Torsten Slok, the chief economist at Apollo Global Management, recently published a finding that should terrify any manufacturing executive who has signed off on recent technology budgets. Looking at what he calls the S&P 493, which excludes the seven largest tech giants, profit margins and revenue per worker remain completely flat. Despite massive spending across corporate America, Slok concluded that the technology has delivered zero financial payoff outside of Silicon Valley.

If you are an operations leader wondering why your own software pilots have failed to move the bottom line, the answer is simple. Generic tools cannot replace human judgment in complex operational workflows. To secure a real financial payoff from your investments, you need targeted execution that eliminates manual friction on the shop floor rather than chasing generic productivity hype.

The S&P 493 Reality Check: Billions Spent, Flat Productivity

Apollo Global Management chief economist Torsten Slok exposed a hard truth for industrial leaders. Examining the rest of the market outside Silicon Valley, profit margins and revenue per worker remain flat despite heavy enterprise software budgets. As Slok bluntly stated, “Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech.”

That gap explains why factories deploying generic software tools miss their targets. Slok noted that intelligence is mostly judgment, and systems struggle with trust, unwritten rules, and ethical tradeoffs. Manufacturing operations require strict adherence to quality and real accountability. Generic tools cannot navigate those constraints without targeted workflows designed for the factory floor.

Why Calculators Don’t Run Factories: The Limits of Raw Intelligence

Judgment versus computation in manufacturing

Most industrial technology budgets are built on a faulty premise. Software vendors treat manufacturing challenges as math problems that can be solved with faster processing power. When operations teams buy generic tools, they get high speed calculation without any underlying grasp of the physical shop floor. The technology can process scheduling data instantly, but it lacks the contextual reasoning needed to handle a sudden machine breakdown or a supply chain bottleneck.

As Apollo economist Torsten Slok pointed out in his recent market analysis, raw intelligence fails when applied to complex operational environments. He observed that AI works well as a calculator, but business success relies on human judgment. Manufacturing operations require constant trade-offs between speed, safety, and material costs that pure computation cannot resolve on its own.

AI may be a great calculator, but in business, intelligence is mostly judgment.

Why unwritten rules and operational trust break generic AI

Factory floors run on unwritten rules, tribal knowledge, and deep operational trust between shifts. A plant manager knows which assembly line tolerates a specific variance in raw materials based on years of physical experience. Off the shelf software ignores these cultural nuances. It optimizes strictly for the metrics fed into its algorithms, which routinely damages product quality and strains team coordination.

Trust breaks down entirely when workers realize the software does not understand accountability or ethical trade-offs. If an automated system pushes an unsafe speed adjustment to hit a daily throughput target, the human supervisor takes the risk. Generic models do not carry reputational or operational liability. Achieving a true financial payoff requires embedding automation into targeted workflows where human oversight and accountability remain fully intact.

The Misconception of Adopting AI for AI’s Sake

Most industrial enterprises buy software because of a vague fear of falling behind competitors rather than a specific operational bottleneck. Software vendors push generic assistants and broad automation platforms that promise to transform the entire enterprise at once. This horizontal approach leads straight to the stagnation highlighted by Apollo Global Management.

When leadership implements technology without a specific operational target, employees end up with tools that do not fit their daily routines. Quality managers and floor supervisors do not need general purpose intelligence. They need narrow solutions that solve specific scrap rate issues, maintenance scheduling delays, or compliance documentation errors.

“Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech.”

Software adoption requires a strict focus on cost reduction or throughput improvement from day one. If an implementation project cannot point to a specific metric on the factory floor that it intends to move, it is an IT science project. Industrial executives must stop treating software as a corporate accessory and start treating it as heavy machinery that needs to earn its keep.

Apollo economist Torsten Slok recently highlighted this divergence by looking at the S&P 493. Outside of the mega cap technology firms building the infrastructure, the rest of the market has seen virtually no margin expansion driven by these tools. That disparity exposes a fundamental truth about operational technology. Buying a subscription to a general intelligence platform does not automatically change a P and L statement.

Operations leaders frequently make the mistake of measuring success by adoption rates rather than cash flow. They track how many employees logged into the new system or how many prompts were submitted during the month. Those metrics measure activity, not productivity. A plant manager might see high engagement with a general writing assistant, but if that tool fails to reduce labor hours spent on shift handovers or lower raw material waste, the investment is a net loss.

Achieving a true AI financial payoff requires embedding algorithms directly into existing physical and digital workflows where margins are bleeding. Consider maintenance operations. A generic chat interface forces a technician to stop working, open a browser window, type out a prompt, and interpret a generalized answer. That friction kills productivity. By contrast, a specialized model integrated directly into the computerized maintenance management system can automatically analyze vibration sensor data, order the correct replacement part before a motor fails, and generate the work order without human intervention. The technician simply receives a notification to execute the repair.

That level of integration demands rigorous process mapping before a single dollar is spent on software. Teams must identify the exact friction points in their current procedures where human hesitation or manual data entry slows down output. Only then can engineering teams build or configure machine learning models that address those precise bottlenecks. Treating technology as a targeted surgical instrument rather than a blanket solution separates profitable automation from wasted capital expenditure.

Turning Intelligence into ROI: What Operations Leaders Must Do Instead

Targeting manual bottlenecks in quality and operations

Most industrial technology implementations fail because they target the wrong layer of the enterprise. Operations teams waste weeks trying to deploy broad enterprise software that promises total transformation without fixing a single specific workflow. Real operational efficiency comes from identifying discrete friction points where manual data entry or redundant checks slow down the production line.

Quality managers should start by examining inspection logs, shift handovers, and compliance documentation. These processes rely heavily on human repetition that creates delays and introduces errors. By mapping the exact steps required to clear a batch or log a defect, you find the exact spots where targeted automation removes friction.

AI may be a great calculator, but in business, intelligence is mostly judgment

That observation from Apollo Global Management highlights why broad enterprise tools stall out. When you apply technology to well-defined operational bottlenecks instead of vague company goals, you stop treating software as a general calculator and start using it as an operational tool.

Measuring tangible ROI through bandwidth and error reduction

Measuring the impact of your technology investment requires looking beyond abstract productivity metrics. Factory executives need to track concrete indicators like hours saved per week on compliance reporting and reductions in scrap rates caused by data entry mistakes. When floor supervisors spend less time moving numbers between spreadsheets, they redirect those hours toward shop floor supervision and process improvement.

Financial payback appears when labor hours shift from administrative overhead to direct value creation. If a tool saves an engineer ten hours a week on manual documentation, that reclaimed time translates directly into higher throughput and lower operating costs. Keep your tracking focused on these operational realities rather than theoretical efficiency gains.

Set clear baselines before deployment and review the numbers thirty days after rollout. If error rates in quality logs remain flat, the implementation missed its target. Practical execution means adjusting the workflow immediately until the metrics show genuine improvement in daily operations.

Source: techcrunch.com