Google Cloud this week published a post on how Latin American small and medium businesses (SMBs) are using AI. The headline number: the number of Latam SMBs using its AI tools grew 8x year-over-year, with Brazil at 9x. The post is timed to the Google Cloud Summit in Brazil and closes with nearly 20 customer cases.
Two details stand out. First, SMBs account for over 60% of Latam employment (UN data) — they are the lifeblood of the region's economy. Second, Google repeatedly pushes a per-task model-selection strategy — "cheaper models for summarization, larger models for complex analysis." Among the closing cases: Argentine ad-tech firm AdGoat processes 10 billion ad requests a year using the Gemini API for automated content analysis and bidding; Brazilian dental device manufacturer Angelus uses Gemini Enterprise for customer support.
What this is
In essence, this is marketing content timed to the summit. Gemini Enterprise is Google's flagship enterprise AI platform, positioned to let non-technical employees build AI Agents — programs where AI autonomously executes multi-step tasks. The signal Google wants to send: AI is penetrating the SMB market faster than expected, and they picked Google.
Industry view
Supporters read this as an early signal of AI democratization. The argument: SMBs carry no legacy IT baggage and skip the "document your processes" phase faster than large enterprises, going straight to AI handoff. "Model selection by task" is also seen as a more pragmatic deployment path than defaulting to the most powerful model.
Skeptics focus on the data source. The 8x and 9x figures are self-reported by Google, with no third-party validation. If last year's base was only a few hundred customers, an 8x jump is still small in absolute terms — the story reads more like victory on a low base. Others note that low-code platforms — no coding required, apps built by drag-and-drop — like those enabling "non-technical staff to build Agents" have a contested track record in enterprise IT. The Latam cases showcase things that "run," but running is still a long way from delivering real business value.
Impact on regular people
For enterprise IT: The Latam SMB "model selection by task" playbook applies to Chinese SMBs as well. Cheap models are enough for classification, summarization, and simple Q&A; reserve the heavy model for serious analysis. You don't need to burn the most expensive compute on every task.
For individual careers: Google's message of "letting non-data analysts use data" means business roles that don't actively learn AI tools will fall behind IT roles in understanding their own operations.
For consumer markets: Advertising and customer service are being reshaped by AI in bulk. Behind AdGoat's 10 billion annual ad requests is a large volume of optimization decisions that used to be made by humans, now automated.