‘Nobody knows anything’ and ‘this time is different’: the phrases that define — and haunt — the AI economy

The article argues that AI’s economy-wide productivity boost is only about 0.1% per year, despite Bank of America’s bullish framing and Goldman Sachs finding no meaningful economy-wide productivity relationship so far. It highlights a theoretical ceiling of 0.66% labor productivity gain today, with adoption slowed by organizational friction and KPI-driven risk aversion. Overall tone is highly uncertain: the piece suggests AI remains transformative in theory, but practical deployment and monetization are still unresolved.

Analysis

The market is still pricing AI like a broad productivity super-cycle, but the article’s core implication is that near-term monetization is bottlenecked by enterprise process change, not model capability. That favors vendors selling picks-and-shovels into experimentation and deployment — cloud, data plumbing, security, workflow software, and systems integration — while pressuring pure-play “model” economics if buyers keep internalizing pilots without scale. The second-order effect is that capital spending can remain elevated even if realized productivity stays muted, because CIOs will spend defensively to avoid being left behind.

For BAC and GS, the read-through is more about client behavior than direct earnings exposure. A prolonged “no playbook” environment tends to support advisory, capital markets, and structured finance tied to AI capex, but it also delays the broad productivity thesis needed to justify a step-function re-rating in cyclicals and labor-sensitive sectors. If AI adoption stays concentrated in a few functions for 6-18 months, the equity market may keep rewarding narrow winners while punishing companies that overhire or overpromise margin gains from AI.

The contrarian point is that consensus may be overestimating how quickly enterprise inertia breaks, but underestimating how quickly incumbent software and services vendors adapt by becoming the de facto deployment layer. If AI companies are forced to build consulting arms, that is evidence of a large services market emerging around model implementation; the economic rent may accrue to firms that own workflow distribution and implementation trust, not the model layer itself. That creates a slower-burn winner set than the market’s current “take-all” framing.

Watch for catalysts over the next 1-2 quarters: enterprise budget season, any evidence of measurable ROI in customer support/software, and commentary from banks about AI-related transaction activity. A disappointment in broad productivity data would likely hit the most expensive AI-adjacent names first, but would be constructive for software/integration vendors that monetize deployment friction. The key risk is a sharp sentiment unwind if capex accelerates without visible revenue lift, compressing multiples across the theme.

Key Decisions for Investors

• Long IBMM / ACCN / EPAM-style deployment beneficiaries vs short a basket of unprofitable AI application names over 3-6 months; thesis: implementation friction monetizes services faster than models monetize productivity.
• For BAC and GS, stay tactically neutral but bias long into any selloff in AI-capex headlines over the next 1-2 quarters; the higher-probability revenue stream is advisory/financing around deployment, not operating leverage from economy-wide productivity.
• Buy 3-6 month put spreads on the most crowded AI beneficiaries with stretched multiples; risk/reward improves if enterprise ROI data stays soft and the market re-prices the timing, not the end-state, of adoption.
• Accumulate quality enterprise software on weakness, especially names embedded in workflow and security, for a 6-12 month horizon; the market is mispricing the toll-booth effect of AI rollout versus the end-productivity story.