Revenue, employment improve with business AI use: MTI report

TEXT START: [SINGAPORE] Workplace use of artificial intelligence is linked to quick gains in revenue and employment, though not yet to higher productivity and profit, a new study has found.

The Dissection

This is an early-adoption correlation packaged as economic reassurance. The article foregrounds 16 per cent revenue growth and 8 per cent employment growth, but its decisive finding is the opposite of the headline mood: businesses saw no statistically significant increase in productivity or profit during their first four years of AI adoption.

That is the complementarity phase. Firms are still hiring people to redesign workflows, manage implementation, interpret outputs, expand capacity, and supply specialized expertise. The gains are also concentrated among higher-income, mid-career workers and foreign Employment Pass holders. This is not broad labor empowerment. It is selective elevation of workers closest to capital and selective recruitment of scarce expertise.

The report’s own evidence reveals a narrowing labor market disguised as expansion: AI-using firms grow, but the valuable positions increasingly cluster around workers who can operate, verify, or control AI systems.

The Core Fallacy

The text mistakes lag-phase co-movement for a structural reversal.

Revenue can rise while employment temporarily rises alongside it. A firm may use AI to expand sales before it automates the labor required to produce and administer that expansion. Net employment can also increase while specific occupations are destroyed, entry-level pathways collapse, and bargaining power shifts toward owners of the systems.

The Discontinuity Thesis asks whether AI preserves economically necessary human labor once cognitive automation becomes durable and institutions cannot maintain human-only domains at scale. This report does not test that question. Its data ends in 2024, before widespread agentic AI, and its strongest evidence concerns firms still building the infrastructure around AI.

“Human judgment” is treated as a durable moat. Under P1, it is merely a higher-value automation target. Under P2, reskilling and policy cannot guarantee a permanent human economic niche. Under P3, the current employment increase can be the prelude to productive participation collapse, not evidence against it.

Hidden Assumptions

• Early complementarity will persist after AI becomes more autonomous and capable.
• Workers displaced from routine functions can be reskilled or redeployed at the necessary scale.
• New, better-paid roles will remain numerous enough to absorb the workers released from old ones.
• Aggregate employment growth represents durable, broadly distributed security rather than temporary expansion.
• Higher revenue and wages will translate into worker power, despite ownership remaining concentrated.
• Government support can preserve labor demand rather than merely redistribute income after labor demand disappears.
• AI-related job postings accurately identify causal AI adoption rather than larger, faster-growing firms that would have hired anyway.
• “Human-hybrid workforces” will remain human-led instead of becoming a transitional layer before AI agents take over coordination itself.

The article also treats net employment as the relevant metric. It does not establish job quality, permanence, hours, insecurity, occupational destruction, wage compression below the favored bands, or who captures the resulting surplus.

Social Function

Primary classification: transition management. Secondary classifications: partial truth and ideological anesthetic.

The report contains real evidence: AI currently complements labor in several sectors, especially during implementation and expansion. But the narrative converts a temporary phase into a reassuring trajectory: adopt AI, and employment will follow. That message encourages firms to accelerate adoption while discouraging workers and policymakers from confronting the ownership problem.

The line about supportive policy producing broader wage gains is an unsupported bridge. Transfers may preserve consumption. They do not restore productive participation or prevent the owners of AI capital from capturing the system’s surplus.

The article is not pure propaganda because it includes the critical warning that cheaper replication of human functions can cause displacement even as revenue rises. That warning is the structural truth. The optimistic framing is the anesthetic wrapped around it.

The Verdict

Accurate as a snapshot; misleading as a forecast. The article captures AI before its terminal phase: revenue rises, hiring rises, productivity remains unproven, and gains concentrate among workers nearest to capital.

The 8 per cent employment increase is not proof that the post-WWII employment–wage–consumption circuit survives. It may be the installation surge for machinery that later removes the need for much of the labor being hired. P1 is not defeated; it is still advancing. If P1 and P2 hold, P3 follows. This is not the system recovering. It is the system expanding around the machine before the machine becomes the system.