware of the hidden cost ledger, has two proposals on the table: a million-dollar retrofit that layers defectdetection AI onto the existing Rube Goldberg machine, or a deeper redesign that merges two stations and halves the sensor footprint. The first option dazzles shareholders with buzzwords; the second promises a quieter form of progress, fewer alarms, shorter training, longer mean-timebetween-failures. Ten years ago the choice would have been obvious, but tenure erosion has a way of clarifying risk. The manager is leaning toward the simpler path. Perhaps that single decision, replicated across thousands of plants, marks the beginning of a different industrial epoch, one where productivity is liberated not by more lines of code, but by fewer, better ones. The arc of American manufacturing has always traced the interplay between invention and discipline, between the engine that can and the human mind that must. Our current plateau is not a verdict of decline; it is an overdue call to realign those forces. Complexity will always tempt, entropy will always lurk, but with the right lens, one that prices orchestration, honors expertise, and elevates causal clarity, we can once again put machines in the service of productivity rather than the other way around. The lesson, as Emilio would say while tightening the last bolt on his watch: “Make it smart enough to work, but simple enough to last.” Sources: 1. Federal Reserve Bank of New York, Liberty Street Economics — “The Mysterious Slowdown in U.S. Manufacturing Productivity” (2024). Liberty Street Economics 2. U.S. Bureau of Labor Statistics — “Gauging the Labor-Force Effects of Retiring Baby Boomers” (Monthly Labor Review, 2000). Bureau of Labor Statistics 3. Moseley et al. — “Product Variety, Product Complexity and Manufacturing Operational Performance: A Systematic Literature Review” (2024). Orbit 4. Ahmed et al. — findings on electronic-record information overload, summarized in “Dealing with Information Overload: A Comprehensive Review” (2023). PMC 5. American Economic Association research brief — “Why Did Productivity Drop After the Great Recession?” (2019). American Economic Association 6. FRED series OPHMFG — “Manufacturing Sector: Labor Productivity (Output per Hour) for All Workers” (updated 2024). FRED 7. U.S. Bureau of Labor Statistics — “Labor-Force Projections to 2024: The Labor Force Is Growing, but Slowly” (2015). Bureau of Labor Statistics 8. Andersen & Hilletofth — “Product Variety, Product Complexity and Manufacturing Performance” (mineral-wool case study, 2018). ResearchGate 9. Imasuen & Lubbe — “Information Overload and the Use of Data Analytics and Visualization Tools in Organizations” (2024). ResearchGate 10. Intereconomics — “Productivity and the Great Recession” (2018). intereconomics.eu #IndustrialTransformation #ManufacturingExcellence #OperationalIntelligence #DigitalTwin #CausalAI #WorkforceStrategy #KnowledgeManagement

#IndustrialComplexity #SmartFactories #Digitalization #TalentRetention #AssetManagement #IIoT #Industry40 #LeadershipInEngineering