Key takeaways
- Mark Zuckerberg is developing an AI agent to streamline CEO decision-making and reduce reliance on staff.
- Meta is aggressively integrating AI into its operations, aiming to speed up work and reduce organizational layers.
- Meta plans significant capital expenditure on AI infrastructure, data centers, and custom chips by 2026.
- The company's AI push may lead to substantial workforce reductions, potentially impacting thousands of jobs.
- Meta is developing in-house AI chips and internal AI tools to enhance efficiency and compete in the AI landscape.
Meta CEO Mark Zuckerberg is developing an AI agent to support his work as Chief Executive of the company, The Wall Street Journal reported.
The system is designed to surface information quickly and support decision-making.
The agent, still under development, retrieves answers that Zuckerberg would typically obtain through internal channels, reducing reliance on layers of staff, the report added.
Zuckerberg’s agent project reflects a broader push across Meta’s roughly 78,000 employees to speed up work, reduce organisational layers and reshape day-to-day roles to compete with smaller, AI-focused startups.
The company is reportedly considering cutting up to 20% of its workforce, potentially around 15,000 to 16,000 jobs, as it ramps up spending on AI infrastructure and data centres.
Meta sees AI adoption as central to its future and is testing ways to integrate it more deeply into its operations. It has encouraged employees to adopt internal systems such as ‘Second Brain’, which can index and query documents, and ‘My Claw’, which interacts with files and communications.
Meta plans to ramp capital expenditure to roughly $115-135 billion in 2026, largely to fund AI compute, data centres, and custom silicon, as part of an aggressive push to compete with OpenAI, Google, and others.
Meta recently unveiled a new roadmap of four in‑house AI chips under its Meta Training and Inference Accelerator (MTIA) programme, with MTIA 300 already in use and MTIA 450/500 focused on AI inference workloads such as recommendations and chat.
The company recently delayed the public rollout of its next‑generation foundation model, codenamed Avocado, after internal tests showed it still lags behind top‑tier competitors such as Google Gemini 3.0 on reasoning, coding, and writing benchmarks.
