Usage is moving toward production
OpenAI published country-level data on how people use ChatGPT, describing a shift from asking questions to producing outputs and completing tasks. The company says people are more than twice as likely to use ChatGPT for “doing” at work—activities such as writing, coding, editing, or analysis—than outside work. It also reports that parts of Latin America, Africa, and Oceania are catching up with earlier adopters.
The dataset comes from OpenAI and reflects messages in Free, Go, Plus, and Pro accounts generally managed by individuals rather than organizations. It is vendor-produced usage research, not a survey of business outcomes, and it does not prove productivity, quality, or financial return. It does, however, show how quickly individual behavior can change before formal company processes catch up.
What happened
OpenAI says its Signals data covers country-level patterns among more than one billion people putting ChatGPT to work. In its Q2 2026 rankings, Peru, Uruguay, and Costa Rica rose the most in messages-per-capita rank. The publisher also says multimedia became the fastest-growing use case, reaching 7.8 percent of messages globally and more than one in ten messages in Brazil and Colombia.
Usage among people over 35 increased in almost every country in the age-classified analysis. OpenAI notes that age is self-reported and that these users accounted for a five-percent higher share of messages than twelve months earlier. The data therefore suggests broader adoption by geography, format, and age, but it should not be confused with audited workplace deployment.
For employers, individually managed accounts are the key detail. Staff may already use AI for customer replies, spreadsheet analysis, translation, proposals, code, and internal documentation. The organization may receive the output without seeing the prompts, source data, model version, or corrections.
Why it matters for growing businesses
Shadow workflow is more important than shadow software. A company may approve a familiar AI service but still have no idea how employees use it inside material processes. One person drafts routine text; another uploads a customer export; another relies on an analysis to set inventory levels. The service name is the same, but the risk and control requirements differ.
Management should inventory business workflows rather than only licenses. Ask where AI output enters CRM, ERP, accounting, customer support, HR, software, or public content. Identify the authoritative record, reviewer, error consequence, and fallback. A low-risk brainstorming task needs lighter controls than an output used to communicate a price, interpret a contract, or post financial data.
Latin American adoption also reinforces the need for multilingual and local testing. Spanish and Portuguese workflows must be evaluated with actual terminology, tax documents, customer expectations, and regional data rules. A system that performs well in generic English examples may misread abbreviations, currency formats, or local legal context.
Turn individual use into a governed pilot
Choose one recurring task where employees already use AI and where the company can measure the full result. Proposal preparation is an example: approved service descriptions, pricing rules, and customer context can be assembled from governed sources; AI drafts a structured proposal; an employee verifies commitments; and the final version is stored with the opportunity in Odoo.
The pilot should capture baseline time, review effort, corrections, conversion, and customer issues. It should also control input data, approved models, retention, access, and logging. If the workflow cannot reproduce why an output was accepted, it is not ready for broader automation.
Avoid measuring prompts or messages as business value. More usage can indicate adoption, confusion, or rework. Measure cycle time, quality, capacity, revenue, cost, and exceptions at the process level. Include the reviewer’s time and the cost of correcting errors.
Practical action checklist
- Survey teams for real AI-assisted tasks, including use through individually managed accounts.
- Classify each workflow by data sensitivity, decision consequence, external impact, and reversibility.
- Define approved services and prohibit confidential or regulated data in unsuitable accounts.
- Keep customer, product, order, invoice, and operational facts in Odoo or another governed record.
- Require source checks and accountable human approval for material outputs and commitments.
- Test Spanish, Portuguese, French, and local terminology with representative regional cases.
- Log the workflow, model or service, source records, reviewer, correction, and final outcome.
- Measure end-to-end time, error, rework, customer impact, cost, and business result.
- Provide a safe route for employees to disclose current usage without punitive assumptions.
- Expand only workflows that show repeatable value and manageable exceptions.
How VDS Consultoria can help
VDS Consultoria helps businesses convert informal AI use into controlled operational improvement. We can map current workflows, classify risk, connect approved tools to Odoo, define human review, test multilingual cases, and build value and exception dashboards. We also help consolidate duplicate services and establish a lightweight approval process for new AI features.
The shift from asking to doing is an opportunity, but it changes the management question. Leaders should no longer ask only whether employees use AI. They should ask which business process changed, what evidence supports the output, who remains accountable, and whether the result improved.
Source
OpenAI, “From asking to doing: How the world is putting ChatGPT to work,” August 6, 2026: https://openai.com/index/how-the-world-is-putting-chatgpt-to-work/