Brazil's small businesses embrace AI to boost productivity - Valor International

TEXT START: More than half of Brazil’s small businesses already use artificial intelligence (AI) in some way daily, particularly in areas such as marketing, customer service, and inventory management.

The Dissection

This is an early-adoption report packaged as labor-market reassurance. It documents 52% usage among surveyed small businesses, 60% planning to adopt or expand AI, and self-reported gains in marketing, customer service, inventory, contracts, maintenance, and expense analysis.

Its central rhetorical maneuver is to treat the absence of immediate layoffs as evidence that AI changes jobs rather than destroys them. That inference is structurally worthless. The lead’s claim of daily use also outruns the reported metric: usage during the preceding two weeks does not establish daily usage.

The article assembles Sebrae, ministries, a business owner, and an economist around a benign narrative: AI removes drudgery and upgrades workers into strategic, critical-thinking, interpersonal roles. Productivity uncertainty is mentioned, but only as a reason to postpone the alarm. The ownership question—who controls the systems and captures the gains—is absent.

The Core Fallacy

No layoffs during initial adoption does not equal no eventual displacement. AI first allows existing staff to handle more output, customers, and administrative load. Once competitors adopt the same tools, that advantage becomes a baseline. Firms then need fewer human-hours per unit of output, producing hiring freezes, attrition, lower wages or hours, consolidation, and eventually explicit cuts.

The relevant measure is not whether an owner fired someone this quarter. It is whether the same economic output can be produced with fewer human hours over time. The article measures headcount, not labor demand, wages, hours, hiring, output per worker, or the counterfactual number of jobs that would have existed without AI.

The salesperson-versus-truck-driver comparison is another weak shield. Job-title complexity is not a moat; automatable task content is. Critical thinking and interpersonal interaction do not remain human property merely because institutions label them valuable.

Hidden Assumptions

• Productivity gains will create enough additional demand to absorb displaced labor.
• Employment loss appears mainly as immediate firing rather than reduced hiring, attrition, wage compression, or reduced hours.
• Jobs are indivisible bundles instead of collections of tasks that can be progressively automated.
• More complex occupations are protected because their human skills cannot be simulated, decomposed, or scaled by AI.
• Small-business AI adoption will democratize productivity rather than intensify price competition and concentrate ownership.
• Self-reported productivity gains are causal, comparable, and measured against a credible counterfactual.
• Workers who become more fluent with AI gain durable bargaining power, rather than becoming more efficient Servitors under someone else’s control.
• Government training and financing can alter the ownership structure of AI capital rather than merely accelerate its diffusion.

Social Function

This is partial truth, transition management, and ideological anesthetic. The survey may accurately capture current adoption and current headcount stability. It then converts not yet into not at all, normalizes labor substitution as worker upskilling, and shifts responsibility onto individuals while leaving capital ownership untouched.

The Verdict

The article does not refute the Discontinuity Thesis; it records its lag phase. Small firms are using AI to expand capacity before competitive saturation forces labor compression. The 90% figure is a snapshot of delayed impact, not evidence that the wage-employment-consumption circuit is structurally safe.