This week, Meta has announced the internal testing of Hatch, its latest advanced agentic AI tool, while simultaneously scaling back on previous pressures for employees to maximize AI tool usage. The announcement marks a shift in how Meta is approaching AI integration within its workforce.
Hatch is designed to autonomously perform tasks on a computer, with capabilities such as web browsing and operating other applications. It represents Meta's most advanced AI experiment to date and has been made available for employees to test on corporate devices. The tool is expected to be publicly released in the near future.
Previously, Meta had incentivized employees to utilize AI tools heavily, with performance evaluations linked to their usage. This led to a 'tokenmaxxing' culture where employees felt pressured to maximize their interaction with AI tools to achieve favorable performance labels. However, Meta has now removed these incentives, emphasizing that employee evaluations will focus on overall contributions rather than AI usage metrics.
Hatch operates as a sophisticated agentic tool, similar to other AI models like OpenClaw. While details regarding its technical specifications, such as model size or training data, were not disclosed, employees have reported Hatch consuming more resources than standard chatbots and coding assistants.
The primary users of Hatch are Meta employees, who are encouraged to experiment with the tool for both professional and personal organizational tasks. Despite some privacy concerns regarding the connection of Hatch to personal accounts, some employees have begun using it to manage appointments and other tasks outside of work.
Work implications: Hatch has the potential to streamline workflows by automating routine tasks, which could enhance productivity for roles involving significant administrative duties.
Originally reported by Wired