Meta’s internal AI assistant, Metamate, is being used by employees to help prepare for performance reviews, according to Joseph Spisak, a product director in Meta’s Superintelligence Labs. Speaking at the TechEquity AI Summit in Sunnyvale, California, Spisak shared that the company’s ChatGPT-style tool can scan through employees’ documents, projects, and feedback to generate summaries of their annual accomplishments. “When at the end of the year we do our performance, and I want to summarize my performance or whatever, I call Metamate and it goes and searches all my docs and what I've done and summarizes what I've done for the year and my accomplishments and feedback on me,” Spisak said. “And it's great.” He added a lighthearted note, joking that employees might try to “reward hack it” by attempting to influence the AI’s output, after a moderator asked whether the system could be manipulated to improve evaluations. Beyond performance reviews, Spisak said Metamate is trained on internal Meta data and is used for a range of other work tasks, including application development. The tool is part of a broader trend at Meta and other major tech companies to embed AI into daily workflows. The company has rolled out several AI-powered tools, such as Devmate, an AI coding assistant, and internal dashboards that track how employees are using AI across teams. While some staff members are actively using Metamate to draft performance summaries and provide feedback to colleagues—often by setting up templates and populating them with examples—others report mixed results. An anonymous Meta employee told Business Insider that the AI often struggles without clear context about individual projects, leading to inconsistent or overly generic summaries. Still, the tool is seen as a helpful starting point for organizing thoughts and compiling evidence of contributions, especially during high-pressure review periods. As Meta continues to integrate AI deeply into its operations, the use of tools like Metamate reflects both the promise and the challenges of relying on artificial intelligence for sensitive, human-centric tasks like performance evaluation.