Google released a detailed survey on how Gemini is being used in the workplace, addressing professionals' growing concern about human replacement by artificial intelligence. The results show that the tool is predominantly used to assist with daily tasks, rather than taking over entire functions.
The study analyzed a total of 14.65 million interactions on platforms providing the AI. According to the data, only 3% of these cases involve the massive use of AI in a work context. Furthermore, 75% of the analyzed tasks consist of simple queries to Gemini. An additional piece of data shows that 10% of conversations in work contexts are classified as 'cognitive,' suggesting that solutions are mostly conceived by employees themselves before being executed with AI assistance.
Queries to Gemini show a strong inclination towards personal use, as 86% of recorded interactions relate to activities outside the professional scope. The survey used the ATLAS (Activity, Task, Landscape and Adoption Study) methodology to collect the information.
In contrast to Google's findings, the market has been facing waves of mass layoffs, often justified by the logical transformation of work driven by automation. Data from the Layoffs.fyi platform records over 120 thousand cuts in technology companies alone. More recently, Microsoft announced the dismissal of 4,800 employees, representing about 2.1% of the global workforce. Previously, Amazon had laid off approximately 30 thousand people, claiming that 'the world is changing faster than ever.' Meta has also reported similar situations.
The Google study found no evidence of automations that justify replacing humans with AI agents. Many technicians use Gemini to obtain specific information for their fields, such as support in electrical or construction services. The report highlights that the use of AI tends to be more associated with leadership roles in regions with higher purchasing power, observing a 2.5% increase in AI usage for every 1% rise in salary bracket.
Google concludes that AI acts primarily as a tool, and not necessarily as a means of complete automation. Although the study indicates that 80% of workers in the US market have at least 10% of their work activities influenced by AI, this does not imply an automation of processes that legitimizes the replacement of humans by LLMs. It is important to note that the survey does not cover private models like Gemini Enterprise and Workspace, which may conceal existing automations.
Google has introduced three new artificial intelligence versions with the aim of intensifying competition against OpenAI and Anthropic. These new offerings include faster, more economical models, and those specialized in digital protection.
The main launch is Gemini 3.6 Flash, presented by Google as the company's most advanced Flash model. This model promises performance improvements in programming, knowledge tasks, and multimodal features, while consuming 17% fewer tokens than its predecessor, resulting in reduced operational costs.
Alongside Gemini 3.6 Flash, Gemini 3.5 Flash Cyber, focused on identifying and correcting software flaws, and Gemini 3.5 Flash-Lite, designed to handle large workloads and autonomous AI agents, were launched.
Security is a key differentiator for this new generation of models, according to Google. Gemini 3.6 Flash has been strengthened to offer greater resistance to misuse in cyberattacks without compromising legitimate applications.
Among the important announcements are: Gemini 3.6 Flash, for being more economical and efficient; Gemini 3.5 Flash Cyber, dedicated to digital security; Gemini 3.5 Flash-Lite, intended for AI agents and large-scale tasks; and the start of Gemini 4 pre-training.
In tests conducted by the company itself, Gemini 3.5 Flash Cyber detected 55 issues in the V8 JavaScript Engine, including 10 vulnerabilities that other models had failed to identify. Because of this, its initial access will be restricted to governments and trusted partners.
While the new models reach the market, Google announced that Gemini 3.5 Pro remains in testing and will be made available when ready. Furthermore, the company revealed that it has started pre-training Gemini 4, indicating that the next iteration of the platform is already under development.
In parallel, Gemini 3.6 Flash and Flash-Lite are beginning to be released to developers, businesses, and Gemini application users. Flash-Lite will also undergo gradual integration with Google Search.
In recent months, the competition for increasingly specialized and efficient models has been fierce among Google, OpenAI, Anthropic, Meta, and various Chinese startups. With these new releases, Google seeks to regain ground in a rapidly evolving dispute, where factors such as cost, performance, and security have gained comparable weight to the quality of responses.
